Maybe I should just rename this section Loot, period. I only posted once in February. But then again, February is a very short month compared to the others… (yeah, good excuse)

This time, lots of stuff in my history, I hope you have some time ahead of you. Otherwise, good fodder for your “read it later”.

Quick service announcement first

The articles on Godzilla 2 and 3 are in progress, but I’m struggling with clip extraction. Especially with subtitles. I haven’t found, with Shotcut, a solution to simply extract parts of a video while keeping subtitles synchronized. Until I figure that out, no article 😞 so I’m falling a bit behind.

I’m open to ideas and alternative solutions, free/open-source ones. (no Adobe or the like) It shouldn’t be too complicated to do either, since I’ll have to do it for almost every movie. Ideally, the subtitles of the extracted clips should not be burned into the video but kept as separate files. If you have suggestions, contact me.

Society

Le code a changé, 3 episodes about potentially catastrophic legislative changes.

• Omnibus : la directive européenne inquiétante 1/3 (Podcast) (fr)
• L’Amérique, colon digital 2/3 (Podcast) (fr)
• Histoire d’un lanceur d’alerte 3/3 (Podcast) (fr)

Nothing new under the sun regarding the digital vassalization of Europe, but good reminders and connections with the past. I really like the comparison between the ToS we accept without reading, and European colonizers upon their arrival in the Americas who had indigenous people read and accept long, impossible-to-read texts.

Concretely, Omnibus aims to “simplify” regulations that are clearly not being enforced. Honestly, saying that the GDPR is complicated is truly an asshole take. It’s one of the simplest and shortest pieces of regulation out there. No, you have to be an asshole to try to overcomplicate the thing in order to bypass it. Because if you’re not an asshole towards your customers or your visitors/users, then the GDPR isn’t hard, and you don’t need cookie pop-ups or anything like that. And when, in the podcast, he keeps saying the GDPR has flaws, honestly he’s only talking about these asshole moves, not real flaws. Disappointing that he repeats it so many times for free. It’s not the GDPR that imposes banners, that forces dark-patterns and organizes the deception of visitors/customers. It’s the companies, for the sake of surveillance capitalism. They participate in it, like collaborators!

On top of that, the “simplifications” they want to make won’t help our companies at all. It’s only for the American tech giants, to leave more room for workarounds, and add even more legislative instability.

• Sondage de DPO à l’appui, noyb affirme que la simplification RGPD est à côté de la plaque (fr)

Speaking of surveillance capitalism and fascism, I learn that Microsoft is being a jerk again (oh surprise), and that lots of games are getting striked on Steam, probably unjustly too. Personally I’m opposed to the concept of intellectual property protection (a topic for another day), so it makes me even angrier that anyone would prevent other people from creating under the pretense that it’s “too” close to what they own (and didn’t even create)! In short, Microsoft is pulling a Nintendo, as usual.

• Microsoft’s AI-Powered Copyright Bots Fucked Up And Got An Innocent Game Delisted From Steam

Internet

• Doit-on confisquer les smartphones des vieux ? avec Audrey Randriamandrato (YouTube) (fr)
• Vérification d’âge (1/4) : vers un internet de moins en moins sûr (fr)
• Vérification d’âge (2/4) : de l’impunité des géants à la criminalisation des usagers (fr)
• Vérification d’âge (3/4) : la panique morale en roue libre (fr)
• Faut-il interdire les réseaux sociaux aux jeunes ? (Article + Book) (fr)

After reading and listening to all of that, I think you’ll have a good idea of what’s coming and what’s problematic. The book I mentioned is very interesting and is filled with real-world examples and cases to back up the arguments. It brings nuance, even on the side of “pro censorship” folks, which makes it easier to grasp the topic with nuance.

With agent orange in power, the true face of the USA is more visible than ever. Not that they were nice political partners in the past, but at least nobody can deny anymore what is clearly visible where before it was filed under conspiracy theories. And that despite the multiple revelations with proof that these conspiracy theories were indeed true!

There are the wars, the real ones some would say, that kill with bombs and bullets. There is also the digital war, which also causes harm, sometimes even deaths. Without bombs, without bullets. But worse, this one ensures US control over all our lives, Europeans but also everywhere in the world. This war started with the advent of the internet, so it’s nothing new. Many people have been warning, educating, and successfully predicting the interference and abuses enabled by our legislation and our mandatory American infrastructure.

AND NOBODY GIVES A DAMN!!! Réveillez-vous !!! (wake up!!!)

Queer & Anti-racism

Change of tone, for a refreshing novelty. A queer show on the media outlet “Histoires Crépues”. Queer content connected to anti-racism, feels good.

• L’amour entre femmes est-il vraiment plus simple ? - Kwir I DI (YouTube) (fr)

In the gay community, on the male side since I know it better, it’s a real nightmare when it comes to racism and all sorts of problematic behaviors. You’d think fags would be more inclined to understand the principles of domination and discrimination, being victims themselves; but no, nothing. It’s even the complete opposite and it’s exhausting in the long run. It reminds me of a very good episode of the podcast “Les couilles sur la table” if you want to learn more about this topic. I can confirm almost everything that’s said in it, having either seen or lived it. Even though I’m far from having a full bingo card, I’ve checked enough boxes to be convinced of the truthfulness of all of it, and how widespread it all is.

• Grindr ou la culture du plan cul | 1/2 (Podcast) (fr)
• Grindr ou la culture du plan cul | 2/2 (Podcast) (fr)

Coming back to the show Kwir I DI, it feels good and I hope to see more episodes in the future. I don’t know of any with non-white men, I’d like that too I think. Especially if it’s just as chill. Because the shows I know of are all straight and very white unfortunately.

To continue on the anti-racist theme, I recommend this video, especially to your friend who keeps telling you that anti-white racism is a problem.

• Le racisme Anti-Blanc selon Naïma Moutchou (YouTube) (fr)

And I warmly recommend these two videos, still on the topic of racism:

• L’histoire du racisme - Partie 1 (YouTube) (fr)
• L’histoire du Racisme - Partie 2 (YouTube) (fr)

You’ll then clearly understand the difference between racism and xenophobia; why capitalism is racist by nature (or almost), and that a world without racism did indeed exist before.

Android

I’m relaying the open letter to keep Android open:

• Keep Android Open

A less open, or not open at all Android means even more US control over Europeans!

GAFAM

I haven’t used WhatsApp in a while. Not that I was thrilled to use it before, it was imposed on me. Some people still try to reach me on it, and complain that I don’t reply, or that I’m no longer reachable. Too bad. If you want to reach me, send an SMS, an EMAIL, or call me. All of that works perfectly fine and is available to EVERYONE.

If you wish, I use other communication networks, but those tend to shrink more and more because they’re all problematic in one way or another. The next one likely to go on my end is Discord.

On this topic, I’d like to share this article:

• WhatsApp chez les militants? coup de gueule ! (fr)

It approaches it from another angle, one that is important to understand and keep in mind.

On a related note, here’s a good article about the manipulation carried out by our dear GAFAM on ourselves:

• Attention à l’économie de l’attention (fr)

Science

• Your Brain Is the Secret Designer of the Universe and Here’s the Mind-Blowing Way Your Consciousness Shapes Reality

I’m not yet sure where to place this site on the scale of relevance or legitimacy. For now I rather trust it, like I would a magazine like Epsilon. So decent, but not to be taken at face value. Disclaimer over, I find that this research, if it turns out to be confirmed, gives an interesting perspective on what our consciousness is, and connects well to the debates around AI and specifically AGI. Artificial General Intelligence, the ultimate boss of AI.

I recommend reading the article, and I won’t share more of my opinion on the topic, as I’m far from having it well formed (my opinion).

You hear a lot about it, often negatively because it’s used by scamfluencers to become beauty standards, but it’s probably one of the most important recent medical scientific revolutions. On top of that I’m very happy to now know how it works.

• The Uncomfortable Truth About Ozempic (YouTube)

Architecture

I had shared the first one with you, here’s the second in the series. So it’s also likely that I’ll share the third one when it’s available 😛.

• SOLARPUNK: la récupération d’une utopie écologique (YouTube) (fr)

I’ve personally always associated “SOLARPUNK” with something positive. An optimistic vision, one we could possibly strive towards. Because it’s often beautiful and makes you want to live there. Well, even though I wouldn’t want to live there myself, I also really like brutalist and cyberpunk aesthetics; so appeal is probably not a reliable metric. And that’s clearly what this video confirms. I wasn’t ready to see solarpunk being so close to fascism and techno-bros. Yet it was all already right before my eyes, but I was probably too blinded by the genuine appeal of the solarpunk idea.

Trains

Fellow rail enthusiasts, a super interesting video about one of the most unique metro networks. A little treat to watch, to once again realize just how good the Chinese are at planning.

• Ce qu’on ne vous a jamais montré sur ce réseau de métro ! (Les Secrets du Métro de Chongqing) (fr)

AI

Yeah, there isn’t a week without some scary novelty/reflection/analysis about AI. The bubble is still here, it’s wobbling a bit but with the consequences of agent orange, hard to see it bursting anytime soon.

• The job losses are real — but the AI excuse is fake

This first article is a bit of a reminder that AI doesn’t necessarily (understatement?) create new behaviors. Capitalism does just fine on its own. AI is just the trendy excuse to try to mask reality. Good article that reminds us of all that.

• The Eternal Promise: A History of Attempts to Eliminate Programmers

In the vein of “AI isn’t revolutionizing anything”, a good reminder of the facts about our history in software development, before the name even existed. It’s both reassuring and scary. Because despite everything, the countless articles promising the elimination of the developer role are omnipresent and weigh on morale. This one counterbalances a bit. However, once again capitalism is killing us even more. I like the optimistic side of the analysis. Each step leads to more and better despite everything, without erasing what exists or the past, consolidating it further instead.

• Open Source After the Extraction
• GitHub Is Thinking About Killing Pull Requests

The harmful consequences are very much here and visible. The revolution is causing harm, to people and to projects. Open source collaboration, as we know it, may be living its last moments. Either we manage to counterbalance the effects of AI on the commons, or we’ll have to find a new way to create these commons while coexisting with the consequences of AI. For now the reactions are more about protection and closing ranks. Examples abound. The closing of contributions, and of certain bug bounties are good ones. Some try to manage differently, like Gnome redirecting almost all its git traffic to GitHub to benefit from free hosting. (without leaving its own forge) But how long will these “free” solutions remain viable (assuming they even are right now)?

• An AI Agent Published a Hit Piece on Me

And to top it all off, terminator AIs are creeping in. And basically the only answer we get to why? is an Oops, sorry. That wasn’t my intention! It would be good to revisit the basics then. Here are a few videos that are NOT recent on the topic:

• Deadly Truth of General AI? - Computerphile (YouTube)
• General AI Won’t Want You To Fix its Code - Computerphile (YouTube)
• AI “Stop Button” Problem - Computerphile (YouTube)

The full playlist

And when AI isn’t doing harm online, its infrastructure takes care of it in the real world. There’s a topic that seems to fly under the radar. The question of noise pollution, and specifically infrasound and its effects on health. This video explores the subject, and the rigor as well as the scientific references make the problem credible.

• Datacenters Behaving Like Acoustic Weapons (YouTube)

This challenges my enthusiasm for datacenters built in urban areas. Which is a good idea on paper, especially to use the heat produced for heating and hot water. Reducing heat losses by putting them to use. But has the question of noise pollution, even inaudible, been taken into account?

To sum up, here are the concerns I know of regarding datacenters:

• Energy consumption is becoming a problem (because of AI)
• Water consumption is sometimes a concern (a disgrace, cf one of my previous Loots)
• Infrasound production is potentially a concern that turns them into an AoE (Area of Effect)

Worth keeping tabs on, if I find more info on this.

At this rate, you’ll end up thinking I’m against AI and that it’s absolute evil. A bit of balance with this positive testimony on the use of LLMs and what they make possible. To take a step back on the topic.

• Ancient code, mental health, and AI tooling

And that’s a wrap! Lots of Loots this “week” 😉😉.
Bisoux 😗

Pretty cool concert. GOGO PENGUIN as you’d expect. The venue’s quality was really a huge plus. The music quality brought out by top-notch sound, as always over there.

GoGo Penguin:

• Chris Illingworth - Piano
• Nick Blacka - Double bass
• Jon Scott - Drums
• Joe Reiser - sound

© Jon Shard / Nicky Rybka

Opening act: Daudi Matsiko

• Daudi Matsiko - Guitar, voice

Society

• Y a-t-il un instinct paternel? (fr) (Youtube | Podcast)

I’ve personally already read and listened a bit on the topic, but this time I find the many scientific references making the argument more solid and put many preconceptions back in their place.

• Pourquoi on vous ment autant sur le Japon (et pourquoi vous y croyez) (fr) (Youtube | Podcast)

Related to the loot about China, a video explaining why our media say nonsense about Japan, and probably by extension about all Asian countries.

China

• Le Soleil Rouge réveille l’Occident (fr) (Youtube)

A short video that puts prejudices back in their place, again. A must-watch.

AI

• Why do AI company logos look like buttholes? • VelvetShark

Fun and serious at the same time. Nothing to add 😁

USA, Fascism, and renewable energy

• You are being misled about renewable energy technology. (Youtube)

As usual, Technology Connections makes a very interesting video, and even though the topic isn’t as quirky as those two 2-hour videos about how dishwashers work, it’s very relevant and well handled here. I really like how he puts orders of magnitude back into perspective to show the stupid things we do as a society.

But I’m sharing this video here not for its main topic, but for its more unusual part about the issues the USA is facing with the consequences of the fascism in place. He steps out of his “duty of neutrality” I’d say. Not that he has one, but it’s like a trait shared by many content creators to especially not talk politics. Even though they do anyway without realizing it. No, here he takes his responsibilities, takes a risk (facing the harassment he will probably endure), and says things clearly. Clearly enough to give precise voting instructions, without hiding the “imperfections” of such choice. With everything happening in the USA right now, it’s nice to see this kind of content. It’s a reminder that people are opposing what’s happening, as BLAST also reminded us in their video FACE À L’EXTRÊME DROITE, LA RÉSISTANCE S’ORGANISE (fr).

I won’t elaborate further on the subject, but I hope to see more creators take these positions publicly, and I wish all the courage in the world to Technology Connections (yeah I don’t know the guy’s name), to withstand the attacks I’m certain he will face.

Tech

Security

• Eargapping

I didn’t think Bluetooth security issues were that serious.

Kubernetes

• Le premier intérêt de Kubernetes n’est pas le scaling (fr)

Just a little reminder that goes hand in hand with Dear friend, you have built a Kubernetes.

FOSDEM

This weekend I was in Brussels for FOSDEM. I go every year since 2019. (online in 2021 and 2022 due to the pandemic). This year, the opening keynote was particularly emotional, with the reminder that FOSDEM is a political event, and that the state of the world isn’t very encouraging right now. I recommend the replay to get a sense of the atmosphere at that moment.

A few talks I recommend watching (as soon as the replays are available):

• Evolving Git for the next decade

Git is 20 years old, but is far from finished. A great talk about the upcoming planned evolutions, and for me the discovery of Jujutsu which serves as inspiration for new ways to use Git. I need to look into it in detail.

• How to Make Package Managers Scream

Another talk I strongly recommend along with its 2018 prequel which is in the same vein. Very funny! The audience was very participative and dynamic. A pleasure to have attended.

• The state of Go

Every year, the Golang room is a must for me, and with the “state of go” presentation a pleasure to watch. Short, simple, funny, and informative. Worth watching to stay up to date with language evolutions. Even the previous years are informative, and watching them in order lets you see the language evolve.

• How to do a Podcast with Free Software?

A very good presentation of the technical and less technical basics of making a podcast. Having been myself in the position of making a podcast, as a host and for sound recording, I find it fair and informative. However you’ll need to watch and read other more technical resources for all sound production questions. In any case, it’s a good starting point.

• Podlibre: Podcast Audio Editing for the AI Age

A presentation of a new “audio” editing tool aimed at podcast creators. I’m super curious to test and play with it because the promise is really interesting and could finally allow us to skip a whole long and tedious process, or expensive proprietary equipment. The project is in its very early stages, so nothing really tangible to play with right now.

• CRA Integration – How FOSS compliance measures support CRA obligations, especially regarding documentation, security updates, and traceability.
• Implementing the Cyber Resilience Act - engaging with open source
• The Geopolitics of Code: From Digital Sovereignty to Global Fragmentation

I was rather in rest mode, and I listened with half an ear. So I don’t necessarily recommend watching the replays, however I found it interesting to see that the Janson room (the largest) was well filled. For all three talks. Questions about European regulation are attracting crowds and I’m the first to be surprised. It turns out that digital sovereignty was quite a selling point this year at FOSDEM because many talks in many rooms mentioned it, when they weren’t talking exclusively about this topic. I suppose regulatory questions too, consequently. We went far from the small rooms barely a quarter full that talked about GDPR a few years ago.

• Hacking the last Z80 computer ever made

A rather short and cool presentation with old and fun hardware to use and hack.

And that’s it, I left after that. I didn’t share everything I saw, only what seemed interesting or quality enough to be appreciated, even as a replay.

On another topic, it seems I’ve broken all my Masto audience records this year with a Toot. No idea why, but it’s funny.

https://mamot.fr/deck/@ztec/115990259401164027

The cool photos I took during the weekend:

Looking forward to the next FOSDEM, in 2027.

This week I’m sending you to YouTube almost exclusively. Sit down, and chill watching those videos that are pretty good. Then read the Ploum article at the end. Or read it first. One way or another it’s all enshittified anyway.

Security

When a whole continent can read your sexts or the credit card number you used to order Uber Eats 2 minutes ago (I’m almost not clickbait) :-p

• 39C3 - Don’t look up: There are sensitive internal links in the clear on GEO satellites (YouTube)

Really interesting conference on satellite communication. I’m shocked by the lack of encryption on so many transmissions, more so knowing that they are blasted all over entire continents. As one of the two presenters says, it makes you rethink your own personal threat model when you know your SMS may be read by everyone in the country with a few bucks worth of equipment.

Architecture & Ecology

• ARCOLOGIES: ces villes utopiques qui devaient sauver la planète (fr)

I wasn’t ready for this video. On one hand it is really nice to watch, very poetic, on an esoteric theme. On the other hand it is addressed with a lot of seriousness and analysis. I really like that he starts by stating that urban sprawl is a problem environment-wise. Arcology comes as a ready-made solution, old as a grandma’s remedy. It is NOT!

Anyway, I write down the word, and put it in my box of concepts and terms to use in dystopian stories.

• Garder sa voiture thermique, bon pour le climat… vraiment ? (fr) (YouTube)

Still on the environment theme. The question raised seems obvious to me. The answer anyway. The video clearly shattered my preconceived ideas on the subject. Like many, I think it is better to use an object as long as possible to minimize its environmental impact. However there are exceptions apparently, like cars. Broadly, any equipment that produces a non-negligible amount of pollution when used is concerned. Even if the question, if seriously asked, would have made me doubt myself, I don’t think I would have done the work seriously enough to reach such a solid conclusion. Watch this video if you have a gas car, or if you’re hesitating to change cars.

AI

• La Fabrique à Idiots (fr) (YouTube)

I share this even if you’ve probably already seen it. If you’re in France anyway. However, this is the first I’ve seen that addresses the topic in as serious and well-referenced a way as this. No bias toward or against AI in their current form interferes with the video content. That’s nice. Micode’s team seems to do a really good job nowadays. Sometimes it is a bit sensationalist, but never for nothing. It gets better and better. Congrats.

Transport

• Chantier Sensible, l’histoire invisible de la passerelle de Massy - Un film de Karine Dana (fr) (YouTube)

A short video, without any narration, without any frills. Simple, and enjoyable to watch, to take the time. I would love to see more like this on “ordinary” projects like this. A chill moment to appreciate.

Social

• L’urgence de la souveraineté numérique pour échapper à la merdification (fr)

After all those YouTube links, you must read this piece by Ploum.

Happy New Year 2026 folks!

I’m not sure I’ll still be able to do it next year given how the world is already going in 2026; And we’re only the 8th… With the USA doubling down on efforts to be one of the worst assholes (euphemism) in the world, it’s not looking good.

Society

• Chili : asséché par les data centers | ARTE Reportage (2025) (fr)

I’m stunned that people thought it was relevant and okay to build datacenters that “consume” water to operate. Like, we’ve known how to do cooling without “just” evaporating water, in closed circuits, for ages. That’s actually what we do almost everywhere I think. We even have datacenters without AC that don’t consume water. I’m really angry by the lack of empathy from people making these kinds of decisions. American companies on foreign soil (for them), so it’s not surprising either…

• Logiciels libres en bibliothèque - BAM : Biodiversité Autour de Moi - Chronique de Gee (podcast) (fr)

(I’m sharing the episode but especially Gee’s segment in it). Another good take on the question of artistic originality and copyright. Added to the topic bucket ;-)

• CETTE RESSOURCE REND FOU LE MONDE ! (YouTube) (fr)

Funny to list it here now, having watched it before the USA did the USA thing again. But well, it remains relevant.

• Wikipedia:Signs of AI writing

An attempt to detect LLM-generated writing. Since it’s Wikipedia it seems serious, but I’m not certain that anything of this kind will hold up very long against the evolution of LLMs. Unlike automatic solutions, this is more intended for human educations. A bit like the recommendations to look for inconsistencies in AI videos. But that only works as long as LLMs are imperfect (which is far from being an eternal guarantee), and doing it humanly will often be less effective than doing it automatically. Yet we know that automatic tools are not up to the task on that front. and will surly never be.

In short, a page to follow, as it’s an interesting community response nonetheless.

• Comment Votre Identité Peut Nous Faire Perdre Une Élection (fr)

Well, I won’t recommend you listen to or watch BadMulch in general; to use his words, he’s often disastrous (désastreux en french). But his series from which this episode is taken is good. I really like that he tries to create words (in the sense of making them enter common vocabulary) for the phenomena he describes. It’s accessible, quite well done (hats off for the motion design and 3D). I highly recommend the previous episodes as well.

Tech

• How I, a non-developer, read the tutorial you, a developer, wrote for me, a beginner

Funny to read. The question that nags me: “if we find the right words, does it even make sense?”

• Today in “Google broke email”

• How Github monopoly is destroying the open source ecosystem

Get off Github. Please. Alternative forges haven’t been lacking for at least 5-8 years. And self-hosting your code has become trivial, even on a raspberry-pi at home. As usual with this kind of injunction, making yourself available elsewhere doesn’t necessarily mean you have to be accessible ONLY elsewhere. Our code can be on Github AND elsewhere. Videos can be on YouTube AND elsewhere. etc…

Art

• The Living Tombstone - “Everything Is Fine (Remix)”

I think it’s cool, and quite “fitting” at the start of this year.

• Trekorama

Explore StarTrek ships as if you were there. At least that’s what a 2000s real estate agent who’s never experienced VR would say. But it’s cool anyway.

You noticed it because you follow me carefully given the numerous (zero) messages received (still zero) concerning the absence of loot of the week last week; (you can finally breathe if you read aloud) but there was no loot of the week last week 😁.
It was Christmas so… well, I have a good excuse this time. Even if my family’s christmas was postponed for the first time in 42 years (yeah I’m 42 years old, since 42 years…), I was busy.

If you follow me carefully (I know you do), you also know that I wrote 2×4k words about Godzilla this weekend too. With real monster excerpts inside! Go read/watch right after this article: ゴジラ Godzilla (1954)

Anyway, tabs ready, go!

Society

• Why Isn’t Online Age Verification Just Like Showing Your ID In Person?

A good article that explain well some of the problems posed by identity and age verification online.

• How We Lost Communication to Entertainment

Also go read the article by the same author that he cites at the beginning to understand what he’s talking about.
I just hope that enough people will continue to use free, reliable and accessible communication solutions. I don’t use XMPP for example because I wouldn’t have anyone to talk to on it today. I communicate almost exclusively with proprietary services like Telegram, Discord, Signal.

Something that deeply shocked and annoyed me recently: I wasn’t responding on Whatsapp anymore (I’m no longer reachable there), and at no point did the person on the other end imagine sending an SMS. SMS wasn’t even part of their mental model of communication with a smartphone. Worse, when they finally sent an SMS (after calling me, and I reminded them of the existence of this “archaic” system), it arrived a good 12-13h later. The fact that their provider (Free I think) delivers SMS with 12h delay doesn’t seem to worry them more than that. For me it’s grounds for non-delivery of service that I pay for, and I would have changed providers right away. SMS is usually very reliable and fast. I use it every day and I have perfectly reliable and fast conversations. And I’m not even talking about SMS with RCS, but good old SMS GSM.

Anyway, this article also made me think of another one I read recently: Domains as “Internet Handles”. The author demonstrates why it is however useful and desirable to have several identities, one per platform. Whether they are free or proprietary. The need not to mix everything, or at least not to impose it.

• Miraculous : les dessins animés deviennent-ils trop politiques ? (YouTube, French)

I didn’t know about it, but now I almost want to watch 😁. (nah, let’s not push it)

• Tout ce que l’on ne veut surtout pas savoir sur le racisme (Podcast, French)

Nothing to say, go listen.

• Vous n’avez pas d’âme (et c’est pas grave) (YouTube, French)

The subject is touchy. It’s like when you arrive with big clumsy hints to say that we (probably) don’t have free will. The reactions are immediately visceral, and I understand perfectly why. Who would want to tell themselves that everything they do, think or choose is not their doing? Nobody. Not even me. But well, when you study the question seriously enough it’s difficult to arrive at other conclusions than free will doesn’t really inhabit us. Unless we revise our definition of free will. This video addresses a bit the same problem in an accessible way (well I think so, because I understood and I’m not a philosopher). I’m adding this to the list of good content about AI. As usual Monsieur Phi does an excellent job, I love it.

Art

• Fan Art : hommage ou plagiat ? | _Underscore | ARTE (fr) (French)

I’ll talk about it soon, but personally (to make it simple) I’m opposed to copyright, authors’ rights, and any commodification of art (except materials themselves). I won’t expand on that today because I have in preparation a very long prose that will go into detail, especially to talk about generative AI. All with nuance, so keep your emotional reactions for later.

Why do I say that? Because this video gives me great pleasure and reminds us that precisely all these notions are purely artificial, and don’t hold up against creation and art in general. That precisely reproduction, plagiarism, adaptation, etc… is as important as original creation, if it still exists. The only place where I’ve seen what fan art can give in the material world, when it’s tolerated, is Japan. It’s normal and tolerated to sell your fan art, which means you can find beautiful art books, like those you find at Waterstones, with drawings of Mario or other manga franchises. And the revenues are not given back to the “rights holders”, but to the artists who produce these books. And the second-hand market being very developed there, you can find entire stores that offer them for sale. And there’s everything in there, like what you would find online.

Nice video about art copy and inspiration without hiding behind definitions that would make them absolutely different.

• How Rural Japanese Trains Still Exist (YouTube)

Since I was talking about Japan, here’s a super interesting little video about regional trains in Japan. Not the main lines, but the tiny, cute ones. Japan still has many of them, and it resonates a lot with what the state of French railways is today and how it got there. One of the phrases that struck me most: (I paraphrase) “Once the high cost part is made, the infrastructure, it’s relatively cheap to operate”. In France, our rails are rotting. The part that cost us a lot of money, time, and more, is rotting. Just because our urban planning and all-car policies, have decided that it wasn’t worth the cost to operate these infrastructures. And now we talk about urban/rural divide everywhere. Even urban areas of medium or low density find themselves landlocked because there’s no longer any train going there. And obviously it’s not as if bus networks or infrastructure adapted to active mobility (bike, walking) also existed for cases where it would be relevant.

Anyway, we let rot something we’re going to have to rebuild one day or another, and it’s another thing that pisses me off. Having to redo the hardest part of the job one day or another 😞.

ps: I don’t want to minimize operating costs. It’s just a question of ratio with construction. I also take urban planning into account in the equation. It’s much more expensive today to cut through territories to run trains or trams than it was a century ago. This cost isn’t necessarily only financial too.

Tech

• How Google Maps quietly allocates survival across London’s restaurants - and how I built a dashboard to see through it

This article could be in the “society” section. How, from a basic desire to find a good restaurant near you, we can analyze and reveal the mechanisms that algorithms hide from us. And that’s the whole problem, it’s hidden. GAFAMs are neither auditable nor transparent while they influence us from everywhere. And Google Maps is surely one of the most emblematic cases.

I wonder what the DSA (Digital Services Act) will be able to do about it, and if Europe will manage to enforce it. We all see how it goes with GDPR, right.

• N’opposons pas conformité et accessibilité (fr) (French)

Ahh accessibility. The first thing I think about when I’m shown or I see a super new website all beautiful (that’s debatable), brand new, with lots of pretty things. The controversy about the french national education website shows how much we’re not asking the right questions.

I would prefer ugly services (that’s also debatable), than what they are today, that is to say slow, resource-hungry, and inaccessible. The article mainly talks about the traditional “we’ll see about that later” when you have the chance to think about it, often it’s just not thought about at all. You who are reading me, if you’re a tech person, is your employer or the service you work on accessible?

Become instrumental and impose this topic. Force it into the agendas. Make sure each sprint includes accessibility tasks. Event better, that the DOD (Definition of Done) includes accessibility criteria.

• Abbreviations done right: The element and why not use it

Another accessibility article. When the simplest thing is still to do nothing. Or almost. Special mention for the pentacle at the beginning of the article 😈.

• CSRF Protection without Tokens or Hidden Form Fields

Here, news about CSRF. Like the author I wouldn’t have really imagined that paradigm shifts had taken place. An essential good read for any WEB developer today, especially if they think they already know how to handle this security issue.

À l’année prochaine 😉

Godzilla is part of popular culture. Known by everyone just like the T-rex or Star Wars. There have been several recent films about him. For example, the excellent Godzilla Minus One by Yamazaki Takashi

Yes, it’s a disaster movie, a monster movie, a film where everything explodes. But Godzilla movies are much more subtle than they appear.

That’s why (but also because there are giant monsters, and everything explodes everywhere) my partner and I embarked on the project of watching all Godzilla films from the first one. To better appreciate their evolution, and through them appreciate the evolution of Japan and probably the USA. Because Hollywood has clearly participated in the franchise, even though Tōhō maintains control over the concept.

Also, the films were produced after Japan’s surrender following the mass murders of civilians perpetrated by the USA on August 6, 1945 at 8:15 AM and August 9, 1945 at 11:02 AM. After which, Japan found itself under American imperialism and has probably remained so ever since.

To enjoy these films to the fullest, and especially to observe the evolutions, I’ll try to share my viewing experience and analysis of these films. The idea is to explore what makes each film special, show what impresses me, amazes me, makes me sad (because yes, Godzilla is above all a sad story), and brings me joy.

Whether it’s about the film, its form and substance, but also the techniques specific to the genre, almost invented by the franchise itself; I’ll share everything, and maybe you too will one day want to watch them all. Well, especially some of them because I suppose not everything is equal in almost 70 years of franchise.

To find my way, I’ll follow this list which traces all the films by year and era. I’ll watch them roughly chronologically. I say roughly because I’ve already seen all the Godzillas since 2014 I think. Also, there are probably crossovers that will justify changing orders sometimes. This will be reflected here too.

I hope to see this project through to the end, because I’m not really known (by the whole world, but especially by myself) for finishing things. Let’s say it’s an ambitious project for me, a kind of new 2026 resolution, but made and engaged before.

Anyway, I’m going to tell you about Godzilla in my own way,
and it starts here, right now, with the very first film from 1954: ゴジラ Godzilla (1954)

Yeah, I put from ‘54 to today, but you’ll have to be patient to catch up with the 38 films not counting all the stuff on the side. Marvel invented nothing with their MCU.

Trigger Warning

I discuss death, war, and atomic mass murders.
The article is light, but at times I’m very serious!

Hello folks. Today I’m going to talk about Godzilla, the very first film. It was released in 1954 in Japan, so yeah it’s not very recent. Yet everything is there. Everything we know about Godzilla is there.

I’m not going to summarize the film or tell you the story. For that I invite you to go read wikizilla or wikipedia.

No, here I’m mainly going to show you some parts to convince you that yes it’s a good film, and yes (also) everything that is iconic in modern Godzillas was there from the beginning! These people were visionaries 😛

Worth watching or not?

I already said it in the intro, but this film is a masterpiece. It’s absolutely worth watching. Place your Michael Bay expectations in the closet for the duration of the movie because even though the film has aged well in many aspects, it remains rather slow and has that characteristic rhythm of ’50s-’60s films. Think of it as 2001: A Space Odyssey, grin and bear it, listen (or read) the dialogues carefully, and put your phone in the closet with Michael Bay.

WATCH this film, it’s absolutely brilliant, especially if you’ve already seen other more modern Godzillas, like Godzilla Minus One from 2023, you’re going to love it.

I’m obviously going to spoil parts here, but rest assured I’m going to carefully hide some interesting bits for you to discover yourself.

And besides, spoilers never ruin a good film (fr) 😉

First roars ever, and in music

Heavy, saturated footsteps announce what we know is coming. Something big, scary, terrifying.

Then the film opens with the iconic, the unique, the very special Godzilla’s Roar. A sound that characterizes him and that will follow him throughout his career. Well I think, I haven’t checked yet, but I’m starting.

The equally iconic and majestic music that we all know (yes yes, you too) comes to dress these sounds as unpleasant as they are enjoyable. Composed by 伊福部 昭 Ifukube Akira he will create what will be both a musical icon and an inspiration for so many filmmakers (Spielberg included).

Roar and music on opening credits

Auto

Personally, this beginning gives me goosebumps since I saw all the (more) recent films. It’s like reverse fan service, returning to the source. A pleasure that’s just beginning.

I’m not going to translate videos, so speech and displayed text will remain in the original versions. It’s not necessary to translate to understand my point normally.

Here, the texts are “only” the names of people involved in the film, it’s the opening credits. They were at the beginning back then, which means the film ends very abruptly without the traditional (for our era) end credits.

A Kaiju film, no THE Kaiju film

This film is often considered as having defined the Kaiju film style in which we basically find:

• The use of miniatures sets for scale effects
• People in monster suits
• Quirky monsters, with… particular faces I would say.
• Often the monsters are huge, hence the necessary scale effects

This film from 1954, contains its share of particularly striking scenes that have aged really well. Well, I say that, but you have to look past the fact that the film is in black & white, and that yes we can see very well that these are miniatures diorama. However, the result remains coherent, and today makes you smile, but doesn’t make the film any less impactful.

Right from the start we have a little scene that reminds us of all this, a simple boat sinking. I find the fire effects particularly convincing. It’s only the subtle movements on the water that betray its real scale.

And a bit later, there’s a storm scene that batters actors, not very good actors. Well, in fact yes, they’re very good. It’s just that the expectations of the era were not the same as today. What I find overacted was the norm before. In any case, it’s fun to watch, and it’s a good way to calibrate my expectations for the rest of the film. Rather than watching it with disdain, I’m going to watch it with wonder, appreciating all overacting at its true value. Sometimes, it’s true, while laughing heartily. That’s part of the experience now.

Storm on miniature

Auto

Clearly, the construction and lighting work is simply striking. Yes we’re not fooled by the scale effects because the cameras of the era didn’t allow what we can today, but dam it’s impressive. Personally, I feel all the love of the model makers and miniaturists in these shots. And it will be like that throughout the film(s).

But where is the monster?

The film takes its time; really. Built around 3 acts, the first takes a good big third of the film. We’re told about folklore, stories of old fishermen who tell that once a monster had been spotted when fish were scarce. Its name? Gojira (Godzilla).

And there you have it, the name of this prehistoric dinosaur is sealed forever. In the film it’s this story that serves as justification to name the monster when journalists need a big title for their special edition.

However, still no monster on screen. There has been destruction, noises, but nothing visual of Godzilla himself. First clue and first characteristic of this invisible monster forged, forever: a huge footprint right in the middle of a village. A radioactive footprint! Our monster is RADIOACTIVE.

Footprint on village

Auto

So there we learn that the thing is big. Look at the size of the footprint. Also, the thing is radioactive!! The Geiger counter goes crazy, we can hear it well.

Godzilla is therefore:

• Big / Tall
• Radioactive

A few minutes later, we see trilobites hanging around. TRILOBITES, the thing supposed to be extinct for hundreds of millions of years. Godzilla therefore brought things from his home, old things. Nothing suggests it’s voluntary on his part, but more that these creatures accompany him as parasites (symbiotic?). He dropped a few on his path.

Godzilla is therefore:

• Old (or at least hangs out with things from 250 million years ago)

And now the drama. A villager sounds the general alarm and all these people who were quietly taking measurements and collecting samples rush to offer a quite striking shot-reverse-shot.

No, actually it’s mostly very funny today 😁. I can very well imagine the actors placing themselves on their mark, with the director insisting: “Imagine that this camera reveals itself to you, roars, is dozens of meters high, is terrifying!!”. Basically, the beginnings of green screen acting,… without green screen.

Godzilla is finally revealed. His size finally comparable to the rest of the world. A flawless composite shot, which even today renders very well.

So admire this person in a suit playing the big monster, in front of a (small) terrified crowd. A childhood dream!

I can’t help but consider the parade of scientists, reporters, and other explorers facing the camera as having a comic effect. Same for the “terror” on each face. But sorry, let’s put this film in its era, it’s terribly well acted!

The revelation at the end with the tracks on the beach… Marvelous!

So yeah, it’s well done, but it’s sparse for now. Almost all is suggestion. It looks like Spielberg (or is it Spielberg who looks like Honda Ishirō? 😏)

The suit

ⓒ Unknown

Before going further, a little word about the suit.

The ShodaiGoji (初代ゴジ).

I’m not going to paraphrase all the articles and wikis that talk about it, but some interesting info must still be kept in mind, especially when we watch the next excerpts.

The suit is a rubber thing on a wooden and metal frame (if I understood correctly), with cotton and fabric inside for a soft effect (no, that’s wrong, it was just so the thing wasn’t just a cutting blade for the actor).

Anyway, the important point is that this thing is far from a simple pajama. The damn thing weighed nearly 100Kg. Imagine yourself in there, while there are iron bars everywhere, that you have to “play the monster” and roll around with it!

Lucky them, all the parts that move on the head are operated elsewhere, it’s someone else who does it. But still, they had to carry the thing around and destroy everything “properly” as the director wanted. No second take, otherwise you have to redo all the miniatures.

Revelations

Well, the old fishermen were right. We knew, we saw the title of the video the film. But the film’s characters had to see it with their own-eyes! (yeah it’s one word for me, “own-yeyes”. It’s our critical eye you see)

Since I’ve already seen quite a few Godzilla films, I notice another recurring motif settling in nicely. Yeah, I didn’t see all the films in order huh, I started with the more recent ones before growing into a better version of myself, one who can grin and bear it.

So, what was I saying? Oh yes, since I’ve already seen quite a few Godzilla films I notice a rather déjà-vu scene. The PowerPoint™ of the era with acetates, like back then (yeah I’m old, there were no “interactive” boards when I was in middle/high school).

This scene, save it somewhere in you mind, it’s going to become classic in the franchise. Personally, I find it like a running gag, but right now it’s serious. It’s the first film, and it’s not joking around.

It’s the catch-up session for all those who were looking at their phone during the entire first part. I told you to put it away, you’re lucky the teacher is doing a catch-up lesson!

We don’t learn much, but everything is recapped. Basically:

Godzilla is:

• Big / Tall
• Radioactive
• Old or hangs out with old things

We even get a mugshot of Godzilla at the end.

Ah yes, I almost forgot, we do learn something new though. The film’s characters make a conjecture about the reason for Godzilla’s coming or awakening. Because yes, the old fishermen speak of an old legend (even for them). So why is he coming back now?

This is where we arrive at the fundamental of what Godzilla represents. At least in this film. Because I find it changes a bit in the following ones.

Godzilla came back because we messed up the seabed, where he lived/hibernated (strike out your least favorite). Moreover we didn’t hold back ! We atomized the thing. Like A-Bomb, H-Bomb, etc…

We all know how annoying it is when the FBI or the GIGN drop by to “politely” smash your door under an atomically overpowered search warrant. So it’s kind of expected for poor Godzilla to be angry and comes to mess up our home. The flowerbeds are unlikely to make it.

It’s suggested that the bigness of Godzilla and his superpowers (we’ll see later) comes from radiation induced by our nuclear tests.

Godzilla is homeless. We most likely exterminated his whole family, and now he’s coming for us.

More seriously, Godzilla represents the consequences of using atomic weapons. This film comes barely more than 10 years after the mass murder of civilians perpetrated by the USA on Hiroshima and Nagasaki.

Even though Japan is in full imperial colonization by the USA, this film talks about the consequences of weapons as powerful as atomic bombs. Godzilla himself will be a destroyer, and whether he has agency or not doesn’t change the basic point: Lots of people died. If Godzilla himself is a sentient and thinking being, his will can then easily be a metaphor the USA will itself as a war nation. A kind of revenge for having destroyed his home (Pearl Harbor?). If American imperialism on Japan is at work in this film, it’s probably in this little bit of lore. A kind of “good” reason to destroy everything. We’ll see later with… the sequels, but this angle of analysis really seems to survive well and strengthen from episode to episode.

Okay, let’s get back to a lighter level, to laugh a bit more with this film.

Destruction scene

Now that our friend, with a regrettably short family tree, is clear to everyone, even to the film’s characters, he’s finally going to come and wreck everything. The poor humans will try to bring out the big guns. A classic here too, where almost all of Japan is put to contribution, a bit like Evangelion. (or is it Evangelion that’s like Godzilla 😏).

We try to electrocute this big lizard with everything we have. The effects are classic for the era, I suppose the film was simply painted to make the sparks effects. We can see how a person can hide in this monster. But it makes him even more touching, when he himself pass though the high voltage lines as if it were a simple spider web.

Note that this suit which already weighs 100kg had to be operated in water! I can’t even imagine the physical endurance needed for that.

Well, all the cannons unleash on Godzilla, while he breaks the Lego towers made by little brother without even realizing that all the artillery is pounding him.

I don’t know how the model makers take the destruction of their works. Are they happy? Sad? Both? Something else? In any case it renders well. It bends as it should, well almost. The smoke everywhere makes it all even more convincing, but completely breaks the perspective and scale effects.

Then, we discover for the first time in IMAX 4D Dolby Atmos, without smell-o-rama, Godzilla’s fetid breath that melts everything. The heat ray that kills.

Spit on miniature

Auto

Note that we see our friend up close, and how hideous he is! Honestly he’s so ugly it’s beautiful.

Well, that’s not all, but he still has plenty of things to destroy, and past human defenses, it’s the city of Tokyo that’s going to take a beating.

Visually, it’s magnificent. The burning buildings are really believable. Too bad for Godzilla who does a camera look that must have been forgot during editing. They could have done another take honestly.

Also, very striking this Hotwheels™ accident 😁. It contrasts quite a bit with explosions worthy of 80s films. I won’t mention the time-lapses, which are really a classic for the era; besides rather subtle here.

The end with that foot, no, that Godzilla thigh that fills the whole frame; Magnificent! To finally demonstrate all of the man in the suit skill to destroy this warehouse by showing only a small piece of this thigh, no of this foot; Magnificent!

And note the quality of Godzilla’s composite when the villagers flee in terror! Great (technical) art!

A small excerpt where Godzilla positions himself politically and shouts ACAB in his way!

I’m showing you this excerpt because what’s obviously interesting here is the compositing of people in buildings, and at the end of the excerpt when we see him in the background of the city. A city that burns in its entirety. Observed by villagers composited at the bottom. A powerful final result.

Another motif, or rather should I say another setup, journalists on rooftops doing their job. They describe, they film, they record. Nothing will stop them from doing their job, until the last moment, not even this big arms T-rex. Well, in this particular film, they’re a bit cowardly, and stayed well away to not smell Godzilla’s bad breath.

Journalist on destruction

Auto

Let’s note that Godzilla seems to dislike clocks. No explanation is ever provided regarding this visceral hatred. It’s a totally free interpretation for the viewer. Poor clock…

Humans may be small, but they have small planes and that’s another setup here. Godzilla versus planes. Here, sorry I’m going to spoil the surprise, the planes, they don’t win.

Planes on Godzilla (well yeah, they fly, they're above)

Auto

The music is also really perfect for this moment. A treat! It almost looks like a Michael Jackson video at the beginning. Almost as if we’d expect Godzilla to start dancing like a zombie. But no, he just lets the big kids play with planes for a few minutes before remembering that Parent 2 said “dinner time” 10 minutes ago already and goes into Tokyo Bay to go eat.

If you have an eye. Or especially just if you finally put your smartphone in the closet like I told you, you’ll spot the sophisticated rocket guidance techniques. The missiles, not the salad huh. Probably something very technological based on c12h22n2O2.

The consequences

Well, here we’re going to get serious again for a few moments.

The film takes a very sinister turn, even in our times. Because if today we don’t perceive the terror and catastrophe through the images, as it could have been the case at the time when we weren’t used to ultra-realistic 3D effects, we fall back very quickly on our good old reality. Having visited the Hiroshima Peace Memorial Museum, and therefore seen some photos and, even more horrible, some drawings and accounts of the consequences of this mass murder of civilians by the USA on August 6, 1945 at 8:16 AM, I can say that this part of the film is rather chilling. It’s a film, and era requires, the images are very “clean”, but suggest largely what’s needed, and don’t spare difficult images.

The representation of Tokyo after Godzilla’s passage resembles in every way Hiroshima a few hours after the explosion. At least, from what has reached us.

The improvised hospital recalls the (numerous) history of these places that served as refuges, and post-apocalypse help centers. Personally, I think back to this school which had “survived” thanks to its reinforced concrete construction technique despite its proximity to the bomb’s detonation point. The film scene plunges me back into it, mainly because the structure and walls are alike. Surely this architecture is widespread in Japan, but here I can’t help but make the connection with the staircase where survivors wrote the names of those they were looking for.

The film makes its point again, in case the viewer had forgotten;
Or I had forgotten;
Or you had forgotten;
Or we had collectively forgotten.

More consequences

Auto

These schoolchildren singing… I don’t have the translation of their songs, the film’s subtitles did not provide them. But it moves me a lot, it’s strong.

I’m not talking about everything to keep part of the film for you to discover, but this second disaster exposition is far from trivial, and fully serves the narrative’s purpose. It advances the story in a… Well, watch the film (or read the complete summary) if you want to know. I don’t want to cover everything here.

Small aside concerning the shot of the image in the TV. It’s another recurring motif in Godzilla films. For me, the images of destruction are put inside another screen, as if they themselves were fiction. A double level of fiction, so that the characters can still (a bit) deny the reality of this desolation, but it’s no longer possible.

The end

Without revealing how, humans, more exactly Japanese humans, end up winning against Godzilla.

Godzilla on ocean floor

Auto

Note that Godzilla walks on the ocean floor. He doesn’t float. At least not here.

This is the end

Auto

The film ends on a boat taking a beating, with a Godzilla dying like a 5-year-old child would make it in their bathtub. At least, at the beginning. After it’s more of a paleontologist’s fantasy.

Note that Godzilla’s death is beyond any doubts. He is defeated, destroyed, literally dissolved in the water. Quick, direct and painful according to the official translators of his roars.

There you go, spoiling is done. He dies at the end. This obviously didn’t prevent the making of dozens of canonical sequels, so don’t feel like you hold all knowledge and power in your hands. There’s still much to discover.

Final thoughts

Well, this film is a masterpiece. The first person who says I’m exaggerating is just not worthy of receiving all this entertainment, and should go back to playing marbles or Pogs.

It bears the mark of its era. It’s unintentionally fun, yet it never tips over into camp. On the other hand, it’s completely outdated, but we forgive it, it’s the era we said.

The film’s purpose is interesting, full of hope (Godzilla is defeated) and sadness (the consequences, and the means used to defeat him). Since I haven’t said everything, you’ll have to dig on your own to understand what I mean by that. Because my goal here is twofold. First to make you live the film like me, or at least a part of it. To share my excitement, my disappointments (there aren’t any here), my sadness, and my reflections and analysis. So that the crazy idea of watching ALL Godzillas goes from a… crazy idea precisely to a revelation of evidence to you (We all have the right to dream). All this for my second objective, to convert you. For you to become a Godzilla fan. But not just a fan of his entertaining aspect, of his destructions, of the chaos he creates (yes all that is super cool), but also of what he represents. Of what Godzilla shows, through the eras. Because remember that Godzilla films have been made and plenty since this one in 1954!!!.

This first opus lays the foundations. We see it, everything is there and will be taken up, remixed, reinvented, copied, plagiarized, innovated. I had to take the time for this film. It’s quite likely that I’ll be much more succinct on the dozens of others. Otherwise it’s going to take me a lifetime to write these articles, hoping I keep up the pace 😅.

See you soon for the next one. I’ve already seen it, just need to clip and write the article.

Thanks for reading me,
Bisoux 😗

It starts in complete darkness. We’re warned that phones aren’t allowed as they would ruin the experience. Will this horde of old folks manage to restrain themselves? Yeah. Surprising but yeah, and not a single flash as a result.

© Léa Rouaud

Anyway, it starts in the dark.

Two shaved heads (or almost) light up slowly on either side of the stage. We can make out the pianist at the keyboard on the right. Then a bit later the clarinetist on the left.

Sitting cross-legged like monks. Dressed in black, only the arms, head, legs, feet and hands are truly highlighted by the lighting.

The music begins with low, muffled sounds that start the enchantment.

In the center, the light gradually reveals the dancer; she’s well-haired. Motionless at first, she gradually straightens up while starting to dance, spinning. And it’ll be like that until the end.

Spinning, spinning, spinning. A short pause in the middle but that’s it; quite an achievement.

The stage is bathed in video projection. Both the backdrop and the floor. Everything is projected onto black, which tends to fully integrate the artists and especially the dancer, anchoring her in the projected reality.

Musically, it’s dreamlike and very intoxicating. A kind of soft trance, like a dream. Lots of repetition, originality, and changes. It reminded me of one of my most singular experiences, Erwan Keravec - In C 20 sonneurs at Le 104.

I was simply captivated for an hour, and for good reasons. It’s this kind of experience that makes me appreciate live concerts.

Something perhaps unique, though if other performances are scheduled, I seriously invite you to attend.

• Yom - Clarinet, composition
• Manuel Peskine - Keyboards
• Ana Elena Tejera - dance, choreography
• Julien Oska Colardelle - stage direction
• Ludovic Palabaud - sound

CC-BY-NC Ztec

I’m late again. I think I’ll have to get used to it and consider it normal. That way I won’t be late anymore.

Tech

Github Action is probably the most successfully CI solutions since Jenkins. It completely killed any other (free) solutions for open source, and start to do the same for companies and private projects. Personally, I’m still convinced of the vast superiority of Jenkins over any (if not all) other CI solutions. I would even prefer it for CD as well in most cases. To be clear, I don’t think Jenkins is for everyone, or every projects, but all complaints I saw about it never goes away with other solutions, Github Action included.

On my side project, I use Forgejo Actions (a copy of Github actions for Forgejo). Soo, yes, I’m maybe not as coherent as I would like to be 🙈 Anyway, this video show how microsoft pushed the product hardly, even pushing bad quality code and architecture. Fun and concerning to watch. (keep in mind that bad code happens everywhere, to every developers. It happens)

• Why Github Why? (YouTube)

Interesting take about “alternate” version of a web page. Never really thought of using that to provide structured version of any page in addition of the HTML one. I will probably try that and keep it in mind for future projects.

The AI side of the question is also interesting, and goes in the opposite direction of everyone wanting to block AI crawlers. I’m not convinced that we should ultimately block AI from crawling our content, either text or art. But doing it in a sustainable way is mandatory, and this can be part of the solution. Today, Developers of AI companies are like babies, and did not learn from the decades of work we did in web crawling, indexing. What they create and deploy have a huge tendency to impose huge load on infrastructure not sized ( and will never be sized ) to handle it. Working on new standards to help those babies developers to do things the right way may help.

• Stop crawling my HTML you dickheads - use the API!

Social

Two useful videos to watch.

• BUDGET 2026 : EN FINIR AVEC LA DETTE [ARGENT MAGIQUE] (YouTube)
• L’ARNAQUE DU “MANAGEMENT BIENVEILLANT” (YouTube)

Interesting article that puts into writing many things I see in online activist circles:

• Pourquoi Mélenchon va gagner en 2027

I’m not as convinced as the author that people are, proportionally, more left-wing, but especially that the majority of non-voters would turn out for the left. It seems to me that we’ve more often observed a certain balance in left/right mobilization on that front. And that’s probably what scares me the most. The fact that many people will mobilize believing they’re voting in their interest but will do so by voting fascist. The most recent example being the last legislative elections, which despite the mobilization for the Nouveau Front Populaire, also saw strong RN mobilization.

I’m intimately convinced that it’s imperative to have the mobilization of racialized people, with real antiracist and decolonial proposals to counterbalance. Create a real proposal that shows a desirable future for EVERYONE, not just for white working classes.

As said in the article, only La France Insoumise represents these values, but they must not pull a Kamala Harris, and go seriously and all in. Because half-measures or compromise on these issues won’t mobilize.

SmolWeb

I came across those two websites recently. Should be a mandatory (short) reading for any frontend web developer and designer. 😉

• Motherfucking Website
• Better Motherfucking Website

Served them right?

This article goes in the long list of examples of how GAFAM and any online provider should never hold any important and relevant part of your life. Backing up is a good practice, but should not be required to keep your data safe from sudden deletion by a provider. As it is about Apple, and the guy seems to have dedicated his life to Apple, this article made me laugh a lot. This guy had the knowledge and brain to know the risks, and still committed.

• 20 Years of Digital Life, Gone in an Instant, thanks to Apple

Already the second Loot, and I’m late. I had to go back over my Blue and Red book (those who know will know). Let’s go:

Tech

A bit of entertainment with this Linus duo. Pretty fun and informative to watch.

• Building the PERFECT Linux PC with Linus Torvalds (YouTube)

An old piece but still very interesting. If you need to create or improve a documentation space, especially a technical one, read this (with plenty of references inside to reed as well):

• How to structure technical documentation: best practices

New initiative to disable JavaScript by default in our browsers. Fun fact: this blog is supposed to be fully functional without JavaScript. 😉

• https://マリウス.com/disable-javascript-org/
• https://disable-javascript.org/

Society

This show is interesting because it tackles a topic that doesn’t get much coverage. I hope they come back to it, with more details and examples in particular. Feel free to skip to minute 14 for the real content of the video and skip the funny quiz at the beginning, which I personally didn’t find that interesting.

• Esclavage, racisme, whitewashing : Ces jeux vidéo problématiques - Disculture EP.2

I now take a lot of distance from Sismique’s episodes (I’ll probably explain in detail why someday), but this one is a pretty good summary and analysis of violence and how we are exposed to it. I like how it connects the decrease in violence with changes in our exposure to it. I hadn’t really realised this shift, nor the effects it causes, especially in terms of decontextualisation and the disappearance of spaces where we can talk about it. The fact that we no longer have suitable, dedicated spaces, even slightly hidden ones, to talk about violence is quite similar to the question of exposure to porn (especially for younger people). Anyway, it’s a great episode: it makes you think and gives a lot of leads to dig into.

• Violence sur nos écrans : la banalisation du mal | #PAUSE - Sismique - Où va le monde ? (podcast)

A very good documentary about Trump’s America and how/why Americans vote for him. Sometimes hard to watch, but very informative without falling into the “we have to hear all sides” trap, because it doesn’t put all viewpoints on the same level.

• Retour chez mes potes trumpistes (1/4)

• Ça va mal finir (2/4)

• Le fils qu’ils n’ont jamais eu (3/4)

• Réveille-toi ! (4/4)

Music

While making the entry Makoto Ozone - TRiNFiNiTY at Le Duc des Lombards I discovered the recording of this concert. Very happy to be able to watch and listen to it again. You can hear me clapping. 🤡

• https://philharmoniedeparis.fr/fr/live/concert/1163708-avishai-cohen-makoto-ozone

A nice video about the fully analog process of making a vinyl record. Very informative and it really makes you think about all the work required to produce all those trendy new records we see everywhere.

• How Lacquer Masters are made changed my mind about vinyl. JUST LISTEN TO IT! (YouTube)

Train

• The fascinating world of Japanese train melodies (YouTube)

Paris / Photography

Seeing photos of all of Paris, including places where I’ve lived, from the 70s is just so cool. It’s a different Paris, still a bit spared from cars, even though their invasion was already well underway.

• C’était Paris en 1970

Other

• https://puginarug.com/ (a fascinating universe to explore)
• https://theuselessweb.com/ the rest of that universe

Discovery of the day: Old School Funky Family at Le Duc Des Lombards. A rare occasion as they don’t come to Paris often.

Lots and lots of energy, fun, and enjoyment. The pyjama style is fun and perfectly matches the chill & joyful vibe they created.

It’s obvious that all members of the band really enjoy playing together. Lots of chemistry and interactions between them. The audience isn’t forgotten either, and they managed to get people standing up at the Duc, a cool but somewhat stuffy place.

On the musical side, it’s fun and funk. A bit of big band sometimes but without fanfare. Electro and classic jazz mix together neatly. The Sousaphone is really nice. Never heard of it before. I like it. The Duc removed the enormous grand piano that usually takes half the stage to make all 8 (eight!!!) musicians fit.

A few listening tracks to discover:

• Bamboche & Chaloupade (YouTube)
• Ceci Est Une Patate ! (YouTube)
• Le Bon, La Brute et Le Centriste (YouTube)
• and on usual streaming platforms.

I strongly recommend listening to them, but most importantly, go see them live if you can. The experience is way better than albums. So much more energy. You can go without hesitation, I assure you.

They remind me of another band with a different style but the same energy on stage: Who Parked The Car. They’re also cramped on small stages (especially at Sunset Sunside) due to the number of people, but they have so much energy it’s really nice to experience.

The Crowd Breakdown: Diverse and young. As usual at the Duc, there’s always a fair amount of tourists. Overall, the audience was really diverse in all aspects. I often do a quick sweep of the crowd at concerts I attend. I’m looking for obvious demographic over-representation, or lack thereof. Absolutely not scientific at all, but it’s interesting and fun to know who came and if you feel you fit in. In jazz, the common case is to see lots of white dudes over 50. Tonight, that wasn’t the case, and it’s cool.

℗ 2025 Pleins Poumons Productions

This is the inaugural loot, the first haul from my exploration. This week’s drops: readings, listenings, viewings, and other encounters that gave me XP or made me think. Whether the drops are fresh or farmed long ago, doesn’t matter.

Everything I pick up and share doesn’t necessarily reflect my opinions, and at no point does sharing something here mean unconditional support for its content or its author. However, if I add something to my inventory, it’s because I believe it’s an interesting item, even if I won’t necessarily equip it.

I won’t pretend to be objective. The loot is already sorted by my criteria, no need to specify each item’s rarity, you’ll see what I equip or what I skip.

I won’t limit myself to tech drops either.

In form, a simple list of items with sometimes lore when I have something to say other than “it’s legendary, go loot it”. As for categories, I’ll probably craft them by educated guess. If you have farming spot recommendations, feel free to share them through one of my networks.

Last thing, the game is available in French and English. So if you don’t have these language packs installed, sorry, you won’t be able to understand everything. If some articles have official translations, I’ll obviously use them, but it’s quite rare.

Alright, let’s go:

Tech

• Send this article to your friend who still thinks the cloud is a good idea

Cloud this, cloud that. We see it everywhere, and the small company that needs to host 3 peanuts does it by default on one of the 3 hyperscalers (Google, Amazon, Azure). It’s annoying from a GDPR regulation standpoint (because it’s incompatible), but also because it makes many companies complex and dependent on these techs.

The article is good, however it’s a bit too one-sided for my taste. A bit as if these techs are never useful. Never in the sense of statistically insignificant. I think it’s more often useful than assumed by the author. However, you need to know how to identify the needs that justify it and the timing. On this last point, I agree with him that no, we can often do without cloud and the techs are really not objective on these two aspects (I add that).

Also, using these tools comes with many (good) practices. Particularly concerning development cycles, with for example the almost obligation to automate deployments, and observability which is almost imposed with patterns rather recognized as effective. And when you know how difficult it is to have healthy development practices from end to end (from product manager to prod through dev), starting from the base proposed by cloud also often makes sense. And the fact that Devs and Architects know how to handle them rather well is also a plus.

Basically, it’s often useless to do cloud, and a simple CDN is enough to handle the load, with a fairly large server. But it’s not necessarily desirable if good dev practices are not there due to a lack of knowledge, or investment in them. Something that is often somewhat mandatory with cloud.

Ideally, a European GDPR-compliant cloud with mature technical solutions would be best. And having one or two dedicated servers should also be seen as a good solution rather than being criticized or dismissed as outdated.

• The Database That Killed My Wife

• Podcast: Chronique de l’emmerdification du Web, détournement de finalité et absence de base légale du RGPD

Social

• La solution pour sauver le Festival de BD d’Angoulême

• Ma première soirée en club parisien

I’m not a drug user, and I never have been. However, I’ve had my share of bad experiences in Parisian clubs, but that was when I was young. Like the author, I’m surprised, and pleasantly so, that such positive and respectful spaces can and do exist now. Hoping this becomes normalized.

• Internet est mort. Voici comment le ressusciter

• Électricité : factures en hausse, qualité en baisse : où est passé l’argent d’ENEDIS ?

• Ce juge français est « interdit bancaire sur toute la planète » à cause de Trump

When you know that associations and activists have been describing exactly what’s happening now for surely more than 20 years, it’s an angry “we told you so” that comes first. If you want to know the next predictions, check out La Quadrature du Net for example and their site Technopolice.

Science and Consciousness

These two links are interesting to read/listen to one after the other (in no particular order). Also, I recommend all of Monsieur phi’s videos on the subject of LLMs.

• Consciousness Could Be an Ancient Trait Evolved Millions of Years Ago, Not a Human Superpower
• Video Podcast: What if ChatGPT was already conscious? - Albert Moukheiber and Monsieur Phi

ⓒ Hiroyuki Seo

Second time for me seeing Takuya Kuroda and his band at New Morning. Still as classy and jazzy as ever. Each story is delightfully told through music, easy to follow.

Those guys are pro, precise and fluid. They know their stuff.

The electro/jazz mix is still as good.

Takuya Kuroda san told the story about the track “Car 16 15 A” like last time. Always a pleasure to listen to it, especially live where the band takes time to develop each movement. A little preview on youtube for the curious, but live is by far even better.

The Crowd Breakdown: Diverse and young. A few tourists as usual. I had the opportunity to talk to a Japanese lady and exchange a few recommendations. It was really nice. As always at New Morning, people stay standing and move well. But don’t forget it’s jazz sooo no pogos!

At Deezer, we decided to move from a homegrown (not really REST) API to GraphQL. Why? That’s not the point here. I’m focusing on the how.

Here’s an account of how we managed this large-scale migration and some of the challenges we faced.

Conflict of interest disclosure: Deezer is my employer 😛 for 10 years now.

This is a talk I gave at the GraphQL Paris meetup in October 2025. It gives a broad overview of the topic to fit within a 30-minute presentation slot.

If you’d like to know more about a specific aspect, let me know on social media so I can cover it in more detail next time I present on this subject.

Presentation slides

You can find the slides here (in English):

• PDF
• Slideshare

Conference video

Watch the full talk on YouTube:

Disclaimer: The video is in French, but the slides are in English.

Thanks for reading me,
Bisoux 😗

A charitable musical evening bringing together numerous jazz artists in solidarity with Gaza.

All ticket proceeds has been donated to humanitarian aid via UNICEF and the France Palestine Solidarity Association.

Participating Artists

Alain Jean-Marie, Clélya Abraham, Naïssam Jalal, Erik Truffaz, Estrella Besson, Lynn Adib, Cyril Atef, Arnaud Dolmen, Tony Tixier, Climène Zarkan, Baptiste Ferrandis (Sarab), Leonardo Montana, Noé Codjia, Neil Saidi 5tet, Charlotte Planchou, Mark Priore, Lukmil Perez, Mohamed Abozekry, Lina Belaïd, Wadie Naim, Jules Jassef, Léo Jassef, Omarhaba, Corentin Giniaux, Enzo Carniel, Melvin Marquez, Camille Prenant, Laure Sanchez, David Paycha, Noé Huchard, Clément Daldosso, Kevin Lazakis, Samuel F'hima, Tom Peyron, Philippe Maniez, Pablo Campos, Malte Arndal, and Zoot Big Band.

BadBadNotGood

• Al Sowinski - Drums
• Chester Hansen - Bass
• Leland Whitty - Guitar, Saxophone
• Felix Fox - Piano
• Juan Carlos Medrano Magallanes - Percussion
• Kae Murphy - Trumpet, EVI synthesizer

©Sylvain Chausse

Opening act

• Yazmin Lacey

©Myah Jeffers

Recording of the show: Naïssam Jalal - ‘Souffles’ Full Concert, Studio de l’Ermitage 2025

• Naïssam Jalal - flute, voice, composition
• Yom - Clarinet
• Emile Parisien - Saxophone
• Robinson Khoury - Trombone
• Sylvain Rifflet - Saxophone

©Alexandre Lacombe

• Thomas Salvatore - Keyboards, voice
• Sebastián Muñoz - Saxophone
• Félix Reneault - Saxophone
• César Aouillé - Guitar
• Ludovic Prieur - Bass
• Malo Ropers - Percussion
• Alejandro Dixon - Drums

©Yoel Reboh

• Moeun Son - Violin, compositions
• Dexter Goldberg - Piano
• Romain Habert - Guitar
• Julia Richard - Double bass
• Wadie Naim - Percussion, Drums

©unknown

• Makoto Ozone - Piano
• Kunito Kitai - Drums
• Shimpei Ogawa - Double bass

Makoto Ozone “Trinfinity” - Momentary Moment (Youtube)

©unknown

Backstory

A few years ago, I designed and set up a GraphQL server at Deezer. Among all the challenges I faced, one was related to performance. Designing a GraphQL server that is both convenient to use and performant is not an easy task, especially in Node.js.

One of the optimizations we made with Jimmy Thomas was to use GraphQL JIT instead of the default GraphQL execution engine. The README claims to improve response time and increase the number of queries that can be executed per second by reducing the CPU time it takes to compute a single query.

What is GraphQL JIT

GraphQL JIT is a Just-In-Time compiler for GraphQL queries. It is designed to take advantage of V8 optimizations to increase performance. It is a drop-in replacement for the default execution engine of GraphQL, with a few restrictions. One of the most important is stated in the project’s README:

All computed properties must have a resolver, and only these can return a Promise.

Depending on how you design your server, this limitation may or may not be an issue. In my case, it was not an issue at all.

The performance gain claim is pretty impressive: up to 10 times faster than the default execution engine.

Why we switched

At the time, we conducted extensive tests on the project and used Gatling for load testing. I specifically crafted some reference queries and used them to compare the performance of all the changes made. This way, we could improve response time and admissible load bit by bit. One of the most important changes was the adoption of GraphQL JIT. The performance gain was substantial enough to make it worth the switch at the time. Unfortunately, all reports and graphs were purged, thanks to Jenkins’ cleaning policies.

Why reconsider it now?

When we conducted the tests and made the switch, the project was barely used. I completely relied on my test queries for benchmarks. Those queries were built to be representative of a theoretical usage, but they were not real queries as no clients of the API existed yet.

Now, the API is live and in use. I have real-world queries to analyze, and I can observe the server’s performance in real-time with real clients.

So, let’s revisit this choice and see if it is still the right one.

Test methodology

Two tests are in order:

• One in production with real customers queries
• One using the old method with Gatling, as before

Production test

To test the two engines simultaneously, I modified the server’s code to randomly start with one engine or the other. When deployed onto the Kubernetes cluster, multiple instances of the service will be started—enough from both engines to gather significant statistics.

The code is quite barbaric but works wonderfully:

1expressApp.use(
2   [...],
3   Math.random() > 0.5
4       ? createJitGraphqlMiddleware({schema})
5       : createJsGraphqlMiddleware(schema),
6);

Once in production, I will analyze metrics to evaluate the impact of each engine. I will primarily focus on:

• System metrics such as CPU and memory usage
• Node.js system metrics such as Event Loop Utilization (ELU), HEAP size, and Garbage Collection
• Response time goals (how much it deviates from the target)
• Average response time and 95th percentile

Lab tests

For the lab tests, I will use what’s already available from my previous tests years ago. I’ve set up Gatling with a few typical queries. The queries and rate of fire are not exactly the same as before because, over the years, I’ve tuned them to be more representative of the real world. When I say “more representative,” I don’t mean they are truly representative. Instead, I’ve simply adjusted them based on our observations in the field and modified them in a way that makes them closer to the real world. Unfortunately, I have no way of confirming this; it’s far from scientific here.

The most notable change is the way scenarios are grouped. Previously, each query had its own rules during the test. Now, I’ve grouped them into two main categories, representing two typical client profiles:

• Light users send small queries. This is standard usage of the API, similar to how any customer would interact with our apps/front-end.
• Heavy users make large and complex queries with lots of fields and nested fields. This is typical of a customer using some costly features of our apps or any rogue client trying to abuse the API.

What’s important is that the two profiles do not fire queries at the same rate. One fires far more frequently than the other.

The final Gatling scenario looks like this:

 1val heavyUser_ConcurentUser = max_reqps/20
 2val heavyUser_ConcurentRequest = max_reqps/20
 3val lightUser_ConcurentUser = max_reqps - heavyUser_ConcurentUser
 4val lightUser_ConcurentRequest = max_reqps - heavyUser_ConcurentRequest
 5
 6
 7setUp(
 8   lightUser.inject(
 9       rampConcurrentUsers(1) to (lightUser_ConcurentUser) during (2 minutes),
10       constantConcurrentUsers(lightUser_ConcurentUser) during (duration - 2 minutes)
11   ).throttle(reachRps(lightUser_ConcurentRequest) in (duration minutes)),
12  
13   heavyUser.inject(
14       rampConcurrentUsers(1) to (heavyUser_ConcurentUser) during (2 minutes),
15       constantConcurrentUsers(heavyUser_ConcurentUser) during (duration - 2 minutes)
16   ).throttle(reachRps(heavyUser_ConcurentRequest) in (duration minutes))
17).protocols(httpProtocol)

The Gatling tests will run on my laptop but will target instances of the server deployed in a development Kubernetes cluster. This cluster is similar to the one in production in form and configuration but not in size. I won’t push the cluster to its limits, so I’m not concerned about the results being skewed by the environment.

I will monitor the same metrics as in the production test. This time, I will not consider the results from Gatling itself because I am running the tests from my laptop and cannot trust the response times gathered by Gatling. Moreover, I don’t really need them.

Results

Production results

Response times

One of the metrics I have is the time it takes for the engine to compute a request and produce a response. It does not include the time it takes to send the response to the client or network delays.

The 95th percentile shows a somewhat greater difference. It is not enormous, but five milliseconds does feel significant.

Regarding response time, we can say, without any doubt, that the `JIT engine’ is not worth it, mathematically speaking. However, we are talking about a five-millisecond difference, so it is not really a big deal either.

System metrics (CPU & RAM)

The service is deployed on a Kubernetes cluster. I have access to the cluster metrics and can see how much CPU and RAM are used by the service.

In Kubernetes, we set CPU and RAM reservations. It is a good practice to indicate to the cluster how many resources the pods will need. For example, we can define that one Node.js process can use up to two CPUs. The graph then shows how much of these two CPUs are used compared to the reservation. The same applies to RAM. Of course, the graph shows values for the entire cluster, not just one pod.

1limits:
2 memory: 512Mi
3requests:
4 memory: 512Mi
5 cpu: 2

This is an example of a configuration for a pod. This means the pod will have two CPUs and 512Mi of RAM at its disposal.

The js-engine’ uses 2% less CPU than the JIT-engine.’ It is a consistent difference, but like the response time gain, it is not that significant. Don’t get me wrong, when you deploy hundreds or thousands of pods, 2% can mean a lot, but I’m not convinced it does in my case.

The RAM usage is a bit more interesting. The js-engine’ uses ~20% less RAM than the JIT-engine.’ Here, a 20% difference feels substantial.

Node.js metrics

I challenge you to identify when the deployment occurred. The ELU (Event Loop Utilization) did not shift a single bit, even with 50% of the requests being handled by the js-engine.’ This graph only shows the overall cluster level, as I cannot get metrics for each engine separately. In my opinion, since 50% of the pods use another engine, if the ELU were significantly impacted, we would see it. I will assume that using either JIT’ or `js’ does not change anything in terms of Event Loop Utilization.

Monitoring the HEAP shows that the `old’ space seems to have reduced slightly. It is not very obvious in the graph but it is visible in the average. The maximum did not change, but the minimum did. This confirms what we observed earlier with RAM usage.

Other HEAP spaces were not impacted at all, and the same goes for Garbage Collection. They all tell the same story as the ELU metrics. They remain unchanged from before, so I didn’t bother screenshotting the graphs.

Lab results

I ran the Gatling test for each engine in two separate deployments. The tests were run simultaneously. Strictly speaking, they could have impacted one another, but I don’t think this effect is significant, as the cluster I was using had enough resources to handle the load. Moreover, the load was not that high, and I kept the request rate well below limits.

Response times

The results are not favorable for the js engine. The difference is substantial. We must remember that this is a lab environment, and response times may differ from those in production. However, we are losing about 50 milliseconds on the 95th percentile and 20 milliseconds on average. These represent slowdowns of approximately 50% and 70%, respectively, compared to the `JIT’ engine.

System metrics (CPU & RAM)

The first notable difference is in CPU usage. The js-engine’ uses 20% more CPU than the JIT-engine,’ which is significant.

Regarding RAM, the difference is smaller but still present, with only a few percentage points of increased usage for the `js-engine.’

Overall, our synthetic tests indicate that the JIT engine is more efficient than the JS engine.

Node.js metrics

The story remains consistent with the Node.js metrics. The `js engine’ uses more HEAP, more Garbage Collection, and has a higher Event Loop Utilization (ELU). Specifically, the ELU increased from 30% to 50%.

What is the meaning of all of this?

Okay, things are getting interesting. The production results clearly show a slight advantage for the js’ engine. This win is small and could, in some cases, be considered negligible. However, it is still a win. On the other hand, the story is entirely different when we look at the lab results. The JIT’ engine is a clear and substantial winner! Moreover, based solely on the lab results, it is a no-brainer to use the `JIT’ engine.

The lab results were essentially the same as those I had years ago. The JIT’ engine is faster, more efficient, and requires fewer resources in a lab environment. Back then, I only had those results to inform my decision and naturally chose the JIT’ engine.

Today, however, the production results complicate the decision:

• The js’ engine is better or at least as good as the JIT’ engine in terms of response time, resource usage, and overall performance.
• The complexity introduced by the `JIT’ engine has a cost.
• The `JIT’ engine has certain limitations. We did not encounter these, so they may not be highly relevant in my context.
• The `js’ engine is effectively the “standard.”

The question has been raised internally, and there seems to be a consensus that standardization is important enough to justify switching back to the `js’ engine. I tend to agree with this perspective.

Why does the optimization not show up in production?

The production and lab tests have opposite behaviors that feel counterintuitive. I believe the main reason lies in how we utilize the GraphQL API.

When we began our GraphQL journey, we had a steep learning curve. One of the biggest challenges was designing the schema and envisioning the queries that would be made. We had to rethink everything, moving away from our accustomed REST APIs.

However, as anyone who has worked on an existing system knows (do I hear “legacy”?), we had to consider many existing scenarios and technical limitations. For example, paginated results were not as common back then as they are today. When designing the schema, we aimed to follow the “state of the art” in this regard, but it was not always straightforward or even feasible.

After years of learning and striving towards best practices in GraphQL, we find ourselves in a much better situation than expected. For instance, we initially feared processing massive queries with numerous nested fields. This concern hasn’t completely disappeared, but looking at the queries generated by our apps, the reality isn’t as daunting as we imagined.

Teams have adapted to the new way of thinking, embracing the limitations and concepts of a GraphQL API. Gradually, they’ve accepted the need to approach things differently. The quality of our GraphQL queries seems sufficient, indicating that we no longer require the `JIT’ optimizations.

One of the best use cases for `JIT’ appears to be when dealing with complex queries that have many nested fields. We currently don’t have that issue, or at least not to a significant extent.

In the future, we may need to reassess this decision based on how we develop our upcoming applications and frontends. But for now, the `js’ engine seems more than adequate.

Conclusion

A few years ago, I designed a GraphQL server and made several architectural decisions. We implemented some obvious optimizations and conducted tests to verify their effectiveness beyond marketing claims, confirming our choices.

Not all decisions were perfect, and there is much to discuss and critique about the project. However, the choice of the ‘JIT’ engine was a good one at the time.

Today, the context has evolved. We have (enough) real-world clients, and both Node.js and V8 have been improved. Additionally, the standard implementations of GraphQL have benefited from years of community enhancements.

Revisiting old decisions has proven to be insightful, if not beneficial. The ability to test a theory in production easily and without the fear of breaking anything is a luxury. I’m grateful for this opportunity.

Ultimately, we can conclude that the real world always surpasses the lab and its benchmarks. While this may not come as a surprise, it’s always a good reminder. If you have the resources and time, take the opportunity to test your hypotheses using real-world data or actual clients whenever possible. For this, it’s crucial to maintain healthy development practices that facilitate easy and safe testing and deployment.

In conclusion:

 The `JIT’ engine seems to be worthwhile. However, I would recommend conducting proper tests, if feasible, before fully committing to it. Gains may be small or even imperceptible, depending on your circumstances.

 Anyway, always monitor, metrics are essential!

Thanks for reading me,
Bisoux 😗

Many thanks to Pauline Munier and Gillian Kelly for their help in writing this article. You can also find this article, on Deezer.io

• Benjamin Asnar - Bass
• Lucas Belkhiri - Piano
• Francesco Marzetti - Drums

©unknown

• Didier Malherbe - duduk, flute, ocarina, Clarinet, Soprano Saxophone
• Loy Ehrlich - hajouj, kora, sanza, vielle, ribab, Keyboards

Guests:

• Steve Shehan - Percussion
• Jean-Luc Di Fraya - Percussion
• Éric Löhrer - Guitar

Hadouk - Teaser album 2024 (Youtube)

©unknown

Andy Emler MegaOctet

• Adrien Sanchez - Tenor Saxophone
• Philippe Sellam - Alto Saxophone
• Guillaume Orti - Alto Saxophone
• François Thuillier - Tuba
• Laurent Blondiau - Trumpet
• Andy Emler - Piano, composition
• Claude Tchamitchian - Double bass
• Éric Échampard - Drums
• François Verly - Percussion

Léa Cazauran Lady Rocks

• Léa Cazauran - choreography, dance
• Julia Ortola - dance
• Rodrigue Lino - dance
• Jesus Gil - dance
• Maëlle Dufour - dance
• Lisa Dwomoh - dance
• Santiago Codon Gras - dance
• Ludovic Piscioneri - dance
• Virgile Garcia - dance

Technical team

• Vincent Mahey - sound
• Marine Stroeher - lighting

Make ’em Move - Andy Emler MegaOctet (musique) + Lady Rocks (danse hip hop) (Youtube)

©unknown

• Meshell Ndegeocello - Bass, Keyboards, voice
• Jebin Bruni - Keyboards, voice
• Abraham Rounds - Drums, voice
• Christopher Bruce - Guitar
• Kyle Miles - Bass
• Justin Hicks - voice

©unknown

Robinson Khoury

• Robinson Khoury - Trombone, various brass instruments, voice, modular synthesizer
• Lynn Adib (guest)
• Anissa Nehari - digital Percussion, voice
• Léo Jassef - Piano, Keyboards, voice

On digital percussion and voice: Anissa Nehari, a musician with sensitive dynamism, captivating and captivated by the rhythms of the world.

On piano, synthesizers and voice is Léo Jassef, a versatile artist of great sensitivity who has been heard alongside Yom.

Robinson Khoury on trombone, various brass instruments, voice and modular synthesizer.

Robinson Khoury “MŸA” - Horizon (Youtube)

©unknown

Opening act

• Bill Summers - Percussion
• Mike Clark - Drums
• Chris Severin - Bass
• Donald Harrison - Alto Saxophone
• Kyle Roussel - Piano

ⓒunknown

• Takuya Kuroda - Trumpet
• Craig Hill - Saxophone
• Takahiro Izumikawa - Keyboards
• Kyle Miles - Bass
• Chris Napo - Drums

©unknown

This summer, I upgraded a project from Node 18 to Node.js 20. The code was not that old, and I did not expect any problems….But as with every Node.js upgrade, unexpected behaviors happened.

Why upgrade?

It is common practice to keep projects up to date. Lagging behind too much may increase the work needed to upgrade when the choice isn’t yours anymore. I could have waited, but I had some time to do it. Node.js 18 is currently in “MAINTENANCE”, which means there are security fixes but no more new features. Maintenance is set to end in June of next year.

Node.js 20 is the current LTS version available. It is in active development and will enter its maintenance phase next year.

Soon after the next LTS (Node.js 22) is released, a migration will be scheduled (and maybe another post?)

How?

This part is pretty simple: change a few versions values in the “package.json” file, then run npm install as usual.

All dependencies versions are managed by renovate bot. Therefore, “package.json” only contains the exact versions of each dependency.

The version is set to Node “20.15.1” because it was the very last version available at the time of the upgrade. Any future minor upgrade will be done automatically without changing anything in the “package.json” file. Docker images are built regularly and target the latest version of Node.js 20 like a rolling release.

Deployment and first results

This is a critical project, but deploying it is usually simple and quick. No one fears its deployment. But as this was a potentially impactful upgrade, I paid more attention than usual.

Project vitals

The main dashboard I look at during a deployment is the project’s vitals. It contains all the required metrics to know at a glance if the project is healthy or not. I won’t go into details here for confidentiality reasons, but I can show you the response time goal of the project:

The purple vertical line is roughly when the deployment occurred. You can see the response time went up. Note: Given that the purple line is not displayed on all graphs, follow the red arrow instead. The graph shows the percentage of responses that match one of the three response time goals we follow. Requests faster than 100ms went from 96.8% to 96.1%—a minor defeat at the first battle.

Sure, we can agree that losing around 1% is perfectly acceptable, but it shows that something happened.

Looking at the average response time, we can see the increase more easily. It went from ~23ms to ~28ms.

I then looked at all the other graphs, metrics, and logs. After validating and considering the project as healthy, it was time to dig in to understand what was happening. As there was no immediate threat to the project’s stability and the response time was still well above the budget thresholds, I could take some time to dig deeper; otherwise, a rollback would have been in order.

System vitals

The project is running on a Kubernetes cluster. I have access to the cluster’s metrics and can see the pods’ CPU and RAM usage. Those unfamiliar with Kubernetes can consider a pods like one Node.js process started inside a Docker container.

CPU increase

I noticed an increase in CPU usage, going from 24% to 30% of the reservation after the deployment—another defeat.

In Kubernetes, we set CPU and RAM reservations. It is a good practice to indicate to the cluster how many resources the pods will need. For example, we can define that one Node.js process can use up to 2 CPUs. The graph then shows how much of these two CPUs is used compared to the reservation. The same applies to RAM. Of course, the graph shows values for the entire cluster, not only one pod.

1limits:
2  memory: 512Mi
3requests:
4  memory: 512Mi
5  cpu: 2

This is an example of a configuration for a pod. This means the pod will have two CPUs and 512Mi of RAM at its disposal.

Memory usage decrease

The memory usage went down. A small victory!

This project always loses weight after each deployment. It is normal, and it gains it back after a while. It takes a few hours to stabilize, but here, it went down more than usual, and at first glance, it will likely stay like that.

Node.js vitals

The first Node.js metric I was interested in is the Event Loop Utilization (ELU). This metric is essential to understand the health of a Node.js process. It basically shows how much time the process is spending doing work in the event loop, from ~0% for an idle process to 100% for a process that is always executing JavaScript code.

The graph shows a slight increase after the deployment, going from 17% to 20% on average. A defeat, again! It is not really surprising as we know the CPU usage increased, but we can clearly see that it had a (small) impact on the JavaScript code running. This may be the reason for the increase in response time.

After that, I looked at the HEAP and the Garbage collector statistics. They both make sense together.

We lost weight on the HEAP, but something is happening with the Garbage Collector. The minor GC is running more often and is taking more time.

Without going into too much detail, the minor GC is a process that targets newly created objects in the HEAP. Another one (the Major GC) targets all “old” objects.

The Minor GC is a fast process that runs often. It is normal to see it called at high volume. However, something changed after the upgrade as we went from a few calls to thousands of them. Even if it is a fast process, it ate up almost 4 seconds of CPU time on the entire cluster. By comparison, we were only eating up less than 1 second before the deployment.

This increase in GC can easily explain the CPU usage and part of the ELU increase, which results in a slower response time.

I might have my culprit here.

Issue identified, now what?

Ok, there is clearly something going on with the Garbage Collector. But what is it?

If we look at the HEAP in detail, there were a few notable changes after the deployment:

The map HEAP space disappeared, and the shared spaces appeared. But most importantly, the new space went from ~33MB to ~8MB.

Do you remember what I wrote a few lines back?

The Minor GC targets newly created objects in the HEAP.

Is this the reason why the GC is now running so often? I think a smaller space means more frequent GC runs because when the space is full, the GC needs to run to free some space. If the space is smaller, it will become full quicker and run more often.

Why is the new space smaller?

There is no configuration of the HEAP that targets the new space in the project. This means Node.JS itself probably changed between versions 18 and 20. But there is one HEAP space that is configured: the old space. We run the server with the following command:

1node --max-old-space-size=300 dist/server.js

I don’t know if this impacts the new space as well. I started searching for references to the new space in the Node.js release notes. At first, I found nothing relevant, but after a while, I made the link between “new space” and “semi space.” The V8 engine under the hood of Node.js uses this wording instead of “new space.” And sure enough, I found a change in the Node.js 19 release notes: https://github.com/nodejs/node/pull/44436

The parameter “–max-semi-space-size” was added because it was itself added to the V8 engine. By following breadcrumbs, I finally found these changes in the V8 engine: https://chromium-review.googlesource.com/c/v8/v8/+/1631593 and https://chromium-review.googlesource.com/c/v8/v8/+/4384482

These changes modified the way the max-semi-space-size is computed. I’m no C expert, but by simply reading comments and some parts of the code, we can see that the new space’s size has changed. The new value is now computed from various others. Unfortunately, I could not find how it is computed exactly as my C++ is a bit rusty, and V8 is not a simple project. I can only make hypotheses at this investigation stage; one is that the new space’s size is computed in relation to the old space’s size.

The first change was introduced in V8 10.6 and is now part of Node.js 20, which uses 11.3. For reference, Node.js 18 was using V8 10.2. This means that going from Node.js 18 to 20, we also went from V8 10.2 to 11.3.

A workaround?

The “new” HEAP space is now computed differently. Upon the upgrade, it went from ~32MB to ~8MB. It is most likely the reason why the GC is now eating up so much CPU time compared to before.

A new parameter was introduced to specify the “new” space’s size: “–max-semi-space-size.”

I tried to set it to the previous value and see if it solved the issue.

I changed the command to start the server to:

1node --max-semi-space-size=16 --max-old-space-size=300 dist/server.js

The value “16” comes from the documentation itself.

--max-semi-space-size=SIZE (in megabytes)

Sets the maximum [semi-space][] size for V8's [scavenge garbage collector][] in
MiB (megabytes).
Increasing the max size of a semi-space may improve throughput for Node.js at
the cost of more memory consumption.
[...]    
The default value is 16 MiB for 64-bit systems and 8 MiB for 32-bit systems. 
[...]

Let’s deploy this simple change and see if it works.

The response time has gone back to the previous values. It’s a win!

The CPU usage is also back to normal, and the RAM usage is still lower than before. It’s a double win!

Finally, we can see that the GC went back to a “normal” behavior. Also, the new space is now back to its original value. It confirms my hypothesis that the new space’s size was the issue.

What happened?

At this point, the issue has been solved in production. But I still need to understand what exactly caused the new space to shrink like this. I hypothesize that the max-semi-space-size is now computed in relation to the old space’s size. Since we specify a --max-old-space-size of 300MB, a relatively low value for Node.js, it would not be surprising that it significantly impacted the “new” space’s size and caused it to shrink that much.

Let’s build a lab project to test and confirm this hypothesis.

Lab project

I found a basic HEAP consumption script online and used the standard “V8” module to obtain HEAP statistics. Here is the resulting script:

 1const v8 = require('v8');
 2
 3// Dumb function stolen somewhere on the internet and modified to allocate
 4// lots of new objects. A bit barbaric, I know, but it does the job.
 5function allocateMemory(size) {
 6    // Simulate allocation of bytes
 7    const numbers = size / 8;
 8    const arr = [];
 9    arr.length = numbers;
10    for (let i = 0; i < numbers; i++) {
11        arr[i] = {"test": Math.random()};
12        arr[i][`${Math.random()}`] = Math.random();
13    }
14    return arr;
15}
16// We allocate memory to trigger the garbage collector 
17// & force the HEAP to grow, including the `new_space`
18allocateMemory(1000000)
19const heapSpaces = v8.getHeapSpaceStatistics()
20console.log(
21	heapSpaces.filter(item => item.space_name === 'new_space')[0].space_size 
22    / 1024 / 1024 // We display the value in MB for convenience
23);

When running this script, the output will be a single number representing the size of the “new_space” in MB. Before outputting it, I ensured that this space was grown to its maximum value by allocating objects in memory.

1$ node test.js
232

Now, let’s run this script with the “–max-old-space-size” set to 300MB using Node 18 and then Node 20.

1$ nvm use 18
2Now using node v18.20.4 (npm v10.7.0)
3$ node --max-old-space-size=300 test.js
432
5$ nvm use 20
6Now using node v20.15.1 (npm v10.7.0)
7$ node --max-old-space-size=300 test.js
832

The “new_space” size is the same for both versions. I was clearly wrong.

Of course, doing the same without the “–max-old-space-size” gives the exact same result.

What is going on? Why is my project behaving differently? At this stage, I can’t understand the reason why.

After more searching in V8 code and, most importantly, a hint from one of my colleagues Zibock, I may have a new hypothesis related to the memory reservation in the Kubernetes configuration.

The hypothesis now is the “new_space” size is computed in relation to the memory reservation of the pod.

Without going into details, Kubernetes memory and CPU reservation are done through the Kernel. Meaning that what is actually enforcing the reservation is the Kernel itself. It is applied directly to the process running inside the pod via the cgroups. Therefore, the process itself can be aware of the limit and adapt.

I could test that in a Kubernetes Cluster, but there may be an easier way to test this hypothesis. Sure enough, Docker allows setting memory reservations, too, via the --memory parameter. More on that here.

Firstly, the results without memory reservation:

1$ docker run -ti --rm -v ./:/ node:18 node /test.js
232
3$ docker run -ti --rm -v ./:/ node:20 node /test.js
432

These results show the exact same result as before, as expected.

Then, with limits:

1$ docker run --memory=512m -ti --rm -v ./:/ node:18 node /test.js
232
3$ docker run --memory=512m -ti --rm -v ./:/ node:20 node /test.js
42

Finally, we have a different result. The “new_space” size is now 2MB. It confirms the hypothesis that the memory reservation is used to compute the “new_space” size.

I now have my culprit. The memory reservation in the Kubernetes configuration is the reason why the “new_space” size shrunk so much.

Conclusion

The overall upgrade to Node.js 20 is a win. The global memory consumption is significantly lower than before, and apart from the few moments when the “new_space” size was too small, everything worked flawlessly.

When using Node.js in a Kubernetes environment, and more broadly in a containerized scenario, you must consider the memory and CPU reservation. It needs to be configured in order to set limits to your process that would otherwise consume more than you expected. Also, setting it to low values requires attention to ensure the process accommodates the limit appropriately.

Today, we saw that setting RAM limits on a process also impacts how much it will allow itself to consume. Production services with high loads or receiving tons of requests can impact performance and latency. If I had not monitored the metrics, we would have missed our target response time, dropping by a whole percent! A shame, I know.

The impacts of CPU limits on Node.js processes would make good material for a next post, as I observed surprising effects when I did some tests a few years ago. Maybe I’ll write about it in the future… So keep your eye on the blog!

The final word paragraph is:

In Node.js, HEAP space is now (since Node 19) sized in relation to the memory limits of the process (among other things).

In a Kubernetes/Container context, where setting RAM limits is a common practice, it can unexpectedly impact the performance of your Node.js process. The “–max-semi-space-size” parameter must be considered in those scenarios.

Update 24th October 2024

Joe Bowbeer opened an issue and proposed a PR on Node.js project to clarify the semi-space behavior. PR still open for review as of today.

Thanks for reading me,
Bisoux 😗

—

Many thanks to Pauline Munier and Gillian Kelly for their help in writing this article. You can also find this article on Deezer.io : https://deezer.io

El Comité

EL Comite ⓒ Anastasia Polak

• Rolando Luna - Piano, Fender, Keyboards
• Harold López-Nussa - Piano, Fender, Keyboards
• Gastón Joya - Double bass, Bass
• Rodney Barreto - Drums
• Yaroldy Abreu - Percussion, congas
• Carlos Sarduy - Trumpet
• Irving Acao - Saxophone
• Flavia Coelho - voice (guest)

Anthony Joseph - Black History

ⓒ AJ

• Anthony Joseph - voice
• Andrew John - Bass
• Thibaut Remy - Guitar
• Rod Youngs - Drums
• Renato Paris - Keyboards
• Colin Webster - Saxophone
• Dave Okumu - multi-instrumentalist (guest)
• Roger Raspail - Percussion (guest)

Le Cri du Caire

• Abdullah Miniawy - voice
• Peter Corser - Saxophone
• Karsten Hochapfel - Cello

The Harvest Time Experiment - Tribute to Pharoah Sanders

• Tisziji Muñoz - Guitar
• Joshua Abrams - Double bass
• Hamid Drake - Drums, Percussion
• Sheila Maurice-Grey - Trumpet
• Idris Rahman - Saxophone, Clarinet
• Emre Ramazanoglu - Drums
• Liran Donin - Bass

Makaya McCraven

• Makaya McCraven - Drums
• Jahari Stampley - Piano
• Matt Gold - Guitar
• Marquis Hill - Trumpet
• Junius Paul - Bass

©Nate Schuls

Opening act: Endea Owens & the Cookout

• Endea Owens - Double bass
• Miki Hayama - Piano
• Shenel Johns - voice
• Louis Fouché - Saxophone
• Kris Johnson - Trumpet
• Lee Pearson - Drums

©EO

Endea Owens & the Cookout

• Endea Owens - Double bass
• Miki Hayama - Piano
• Shenel Johns - voice
• Louis Fouché - Saxophone
• Kris Johnson - Trumpet
• Lee Pearson - Drums

Makaya McCraven Quartet

• Makaya McCraven - Drums
• Jahari Stampley - Piano
• Matt Gold - Guitar
• Marquis Hill - Trumpet
• Junius Paul - Bass

Melissa Lesnie Sextet - Tribute to Amy Winehouse

• Melissa Lesnie - voice
• Victor Pitoiset - Guitar
• Nicola Sabato - Bass
• Erwan Morisse - Drums

ⓒunknown

Shibuuya!

• Pierre-Louis Varnier - Keyboards, sampler SP-MKII
• Japhet Boristhène - Drums, SPDSX
• Lucas Benedetti - Bass

SHIBUUYA! | NEONAIR LIVE SET

ⓒunknown

ⓒunknown

• Robinson Khoury - Trombone, various brass instruments, voice, modular synthesizer
• Lynn Adib (guest)
• Anissa Nehari - digital Percussion, voice
• Léo Jassef - Piano, Keyboards, voice

On digital percussion and voice: Anissa Nehari, a musician with sensitive dynamism, captivating and captivated by the rhythms of the world.

On piano, synthesizers and voice is Léo Jassef, a versatile artist of great sensitivity who has been heard alongside Yom.

Robinson Khoury on trombone, various brass instruments, voice and modular synthesizer.

Robinson Khoury “MŸA” - Horizon (Youtube)

©unknown

Son Moeun Project - Reflet

• Moeun Son - Violin
• Dexter Goldberg - Piano
• Galindo Cuadra - Guitar
• Florent Allirot - Double bass
• Gabriel Ferrari - Drums

ⓒunknown

Emma Rawicz Quintet

• Emma Rawicz - Tenor Saxophone
• Ivo Neame - Piano
• Jim Hart - Vibraphone
• Kevin Glasgow - Bass
• Marc Michel - Drums

ⓒACT Gregor Hohenberg

Theo Croker Quartet

• Theo Croker - Trumpet
• Eric Wheeler - Double bass
• Idris Frederick - Keyboards, Piano
• Jeremiah Collier - Drums

ⓒunknown

A mini-festival presented by Resolution label featuring 3 emerging groups from the new Parisian scene:

Kalupto

• Calypso Larrazet-Llop - voice, Keyboards
• Victor Gonon - Guitar

Indie-Pop and Folk-tronica

Who parked the car

• Laura Wamba - voice
• Sebastián Muñoz - Saxophone
• Félix Reneault - Saxophone
• César Aouillé - Guitar
• Thomas Salvatore - Keyboards, voice
• Ludovic Prieur - Bass
• Malo Ropers - Percussion
• Alejandro Dixon - Drums

Presenting their EP “Drafts” - Soul, Jazz, Funk, Afro Groove, Acid Jazz, Hip Hop and Electro.

Videos:

• Amande • Live at New Morning (Youtube)
• Only Missing You • Live at New Morning (Youtube)
• Trapped In Forever • Live at New Morning (Youtube)
• Cesar’s Dream • guitar solo • Live at New Morning (Youtube)

Notap

Pop arrangements from the 2000s to 2020s.

Loreena McKennitt

ⓒunknown

Opening Act:

• Grégoire Jokic

Thylacine Symphonique

• Thylacine - composition, performance
• Uele Lamore - direction and arrangements
• Orchestre Lamoureux
• Bravinsan - Piano

ⓒunknown

Reader: I don’t have the time to read, what are you talking about?

Blogger: I plant peppers and show how I turned an Ikea Kallax into a greenhouse for my apartment. It’s cool and works great. I hope, I still don’t have any peppers.

The beginning, plants

I have recently developed a new interest: plants! It’s cool and makes the room nicer. However, this new hobby takes time and requires attention. Also, I need to confess that my final score is not great. This clearly proves I’m still a padawan. I’m currently in the learning phase and I try to learn as much as I can from my mistakes.

• Too much water here
• Not enough water over there
• Not enough light on this one
• Too much water here as well
• Pot too big there
• Drawing for the last one

Anyway, I learn day by day. Except maybe about watering, where I think I may have a sadistic wish to drown my plants. I don’t see any other explanations.

Seeing plants grow is pleasing and brings enjoyment to me. However, the market’s offering of plants seems to always revolve around the same kinds. Once you have a Monstera, Yucca, Ficus, or Pachira, stores seem to always have the same things.

This has sparked a new interest for me: growing my own plants. Ideally, exotic ones that are not often seen here.

Obviously, I’m the very first in the world to grow such an interest, and I’m also the very first to try this experience. What an original I am!

The beginning II, seeding

Here I am in a new world, the world of cultivation. How does one make seeds sprout? I can already hear you saying it’s easy: Put the seed in the ground, pour some water, and wait!

Nope. If it were so simple, there would be no science, nor would there be industrial agronomy :-p. As a young padawan, you can guess I went in almost blindly anyway with little to no thought and just put seeds in good soil, then patiently waited!

Indor sprouting

I’m an urban dweller, living in Paris in a typical Haussmann-style apartment. My windows face south and west, but I don’t receive a lot of sunlight during the day. Despite these challenges, I’m determined to make plants thrive and sprout in less-than-ideal conditions.

To summarize:

• Low light
• Indoors
• Any time of the year because I’m not patient!

The beginning III, first fail test : Flamboyant

I recently bought Flamboyant (Delonix regia) seeds, a beautiful tree!

First issue: It’s a big tree and needs to live outside. These problems will of course be problems for future me. Good luck, future me!

Second issue: It needs a lot of light. Therefore, it probably won’t fully grow in my little flat unless I use growing lights.

Here I am searching for good growing lights compatible with a living space.

Artificial indor Sun

On Amazon, I found lots of options using LED lights. I chose a random one with 8 LED strips and a timer allowing for automatic switching on/off of the light. For 50 € I went for it to try.

I don’t provide any links, but you can easily find these products yourself with different prices and options. I set up everything basically in a square of one of my existing furniture, an Ikea Kallax lying around in my living room.

Seeding

I am now ready to seed, but as I’m really ignorant, I decide to look around on the internet to seek some advice. I find this page. I read and learned that I need to “scarify” seeds then put them in water for two days.

“Scarify” means grinding the seed to remove the outer layer that usually protects the seed interior. They explain everything here.

This process will help the sprouting of the seed. As I didn’t have any sandpaper to do it, I skipped this step and directly put the seeds into water.

I did not take pictures, but after two days, each seed was clearly opening with a little extra white/green tips on one side. It was like it was sprouting already.

Finally, I put each seed in potting soil, one seed per pot. I used the most basic potting soil I could find at a local store.

Sprouting

Out of all the seeds (10, I think), only two sprouted and only one survived more than two days.

Score of 1/10 for my first try.

The ultimate sacrifice

Once the Flamboyant sprouted, I moved it near the window for better lighting. LEDs are good, but the sun is best anyway, even in my situation.

This is the last known picture of my little Flamboyant, still alive. A few weeks later, all the leaves died, and it stopped growing… forever.

Flamboyant conclusion

It is obvious it didn’t have enough light to grow. The Flamboyant needed more than I could give it, using the growing light or the real sun through the window. Moreover, I need to be careful with watering. The cause of death is not clear, but with the lack of light, drowning was probably not helping the plant to thrive.

The beginning IV, second test: Pili Pili

While cooking, I kept some seeds of a small pepper called “bird’s eye chili” or Pili Pili. Tasty, spicy, and good looking.

Online, I found out that like the Flamboyant, you need to leave the seeds in water for 12 hours before seeding them. Also, sprouting is best between 28°C and 30°C, and the temperature needs to be stable and go down below 20°C.

My living room temperature does not go under 20°C, so there’s no issue there. But I’m not really fond of living in a 28-30°C living room. Therefore, I tried to create a heated greenhouse.

Kallax heated greenhouse

I built the greenhouse with the same tools as before but assembled differently. Instead of having 8 LED strips spread over different squares, I put all of them in one. I added aluminum foil to reflect light and heat radiation. The goal is to get the most out of the light and radiation emitted by the LED strips. The Kallax is black; by itself, it would have absorbed most of the light not directly reaching the plants. Having the aluminum foil helped as most rays of light had another chance to hit a plant.

I made that with what I had lying around. I screwed old towels to almost seal off the back and front of the squares. The front towel is only screwed at the top, allowing me to raise it to look inside. This is not sealed perfectly, but I hope it will reduce enough airflow to allow the temperature to rise to 28-30°C.

Another cool thing: once closed, the light inside does not bother me anymore in the living room. Days of a purple-lighted living room are over. I will be able to set up the timer without any bother from the light when I watch a movie.

I set up a temperature and light probe and started a timer cycle.

This is not perfect, but we can clearly see that when the lights are on, the temperature stays around 29°C which is enough. If I disable the timer (and live the light always on), the temperature says.

Bird’s eye chili seeding & sprouting

With such temperatures, I quickly seeded all my Pili Pili. Given my last success rate (1/10), I put between 3 and 5 seeds per pot. I had hopes of getting some of them to sprout, at least one per pot.

In only a few days, the first sprouting happened. This time, I had almost a 100% success rate.

With so many seedlings, I decided to conduct some tests by leaving one mini greenhouse in the lighted square, and one beside the window. We can clearly see the impact of growing lights after just a few days. My window does not let enough sun in.

Repotting

After a few days, when all the true leaves started to grow, I decided to repot all the seedlings as they had already outgrown their pots. I did not expect all seeds to sprout and grow like that in each and every pot.

I tried to separate some plants, but many of them were already intertwined. I took the cautious path and left some of them as is to reduce the risk of losing any.

This choice may have consequences in the future. We’ll see; I’m confident future me will keep you informed!

Kallax greenhouse PAF improvement

The next step after the first towel prototype was to enhance the PAF (Partner Acceptance Factor) of the solution. Make it less ugly for a living-room space. Who would like to have such a horrible, ugly towel thing near your TV or your best decorating thing.

Ikea sells doors that feel ideal for this purpose. Moreover, the door comes with a complete seal of the square, making the full square sealed up from both front and back. Perfect!

Without hesitating, I get one to test it. After following thoroughly the really long and complex manual, here I am with a closed square with Growing LED lights inside. I took the time to put aluminum foil on each side as before, including the door. This last part is more important than it looks. (More details on the next episode)

I find the result clean and okay. The PAF definitely increased. Performance is also a win.

More doors will be acquired to further increase the PAF of the other squares and finally remove this ugly towel.

The beginning, conclusion. Will I have pepper to eat ?

You (almost) already have all the technical information to answer this question by yourself. But, I did not explain how to do the math. This will be the next article’s main focus, to include more science into the mix. I will also explain how I chose the LED strips for the 3rd square, with some hints to make better choices in this regard.

I’m still a young padawan, so don’t take all I write too seriously as I’m in the learning process. Of course, I don’t really know if I will have any fruit one day as all plants are still growing right now. But if the science I found works, I have a non-zero chance to have some.

Thanks reading me,
Bisoux 😗

P.S.: This article was written when I was still living in Paris. I have since moved to a brighter place that will have some (hopefully good) consequences.

Avishai Cohen & Makoto Ozone - The Amity Duet

Listen and watch

Tracklisting:

1. On Green Dolphin Street - 09:29
2. Gesture #2 - 05:00
3. Oberek - 12:37
4. Lazy Uncle - 09:19
5. Intervention - 02:09
6. Remembering - 06:51
7. The Ever Evolving Etude - 06:36

Etibar Asadli Trio - Mugham electro/oriental project

• Etibar Asadli - Piano, Keyboards
• Ranto Rakotomalala - Bass
• Martin Wangermee - Drums

ⓒunknown

Who Parked The Car

• Laura Wamba - voice
• Sebastián Muñoz - Saxophone
• Félix Reneault - Saxophone
• César Aouillé - Guitar
• Thomas Salvatore - Keyboards, voice
• Ludovic Prieur - Bass
• Malo Ropers - Percussion
• Alejandro Dixon - Drums

ⓒYoel Reboh

Ballet in one act, five tableaux and an apotheosis (1912)

Based on Charles Perrault, Madame Leprince de Beaumont, Madame d’Aulnoy

Artistic team

• Maurice Ravel - Music
• Martin Chaix - Choreography
• Patrick Lange - Musical direction
• Camille Dugas - Sets
• Aleksandar Noshpal - Costumes
• Tom Klefstad - Lighting

© Julien Benhamou / OnP

© Julien Benhamou / OnP

© Julien Benhamou / OnP

Singer and multi-instrumentalist Oscar les Vacances has a fine and incisive writing style. He talks about adolescence, questions modernity, masculinity, speed, and tries to make us travel into the depths of our imagination and our unspeakable fantasies. An intimate face-to-face concert especially for you…

• Oscar les vacances - voice

Fusion between mugham and jazz.

• Etibar Asadli - Piano
• Chris Jennings - Double bass
• Yoann Schmidt - Drums

ⓒunknown

• Ana Layla - voice
• Dré Pallemaerts - Drums
• Victor Foulon - Double bass
• Raphael DeBacker - Piano
• Diego Di Vito - Guitar

ⓒunknown

Lakecia Benjamin “Phoenix”

• Lakecia Benjamin - Saxophone
• Zaccal Curtis - Piano
• Elias Bailey - Double bass
• EJ Strickland - Drums

Theo Croker

• Theo Croker - Trumpet, voice, laptop, efex
• Mike King - Keyboards
• Eric Wheeler - Bass, Double bass
• Shekwoaga Ode - Drums

• Yonathan Avishai - Piano solo

©Annie-Claire Alvoët

©Annie-Claire Alvoët

• Man with a mission

ⓒunknown

• Andrew McCormack - Piano
• Kyle Eastwood - Bass
• Gautier Garrigue - Drums

© DR

I recently released a small project called poulpe.ztec.fr. I personally use this repository as a library in one of my other projects. When I do so, my import statement looks like this:

1// main.go
2import git2.riper.fr/ztec/poulpe

I have two issues with this approach:

• It’s ugly, right ?
• If I ever change the location of my library, the import statement will break. (You may have already noticed the ‘2’ in the name.)

Most open-source projects rely on GitHub and simply use ‘github.com’ imports. However, as you may know, GitHub is now owned by a company doesn’t care about open-source anymore. To promote the safety of open-source, we should reduce or remove our dependency on GitHub.

Having thousands of imports that specifically target GitHub.com is not ideal for achieving this goal.

TL;DR

We can use our own domain name as long as we have ownership of the hosting at that address.

1// main.go
2import poulpe.ztec.fr

GOPROXY

You can use GOPROXY to specify an intermediary between your go get command and the version control system (VCS), such as GitHub. This is a nice way to ensure resiliency to GitHub unavailability, among other things. However, using GOPROXY does not solve the issues of ugliness and specifying the location by name.

Replace

In your go.mod file, you can use the replace directive to specify where to fetch a package.

1// go.mod
2replace github.com/ztec/poulpe v0.3.1 => git2.riper.fr/ztec/poulpe v0.3.1

This is useful, but not practical to do on every package. Additionally, it is the responsibility of the library user to do so.

Instead of

Alternatively, you can globally change where and how to load library code based on domains. This is often used to fetch libraries using ssh instead of https, allowing the use of credentials. It is as simple as adding this kind of configuration to your ~/.gitconfig file.

1[url "ssh://git@github.com/"]
2	insteadOf = https://github.com/

or

1[url "https://git2.riper.fr/ztec"]
2	insteadOf = https://github.com/ztec

This is somewhat useful as well, but it requires the library user to perform the task and does not give you, the owner, control. Additionally, you still need GitHub.com to deliver the library.

Custom name

If you think about it, the only thing you can be sure you own is your domain and what’s hosted behind it. Golang allows you to redirect any URL to a proper Golang repository. It is easy and gives you control, not the library user.

For example, I changed my package name from git2.riper.fr/ztec/poulpe to poulpe.ztec.fr.

1// go.mod
2module poulpe.ztec.fr

Just doing so will work for the project itself. However, anyone who wants to use your library won’t be able to use it with the new name. Worse, they will probably have a nightmare to use it as Golang checks the name of the module in the go.mod file against the import name and fails if they do not match.

To make it work for everyone, you can simply add this to the HTML file at the URL https://poulpe.ztec.fr.

1 <meta name=go-import content="poulpe.ztec.fr git https://git2.riper.fr/ztec/poulpe.git">

When doing go get, Golang will check the URL and search for this meta.

Perks of custom name

With the custom name, you now have full power over the naming of your package. You can also change its location if you wish. There is no reference to GitHub, although you can still host it on GitHub. The name uses a domain that you probably own.

You also reduce the imprint of GitHub on projects using your library and can decide to migrate your code wherever you want.

Thank you reading this,
Bisoux 😗

More detailed informations:

• when-should-i-use-the-replace-directive
• are-there-always-on-module-repositories-and-enterprise-proxies
• how-to-use-a-private-go-module-in-your-own-project
• private-module-repo-auth
• https://pkg.go.dev/rsc.io/go-import-redirector
• Remote_import_paths

• Georgia Cécile - voice
• Euan Stevenson - Piano
• Edouard Pennes - Double bass
• Max Popp - Drums

ⓒunknown

Nobuo Uematsu

ⓒunknown

I talked about it in my previous post, I made an emoji search engine. It is integrated into my systems. But as I had sometimes, and it was quite fun to do it, I made a standalone version that I published online

Now, you can test it and see the code by yourself.

• The preview: https://poulpe.ztec.fr
• The code: https://git2.riper.fr/ztec/poulpe or https://github.com/ztec/poulpe

Thank you reading this,
Bisoux 😗

For a project of mine, I had to handle emojis. The goal was to create a search engine to find emojis. I am not starting from scratch and have to include my code into an already existing Go program. So, let’s see how to build a little search engine for emojis in Go.

TL;DR: The full result is available here: git2.riper.fr/ztec/emoji-search-engine-go

You can also test and see the final result. All details here: poulpe.ztec.fr - The emoji search engine open-sourced

🐗 Emoji! Get them all!

First, I had to find the list of all available emojis. Easy! You can find it on the official Unicode website.

https://unicode.org/Public/emoji/15.0/emoji-test.txt

This file look like this

[…]
# group: Smileys & Emotion

# subgroup: face-smiling
1F600                                                  ; fully-qualified     # 😀 E1.0 grinning face
1F603                                                  ; fully-qualified     # 😃 E0.6 grinning face with big eyes
1F604                                                  ; fully-qualified     # 😄 E0.6 grinning face with smiling eyes
1F601                                                  ; fully-qualified     # 😁 E0.6 beaming face with smiling eyes
1F606                                                  ; fully-qualified     # 😆 E0.6 grinning squinting face
1F605                                                  ; fully-qualified     # 😅 E0.6 grinning face with sweat
[…]

Basically, there is the Unicode code, the character itself, and a description. The file is dedicated for machines, so it should not be hard to load and parse it.

Before starting head down, let’s do our due diligence and check what the community has already done.

I found:

• github.com/enescakir/emoji, last updated in 2020.
• AkinAD/emoji, last updated in 2022 (a fork of the first).
• github.com/kenshaw/emoji, last updated in 2021.

Looking at the code of those libraries, I found something fascinating: https://raw.githubusercontent.com/github/gemoji/master/db/emoji.json

This file, which is regularly updated, is perfect and can be parsed with less effort. Moreover, it already contains some metadata like aliases.

This is the way, lets load this file

 1package pouet
 2
 3import (
 4	"encoding/json"
 5	"github.com/go-zoox/fetch"
 6)
 7
 8type EmojiDescription struct {
 9	Emoji          string   `json:"emoji"`
10	Description    string   `json:"description"`
11	Category       string   `json:"category"`
12	Aliases        []string `json:"aliases"`
13	Tags           []string `json:"tags"`
14	HasSkinTones   bool     `json:"skin_tones,omitempty"`
15	UnicodeVersion string   `json:"unicode_version"`
16}
17
18type GithubDescriptionResponse []EmojiDescription
19
20func fetchEmojiFromGithub() (results []EmojiDescription, err error) {
21	response, err := fetch.Get("https://raw.githubusercontent.com/github/gemoji/master/db/emoji.json")
22	if err != nil {
23		return
24	}
25	err = json.Unmarshal(response.Body, &results)
26	return
27}

I used github.com/go-zoox/fetch to fetch the file just because I’m a lounger

🦓 Emoji, Scan them all!

In my program, I already use Bleve to index and search through other things, so I will do the same here. The operation is pretty simple as I do not have to store anything and can keep an in-memory index.

 1package pouet
 2
 3import (
 4	"fmt"
 5	"github.com/blevesearch/bleve/v2"
 6	"github.com/sirupsen/logrus"
 7	"strconv"
 8	"strings"
 9)
10
11var (
12	index  bleve.Index
13	emojis []EmojiDescription
14)
15
16func indexEmojis() error {
17	// we create a new indexMapping. I used the default one that will index all fields of my EmojiDescription
18	mapping := bleve.NewIndexMapping()
19	// we create the index instance
20	bleveIndex, err := bleve.NewMemOnly(mapping)
21	if err != nil {
22		return err
23	}
24	// we fetch the emoji from the internet. This can fail, and may be embedded for better performance
25	e, err := fetchEmojiFromGithub()
26	if err != nil {
27		logrus.WithError(err).Error("Could fetch emoji list")
28		return err
29	}
30	emojis = e
31	for eNumber, eDescription := range emojis {
32		// this will index each item one by one. No need to be quick here for me, I can wait few ms for the program to start
33		err = bleveIndex.Index(fmt.Sprintf("%d", eNumber), eDescription)
34		if err != nil {
35			logrus.WithError(err).Error("Could not index an emoji")
36		}
37	}
38	index = bleveIndex // we make the index available
39}

Once indexEmojis is called, I have an index ready to be used to search for emojis. Let’s test it.

 1package pouet
 2
 3import (
 4	"fmt"
 5	"github.com/AkinAD/emoji"
 6	"github.com/blevesearch/bleve/v2"
 7	"github.com/sirupsen/logrus"
 8	"strconv"
 9	"strings"
10)
11
12var (
13	index  bleve.Index
14	emojis []EmojiDescription
15)
16
17func Search(q string) (results []EmojiDescription) {
18	if index == nil {
19		// no Index mean indexEmojies was not called yet or did not finished. No results (boot process)
20		return
21	}
22	// we create a query as bleve expect.
23	query := bleve.NewQueryStringQuery(q)
24	// we define the search options and limit to 200 results. This should be enough.
25	searchrequest := bleve.NewSearchRequestOptions(query, 200, 0, false)
26	// we do the search itself. This is the longest. Approximately few hundreds of us 
27	searchresults, err := index.Search(searchrequest)
28	if err != nil {
29		logrus.WithError(err).Error("Could not search for an emoji")
30		return
31	}
32	
33	// we return the results. I use the index to find my original object stored in `emojis` because it's simpler. Optimisation possible.
34	for _, result := range searchresults.Hits {
35		numIndex, _ := strconv.ParseInt(result.ID, 10, 64)
36		results = append(results, emojis[numIndex])
37	}
38	return
39}

I decided to use the NewQueryStringQuery as it allows for many options of searching directly from the query string. I will be able to search within fields or add modifiers. I use it a lot for my other purposes; it might be less useful here, but it does not cost a lot, so I kept it.

Open your mind and imagine a clip-show of me adding the above code in my program and creating the interface to send the query and display the results.

🧋 Fuzzy search

Cool, the results seem promising. But there seems to be a problem.

I should have an emoji here, 🤗 to be exact. If I add the s to the query, it finds it, but not without it. Let’s try to enhance the search for this kind of purpose by adding a bit of fuzziness to the search.

The idea is to allow some inexact words to match the query. For that, we will use what’s called Levenshtein distance. In fact, as I already said, I’m a slacker, so Bleve will do it for me. Unfortunately, I could not find any way to add a default fuzzy parameter to query search. I still can add ~ to words of my search to enable fuzzy search on them. As my expectations are not high, I will do it the “just hack it” way. If I have no results from the query, I will do a specific fuzzy search instead.

 1func Search(q string) (results []EmojiDescription) {
 2	if index == nil {
 3		// no Index mean indexEmojies was not called yet or did not finished. No results (boot process)
 4		return
 5	}
 6	// we create a query as bleve expect.
 7	query := bleve.NewQueryStringQuery(q)
 8	// we define the search options and limit to 200 results. This should be enough.
 9	searchrequest := bleve.NewSearchRequestOptions(query, 200, 0, false)
10	// we do the search itself. This is the longest. Approximately few hundreds of us 
11	searchresults, err := index.Search(searchrequest)
12	if err != nil {
13		logrus.WithError(err).Error("Could not search for an emoji")
14		return
15	}
16	
17	// If we have no results we try to do a basic fuzzy search
18	if len(searchresults.Hits) == 0 {
19		// this time, we create a fuzzy query. The rest is the same as before. CopyPasta style. 
20		fuzzyQuery := bleve.NewFuzzyQuery(q)
21		searchrequest := bleve.NewSearchRequestOptions(fuzzyQuery, 200, 0, false)
22		searchresults, err = index.Search(searchrequest)
23		if err != nil {
24			logrus.WithError(err).Error("Could not search for emoji")
25			return
26		}
27	}
28	// we return the results. I use the index to find my original object stored in `emojis` because it's simpler. Optimisation possible.
29	for _, result := range searchresults.Hits {
30		numIndex, _ := strconv.ParseInt(result.ID, 10, 64)
31		results = append(results, emojis[numIndex])
32	}
33	return
34}

This time, I have my hugs emoji when I search for hug. I also have other results, but that’s fine for me. I don’t expect to have only one result and picking the right one as long as it is visible on screen with few to no scrolls is ok for me.

side note: I could have also used prefix search, but I don’t always search using the beginning of an emoji’s name, so I prefer using the fuzzy search.

🟪 Skin tones

If I search for ok hand I find the emoji, right? But as you can see, there is only the standard variation - the yellow one. I would like to have the skin tone variation as well.

Open your mind again and imagine a narrator with a deep voice popping in your head and saying the following: “Zed did not know how hard it would be to include those fancy skin toned emojis. Hours would pass before he finally understands.”

Before continuing, I need to explain (what I understood) how emojis and UTF-8 works for skin tone variations. UTF-8 characters can be composed together. This allows to create what we call ligatures. Basically, you take the code of two characters, and you smash them together into one UTF-8 character. On your screen, if your reader and font are compatible, you will see a different character in place of the two characters. The beauty of it is that even if your reader or font are not compatible, you will still see the original characters anyway. Cool, right?

Emoji skin tone is handled like this. You “smash” together the original emoji and the skin tone color to create a new emoji of the original one, but the yellow is replaced by the skin tone.

1F44C                                                  ; fully-qualified     # 👌 E0.6 OK hand
1F44C 1F3FB                                            ; fully-qualified     # 👌🏻 E1.0 OK hand: light skin tone
1F44C 1F3FC                                            ; fully-qualified     # 👌🏼 E1.0 OK hand: medium-light skin tone
1F44C 1F3FD                                            ; fully-qualified     # 👌🏽 E1.0 OK hand: medium skin tone
1F44C 1F3FE                                            ; fully-qualified     # 👌🏾 E1.0 OK hand: medium-dark skin tone
1F44C 1F3FF                                            ; fully-qualified     # 👌🏿 E1.0 OK hand: dark skin tone

The first column contains the UTF-8 code for each emoji. As you can read, the first one is the same and is the ok_hand emoji itself. The second code is for the skin tone, so we have a list of each available skin tone.

1	tones := map[string][]rune{
2      "light skin tone" : []rune("\U0001F3FB"),
3      "medium-light skin tone" : []rune("\U0001F3FC"),
4      "medium skin tone" : []rune("\U0001F3FD"),
5      "medium-dark skin tone" : []rune("\U0001F3FE"),
6      "dark skin tone" : []rune("\U0001F3FF"),
7	}

The original library, and most libraries I’ve seen, handle emojis as strings and manipulate them with string methods using the \Uxxxxxxxx format. Golang has a type dedicated to UTF-8 character management: the rune. I decided to use it. Unfortunately, there are not many examples online of rune usage, especially with ligatures. I used the string representation to easily make the connection between the UTF-8 code and the runes in Go.

Now, we need to create a new emoji for each skin tone. Not all emojis can support skin tone. I could parse the original Unicode file, but if you paid attention, the JSON file I fetched already has this information.

 1func enhanceEmojiListWithVariations(list []EmojiDescription) []EmojiDescription {
 2	tones := map[string][]rune{
 3        "light skin tone" : []rune("\U0001F3FB"),
 4        "medium-light skin tone" : []rune("\U0001F3FC"),
 5        "medium skin tone" : []rune("\U0001F3FD"),
 6        "medium-dark skin tone" : []rune("\U0001F3FE"),
 7        "dark skin tone" : []rune("\U0001F3FF"),
 8    }
 9	for _, originalEmoji := range list {
10		// we only add variations for emoji that supports it
11		if originalEmoji.HasSkinTones {
12			// we do it for every skin tone
13			for skinToneName, tone := range tones {
14				// we make a copy of the emojiDescription
15				currentEmojiWithSkinTone := originalEmoji
16				
17				// This is the important bit that took me hours to figure out
18				// we convert the emoji in rune (string -> []rune). An emoji can already be composed of multiple sub UTF8 characters, therefore multiple runes.
19				// we append to the list of runes the one for the skin tone.
20				// finally, we convert that in string using the type conversion. Using fmt would result in printing all runes independently
21				currentEmojiWithSkinTone.Emoji = string(append([]rune(currentEmojiWithSkinTone.Emoji), tone...))
22				
23				// we adapt the description and metadata to match the skin tone
24				currentEmojiWithSkinTone.Description = fmt.Sprintf("%s %s", currentEmojiWithSkinTone.Description, skinToneName)
25				aliases := []string{}
26				for _, alias := range currentEmojiWithSkinTone.Aliases {
27					// we update all aliases to include the skin tone
28					aliases = append(aliases, fmt.Sprintf("%s_%s",alias,strings.ReplaceAll(strings.ReplaceAll(skinToneName,"-", "_")," ", "_")))
29				}
30				currentEmojiWithSkinTone.Aliases = aliases
31                // I cleared the unicode version because some emoji with skin tone were added way after their original. I could parse the unicode list,
32				// but I'm a loafer, so I did not.
33				currentEmojiWithSkinTone.UnicodeVersion = "" 
34				// we add the new emoji to the list
35                list = append(list, currentEmojiWithSkinTone)
36			}
37		}
38	}
39	return list
40}

The key 🔑 here is this line:

1currentEmojiWithSkinTone.Emoji = string(append([]rune(currentEmojiWithSkinTone.Emoji), tone...))

I’m no expert in Go, and I probably missed something, but after hours of playing with fmt to print the emoji with ligatures and failing, always two characters were displayed. I inadvertently used type conversion, and it worked. I have no idea why it took me two 🖕🏿 hours to do it.

We now have our skin tone variations! 🎉

🚫 Unsupported Emoji

My computer and my phone do not support Emoji version 14 and higher well. But, as I said earlier, the beauty of UTF8 ligatures is that even if you cannot process them fully, you get the characters. This way, you can still understand the intention.

If you want to test it yourself and tinker with it, you can find the full working example in this repository git2.riper.fr/ztec/emoji-search-engine-go.

You can also test and see the final result. All details here: poulpe.ztec.fr

⁉️ Why did I do all of that ?

First of all, why not. Just playing with emoji is fun, sort of. But mostly, my goal was to have on hand an emoji finder to easily copy them elsewhere. Every system I found online had flaws that were annoying.

• way too slow to load or search
• way too useless. I don’t want to search for my emoji in a huge list of yellow icons

The best solution I had was a shortcut to the unicode list itself. Not all common names are present, so I had to learn some official names. The big reason I decided to build my own search “engine” on that is “unicode website was down!” ⬇️

Yeah, I might be the only one in the world that knows that the unicode website sometimes does not respond, and be impacted by that! 🤣

⏭️ What’s next ?

This engine is really simple and basic. There are already ways to improve it. I have already included a reverse search, even if I did not talk about it here. The indexing engine is somewhat powerful and cool, but still misses some obvious cases. I’ll see what’s bothering me the most upon usage and improve it. Maybe the result will be open-sourced, but for that, I have a lot to do. Emojis are not the only things I have in my little search engine. 😉

Thank you reading this,
Bisoux 😗

On my desk at home, while working remotely and doing a video call, I turn on my webcam. I find it more pleasant for others to be able to see me. I also appreciate it when others do the same, but everyone has their preferences on this subject.

To ensure that my image is clear and clean, I installed a lamp that shines a lot of lumens on my face. This way, I’m sure I’m visible! I used to forget to turn it on or worse, turn it off. So, I automated the process.

The lamp I use is a Philips Hue Play, which can be controlled through the entire Philips Hue ecosystem.

I also have a tablet on my desk with big buttons to turn the lamp on and off. It’s convenient, but it still requires an extra gesture.

##API to turn on the lamp

Philips Hue lamps can be controlled through the bridge, which has a fairly simple REST API.

After some research on the internet, I built a script snippet to turn the lamp on or off:

 1#!/bin/bash
 2HUE_BRIDGE_IP="10.20.0.4"
 3HUE_USER_NAME="secretCodeYouGetByPressingTheBridgeButton"
 4
 5echo "Turning off the light"
 6curl --insecure -X PUT -d '{"on": false}'  "https://${HUE_BRIDGE_IP}/api/${HUE_USER_NAME}/lights/4/state"
 7
 8
 9echo "Turning on the light"
10curl --insecure -X PUT -d '{"on": true}'  "https://${HUE_BRIDGE_IP}/api/${HUE_USER_NAME}/lights/4/state"

To access the API on the bridge itself, this page describes the steps to follow: https://developers.meethue.com/develop/get-started-2/

There is a process to obtain the “username” which serves as a secret code. To put it briefly, you need to send the following request:

1curl -X POST -d '{"devicetype":"nom_de_votre_script"}'

Quickly press the button on the bridge and start again. The code is then returned, and I write it down, notably in the HUE_USER_NAME variable. This code must then be inserted in the URL between /api/ and the “resources” path.

Next, I had to understand how the API works. In summary:

• There are “resources” that represent the scenes that we have configured (via an application, for example). The scenes define the desired state for each lamp in the scene.
• There are “resources” for each lamp to modify them.
• There is, of course, the possibility to retrieve the list of scenes and lamps.
• Applying a scene involves obtaining the list of desired states for each lamp and making a call for each with the scene values.

I didn’t go further. There are probably subtleties for dynamic scenes or certain features. After some trial and error, I discovered that the number of my lamp is 4.

Webcam usage detection

I did some research on the internet, and I don’t quite remember where I found the answer. To detect if my webcam is being used, I just need to run lsmod and “look” at the status in front of uvcvideo. It looks like this:

1IS_CAM_IN_USE=$(lsmod | grep uvcvideo|head -n 1|awk '{print $3}')

This command outputs a 0 or 1. If there’s more than one webcam, it’s likely that the method needs to be adapted a bit.

Webcam settings

With the light on, no need to let the webcam adjust it setting itself. I can set it how I want with fixed exposition time, sensibility and white balance. This way, I won’t look sick because the white balance is wrong.

 1v4l2-ctl --set-ctrl exposure_auto=1
 2v4l2-ctl --set-ctrl exposure_auto_priority=0
 3v4l2-ctl --set-ctrl exposure_absolute=120
 4v4l2-ctl --set-ctrl white_balance_temperature_auto=0
 5v4l2-ctl --set-ctrl white_balance_temperature=3500
 6v4l2-ctl --set-ctrl brightness=200
 7v4l2-ctl --set-ctrl contrast=110
 8v4l2-ctl --set-ctrl gain=170
 9v4l2-ctl --set-ctrl power_line_frequency=1
10v4l2-ctl --set-ctrl zoom_absolute=100

Those are my settings. Play with them to find your best on your hardware.

Final assembly

To build the final script, we just need to put everything together

 1~/bin/auto-cam.sh
 2#!/bin/bash
 3HUE_BRIDGE_IP="10.20.0.4"
 4HUE_DEVICE_TYPE="linuxauto"
 5HUE_USER_NAME="AH1gq0t6MlKDi4cVYQuxaD-35vgq-ves2W57Tz4G"
 6
 7IS_CAM_IN_USE=$(lsmod | grep uvcvideo|head -n 1|awk '{print $3}')
 8STATE_FILE="/tmp/hue_state"
 9HUE_STATE=$(cat ${STATE_FILE})
10
11if [ "${IS_CAM_IN_USE}" == "0" ]; then
12	echo CAM NOT IN USE;
13	if [ "${HUE_STATE}" != "0" ]; then 
14		echo "Turning off the light"
15 		curl --insecure -X PUT -d '{"on": false}'  "https://${HUE_BRIDGE_IP}/api/${HUE_USER_NAME}/lights/4/state"
16		echo "0" > ${STATE_FILE}
17	fi
18
19else
20	echo CAM IN USE;
21	if [ "${HUE_STATE}" != "1" ]; then
22		echo "Turning on the light"
23		curl --insecure -X PUT -d '{"on": true}'  "https://${HUE_BRIDGE_IP}/api/${HUE_USER_NAME}/lights/4/state"
24		echo "1" > ${STATE_FILE}
25		v4l2-ctl --set-ctrl exposure_auto=1
26		v4l2-ctl --set-ctrl exposure_auto_priority=0
27		v4l2-ctl --set-ctrl exposure_absolute=120
28		v4l2-ctl --set-ctrl white_balance_temperature_auto=0
29		v4l2-ctl --set-ctrl white_balance_temperature=3500
30		v4l2-ctl --set-ctrl brightness=200
31		v4l2-ctl --set-ctrl contrast=110
32		v4l2-ctl --set-ctrl gain=170
33		v4l2-ctl --set-ctrl power_line_frequency=1
34		v4l2-ctl --set-ctrl zoom_absolute=100
35	fi
36fi
37echo "END" 

I added a state file to remember whether the lamp is on or off. This is not mandatory, but it saves me from making an API call every second. In fact, when I start my computer, I run the following script, which simply runs the previous script every second.

1~/bin/cam-watcher.sh
2#!/bin/bash
3while true; do
4	~/bin/auto-cam.sh
5	sleep 1
6done

And voilà! When I use my webcam, the light turns on automatically. It resumes the last setting I used. Therefore, I can still adjust its brightness with the tablet (because sometimes it’s too intense for my sleepy morning eyes).

About de Philips Hue

I fell into the Philips Hue trap a few years ago, but I am satisfied with it - probably a Stockholm syndrome. I haven’t tested other brands, including Elgato for my webcam lighting. So, I don’t know if it’s easy to control them without installing a spy app on your phone. Regarding Philips Hue, everything is done locally, and the APIs are relatively easy to access. This makes it easy for the community to build around it. I was pleasantly surprised by how little time it took me to create this script. Writing this article was probably longer.

Thank you reading this,
Bisoux 😗

• Giacomo Smith - Alto Saxophone, Clarinet
• Joe Webb - Piano
• Anders Fjeldsted - Double bass
• Snorre Kirk - Drums

ⓒunknown

First Aid Kit

ⓒunknown

• Avishai Cohen - Double bass
• Guy Moskovich - Piano
• Roni Kaspi - Drums

ⓒunknown

TL;DR:

Ext4 partitions reserve 5% of the volume to ensure enough free space is always available. I have tuned my ext4 filesystem to reduce reserved space from 5% to 0.05% to maximise usage on a multi Terabytes volume. For that I use the following command

1sudo tune2fs -m 0.05 /dev/vda

I have NAS

I have a personal NAS that I manage myself. No TrueNAS or anything pre-made. I manage it myself on a computer. It’s a Debian based distribution (proxmox to be exact) with hardware RAID drives attached.

I have a big LVM volume with an ext4 partition where I put all my files inside. I have this exact volume for 3-4 years now. It grew whenever I add drives to it. From few TeraBytes it is 24To large now.

I have a monitoring in place to follow its usage. And you guessed it, empty drive space does stay empty very long.

As I’m not able to grow any Simoleon Tree at my place. I cannot add drive endlessly.

I perform some cleaning on a regular basis to try to keep the used space under control. I mostly delete anything not relevant anymore. From Movies files to project dataset I abandoned I can sometimes free multiple Terabytes of data.

Cleaning day

Today was cleaning day. After doing the usual stuff, I go on the recycle bin to permanently delete some dangling files. I do a df on the server to check free space after that and find odd values.

Who did steal my 1.2To ?

I check with a reliable calculator (my brain isn’t) and yes: 24-20=4 not 2.4. I know there is always some loss in filesystem to hold for maps, journals, or whatnot. But 1.6To seems excessive.

After some search on the great library of Internet, I found that ext4 filesystem, by default, reserve a small portion of any partition made to ensure there is always free space on a volume.

This seems at the utmost importance for system partitions because no space mean no shell. Believe me, I’m old enough to have worked on ancient system that did not have any protection in this regard, and it was a real challenge to get a shell on those computers. (considering you do not have physical access to it, of course)

Another usage of this reserved space is to fight against fragmentation. As I’m running on HardDrive and not SSD, Fragmentation is still a thing I need to consider. My volume is for cold storage mostly. I have some hot files that change a lot, but they represent a minimal amount of the total data on the drives. Fragmentation should be low.

Get my terabyte back

As I don’t have a real usage of this reserved space, It is safe for me to remove it or at least reduce it a lot. So let’s go. Simply run the following command and done

1sudo tune2fs -m 0.05 /dev/vda

And voilà !

So what ?

I manage my NAS by hand, and learn with it. Today I learned. Goal achieved.

• Now there is safety to avoid being locked away from a servers (when you have credentials)
• Ext4 reserve 5% of a volume and this is not really useful if the volume is not used as root, or fragmentation is not a concern.
• e4defrag exist and can be used to report on fragmentation and de-fragment a volume, but is really really really long.

Thanks reading me,
Bisoux 😗

I take a lot of picture, and 2022 does not make any exception. Most of them stay private, but I like sharing a few. You can see them on this post, and on Flickr. I will post on social platforms the one I like the most.

There is no rules either regarding topics, groups or themes. This page will be updated to include new pictures each time I feel it.

Le’ts go

Plants

I find it really hard to take pictures of plants that are relevant and pleasant to the eye. Often, the light is hard to catch, and the dynamic make it hard to capture the scene, moreover when I only have my phone on me.

Photos de musee

Philharmonie de paris

I really love this building. It always catches my eye. You can enjoy it by being close or far away. It’s always a different expérience. Reflections always amaze me and I really like the way the tiles on the walls are shaped. It creates mouvements following curves, shaping the reflections in all directions.

Concerts

Often, Concerts pictures are awful. Not because the subject or composition is bad, but mostly because it needs good hardware to have decent results. Sometimes, even if the file quality is far from perfect, results feel good anyway.

Divers

Paris is beautiful

I made dedicated an article on some pictures of Paris. You can follow this link to watch more Pictures of Paris I took this year (2022).

Thanks reading me watching my pictures,
Bisoux 😗

This page may be updated with new pictures. To get the news, follow me on twitter.com or mamot.fr

Rag’n’Bone man

• Rag'n'Bone man

Mathilda Homer

• Mathilda Homer - opening act

ⓒunknown

Spoiler: I regret buying them. Its trash. It had one job, and failed at it. No good points can make it ok as long as no music reach my ears.

Why Nothing ear (1)

First, let me explain how I have chosen them and why. I’m not a wireless guy usually. I still use wired headphones and earbuds, and will probably continue for a long time. With covid, and mask, it was cumbersome to put on and off earbuds because of the mask. It’s not an existential issue, I give you that, but it was enough for me to look for wireless earbuds anyway.

My key buying points

• Small, not creating a big visual ball in ears (Unlike some sony’s propositions)
• In-ear design as I always preferred this for noise isolation and quality (to lower the volume)
• Does not require to sel organs to buy it
• Ultimately, comfortable enough but this is really hard to consider without testing them myself beforehand.

After some searches I ended up with few candidates with two main eligible categories

• Good but not so expensive earbuds (around 100€)
• Really cheap (between 9€ and 40€ shipping included )

I’m able to afford ~100€ for this and I think that the really cheap ones might be garbage just because of the price.

More expensive felt too much for the requirements and usage I expect to have.

The final choice was the Nothing ear (1) because of the form factor, size, and quality reviews. Other candidates were also good options, and all of them was meeting requirements but ultimately the price convinced me, as the Nothing ear (1) were between 10€ and 30€ cheaper for the same features as others.

First experience: promising

Upon reception, I tested them for few days, mostly indoors at home. Quality was good enough (it was not my main concern anyway). Range was almost covering my 40m² flat, with the phone in center. Noise cancellation was good, especially covering hoover’s noise. At this point they were the perfect cleaning accessory to listen to music or podcast when doing chores.

Real usage: they are useless

At some point, I had to get outside, to go to work or just wandering around. This was the real test because it is outside I intended to use them.

To put it simply

As soon as I was outside, music started to constantly cut, degrade, cut and degrade again.

By “cut” I mean, no sound for 2-3 seconds. By degrading I mean, “mono phone quality with real bad compression artefacts” sound for 10-20s.

For you to fully understand what I just said, I made a quick simulation of what it is to experience those earbuds. You have to listen until the end to really understand. Micro drop are also a thing in real life.

The only ways I found to avoid those issues was keeping my phone into my hand, in front of me. This is unacceptable for me as the place of my smartphone is in my pocket, not around my neck. The distance between my pocket and my ears is roughly 1m ! I even tested with my phone in my interior jacket pocket, same behavior.

To put it simply, I cannot use those earbuds at all outside, moreover if I need my hand for something else.

I tested all the different latency options. No change. I tested on a one plus 5, one plus 7t Pro, and finally on my Sony Xperia one mark IV, same behaviour. I consider all those phones either top of the line or excellent and don’t consider them the source of the issues.

Unfinished product

As useless as it is, it’s not the end. I don’t know the conditions in which the product was designed and built, but it was probably terrible. There is many little things that shows that nobody involved in the making of this thing cared at all about its usage, or even used it.

It does not memorise parameter like noise cancellation level. I have to re-enable it each time I connect the ear buds. The nothing ear (1) App still display the option as enabled, but I have to manually disable it and re-enable it to make it work again. To make this worst, half the time, only one activate the noise cancellation. I don’t know if you tried this before, but it is really unconformable to have only one ear with such option activated. With or without music. To fix that I always have to put them in the box again and redo the whole process hoping it works this time.

Some gestures and touch actions cannot be disabled at all. Not a problem if they were not getting in the way. I personally disable the “Pause when removing earbud” feature, mostly because it never worked for me. Consequently, when I take them out I always accidentally activate touch actions. Annoying when removing one to speak to someone, I always end up with a pause or worse, a skip on my music app.

Transparent mode is a joke to. Really, don’t bother at this point. It is intended to hear what’s going on outside mixed with the music. In reality, music quality become bad and I don’t hear much of the outside things or people speaking (Even if it’s better than other modes).

I simply never successfully used the touch and hold feature to change the noise cancellation mode.
Maybe I do not know how to use my fingers, but it never worked when needed. I could only make it work when I did not want to, all the time I really needed them it ended up skipping or pausing the music. (one of the un-disablable touch actions). I finally simply disabled this feature at all.
That would not be an issue if buds simply kept memory of the noise cancellation options they had last time I used them.

The sound quality is good, when on the highest level. As it goes to garbage mode pretty all the time, It’s better to make a call and listen to the crappy waiting music than listen to those ear buds.

Often, one of the buds does not connect properly. It just does not connect at all. Putting them in the case again and retrying the connection always solved the issue. Not a big deal, just annoying because frequent.

All those details clearly show this product was finished in soo bad conditions that I do not wish to know how badly people were treated.

The good things

All the reviews online already point to the good things, nothing new here, but for fairness, I wish to list some of them.

When you have the chance to have the highest quality codec working, sound quality is good. I don’t expect more for a nomad listening device anyway. I personally add an EQ to lower the basses, but it’s my preference.

The case is small enough for me and do its job. The red and white color do not help me as I’m always putting the wrong one first (You know, like you have to test the 3 side of a USB plug). You may have a better brain than me tho.

The autonomy is plenty enough for me. Only had them reach the end once in all the time I had them. It was a long cleaning day!

The low latency mode still has a noticeable lag, but it’s enough when you watch a movie or a tv-show for me.

And …, that’s all. Nothing good bring the nothing ear (1) to the level of any simply functioning earbuds, unfortunately.

Conclusion: I regret buying them

It’s my first time with wireless earbuds. It’s bad. The product quality is bad, and the only feature required to be expected to works by this product is not … working at all. By itself, this is a big disappointment, and I really feel I just lost 100€

All the details of badly implemented stuff, and broken feature are soo many clues over the conditions this product was made. The nothing ear (1) is unfinished, release in an alpha state, even 6-8 month after I bought them (firmware and app updated the latest available in October 2022).

To conclude, don’t bother buying the nothing ear (1). I will personally boycott Nothing for all their futur product, because there is too much red-flags here to gamble again on another of their product.

I will get back to my wired earbuds I use for more than 20 years now, that cost 10€ and have an ok sound that just works.

Nothing ear (1), the earbuds to hear nothing!

Thanks reading me,
Bisoux 😗

ps: This is my experience and thoughts, Your mileage may vary and if you find yourself enjoying your Nothing ear (1), please continue, nothing wrong with that. Moreover, there is still a chance my earbuds were defective, but I’m not convinced of that.

TL;DR:

I just released a library to read and parse data from french electronic energy meters (linky or older models).

https://git2.riper.fr/ztec/go_edf_teleinfo also on github

Once upon a time

Since 2018, I have a raspberry-pi hanging near my EDF energy meter. Following some tutorials back then, I have connected them to follow my energy consumption in real time.

Without going into the details, EDF electronic meter (even before the ugly linky was imposed) have 3 connector on the right lower hand side. Two of them allow to receive constant data about the meter and the instantaneous state such as power consumption. I won’t detail how I did it because there is a lot of better content that go in details online. Just do some duckduckgo search with “EDF, teleinfo, raspberry-pi” for example.

Unfortunately, I broke all my data platform end of 2019. No mor history (visage triste). However, I never removed the raspberry-pi. So, it is still there, hanging and connected to the meter, powered down.

Meanwhilem ENEDIS the French electrical grid operator, changed my meter by one of the new Linky. The technician that installed was a bit surprised at first to see two wire connected to the meter like that. After some explanation he understood, and even re-installed them carefully on the new meter. After that, the raspberry-pi stayed there, connected, but off.

With recent european electricity price “changes”, I was curious again to see my electrical trend (Visage innocent) At this point, I don’t even know if the new meter have a working Téléinfo. I don’t know either if my code still works. Speaking of code… (tête en pleine réflexion)

I search in my old backups (thanks backups (visage heureux). You can read more about them on Borg backup) I’m lucky and find my old code. I put it in a repository, add some debug lines, compile and test int on the raspberry-pi I switch on again for the ocasion.

It works! Nothing to update, re-write. It worked on the first try. I now have logs like this:

Oct 26 19:27:31 compteur plumbus[3960]: time="2022-10-26T19:27:31+02:00" level=info msg="EDF PAYLOAD" HCHC=5605491 HCHP=12906199 HHPHC=A IINST=3 IMAX=90 ISOUSC=30 OPTARIF=HC.. PAPP=750 PTEC=HP..
Oct 26 19:27:32 compteur plumbus[3960]: time="2022-10-26T19:27:32+02:00" level=info msg="EDF PAYLOAD" HCHC=5605491 HCHP=12906199 HHPHC=A IINST=3 IMAX=90 ISOUSC=30 OPTARIF=HC.. PAPP=760 PTEC=HP..
Oct 26 19:27:34 compteur plumbus[3960]: time="2022-10-26T19:27:34+02:00" level=info msg="EDF PAYLOAD" HCHC=5605491 HCHP=12906200 HHPHC=A IINST=3 IMAX=90 ISOUSC=30 OPTARIF=HC.. PAPP=770 PTEC=HP..
Oct 26 19:27:35 compteur plumbus[3960]: time="2022-10-26T19:27:35+02:00" level=info msg="EDF PAYLOAD" HCHC=5605491 HCHP=12906200 HHPHC=A IINST=3 IMAX=90 ISOUSC=30 OPTARIF=HC.. PAPP=800 PTEC=HP..
Oct 26 19:27:36 compteur plumbus[3960]: time="2022-10-26T19:27:36+02:00" level=info msg="EDF PAYLOAD" HCHC=5605491 HCHP=12906200 HHPHC=A IINST=3 IMAX=90 ISOUSC=30 OPTARIF=HC.. PAPP=790 PTEC=HP..

I put all of that in my new data hoarding platform, and here me again with nerdy graphs.

You can now make fun of me and my outrageous energy consumption. It’s not pretty I know. I will talk about that maybe one day.

Go library

So, why do I invite you in my life like that ? I decided to publish the code that works and that I stole I don't remember where did years ago. Maybe you to build a personal home automation “platform” using go as your main language. With this library, you will have less to write I hope.

https://git2.riper.fr/ztec/go_edf_teleinfo also on github

I won’t lie to you. It is not my best code of art. First because it’s from 2018, when I started with go. Then, because I did nothing to improve it. I just publish it as is, with no guaranties. Not even Tests !!!

If future me’s Panic monster wake up some day, maybe some tests will appear, and the ENEDIS specification coverage could be increased. Good luck!

Ça fé quoi ? (what does it do ?)

The library has 3 things:

• A mean to identify the beginning and end of teleinfo data
• A mean to parse teleinfo data and validate it thanks to checksums included
• A simple struct to hold all the resulting data, that is auto-completion friendly for you loved IDE

More information on the readme.md.

What does it need ?

Basically, you need to get the teleinfo data sent by the EDF meter. The simplest way, is ti use the raspberry-pi UART, configure it with the proper parameters then open and read the interface in you program.

/!\ WARNING /!\ /!\ WARNING /!\ /!\ WARNING /!\ /!\ WARNING /!\ 

Do not connect the raspberry-pi to the metter directly!
Please check onlines tutorials based on optocoupler, 
or use secial purpose equipment such as the one you can find for 15e

/!\ WARNING /!\ /!\ WARNING /!\ /!\ WARNING /!\ /!\ WARNING /!\ 

Finally, speaking go, it looks like this:

Extracted from Readme.md

 1fi, err := os.Open("/dev/ttyAMA0") // Open the interface. It must be already configured with correct parameters
 2scanner := bufio.NewScanner(fi) // Creating a scanner reading incoming data from interface
 3scanner.Split(go_edf_teleinfo.ScannerSplitter) // Adding a "content splitter" to identify each teleinfo messages
 4for {
 5    for scanner.Scan() {
 6        teleinfo, err := go_edf_teleinfo.PayloadToTeleinfo(scanner.Bytes()) // Reading the latest packet  
 7        if err != nil {
 8            fmt.Printf("ERROR %s. %#v\n", err, teleinfo)
 9            continue
10        }
11        fmt.Printf("EDF TELEINFO PAYLOAD %#v\n", teleinfo) // You can now use this data as you wish
12    }
13}

On my side, data are sent in a prometheus collector. I configured the prometheus server to scrap metrics every 5 seconds

That look like this

 1teleinfo, err := go_edf_teleinfo.PayloadToTeleinfo(scanner.Bytes())
 2if err != nil {
 3    services.GetLogger().WithError(err).Error(err)
 4    continue
 5}
 6edfPAPP.Set(float64(teleinfo.PAPP))
 7edfPAPPHistogram.Observe(float64(teleinfo.PAPP))
 8edfIINST.Set(float64(teleinfo.IINST))
 9edfIINSTHistogram.Observe(float64(teleinfo.IINST))
10edfHCHC.Set(float64(teleinfo.HCHC))
11edfHCHP.Set(float64(teleinfo.HCHP))

Contributions ?

If you want to participate and fix my ugly code improve the library, please do. Open a PR on github and you will receive tons of virtual hugs as thanks.

If anyone on earth use this code one day, coool. If not, future me will anyway. he will probably break everything again, and start from scratch again like today, in 5 years. (I suspecte he does not learn from mistakes)

Thank you reading this,
Bisoux 😗

You are a developer, and you heard of Apache Kafka, the message broker, and you find it probably complex and difficult to work with. Today, I will try to give you the basic knowledge about kafka you need to know to be able to work with it on a daily basis.

This page is what I compiled over the years as my simplified view of Kafka and I often use it to explain Kafka to my pairs.

This Post contains lots of content from already great articles written by others. If you want to refer to the original materials, you can skip to the end and watch the “source” section. I’ve put them all.

Basic knowledge about Kafka

In order to correctly use Kafka as a developer, you must know and understand a bit about how it works externally AND internally.

Remember that this post is simplified and skip some details to make things simpler. If you have a doubt about a detail, you can check the official documentation which is clear and exhaustive: https://kafka.apache.org/documentation/

What is Kafka?

Kafka is a message broker. Its goal is to receive messages and pass them as is to something else. Nothing more. It is also associated to “stream” processing as it can do it in real-time.