Jean-Baptiste "JBQ" Quéru

I'm JBQ / Djaybee, Husband, Immigrant, Veteran, Highly Sensitive Person #HSP. He/Him. I write about tech and other things. I'm fluent in French and English, I live in Greece, previously France, UK, USA.

I like:

#skiing, #hiking, #biking, #games, #photography, visual #astronomy, #RetroComputing, #DemoScene, #PixelArt, #painting, #knitting, #weaving, #crochet, #BeadWeaving, #CrossStitch and #BlackWork embroidery.

Profile: photo of me
Header: logo of the MegaBusters
Auto-delete: 1 month

CodeBerg: https://codeberg.org/jbqueru
LinkedIn: https://www.linkedin.com/in/jbqueru/
Mastodon in French: https://mastodon.top/@jbqueru
Demozoo: https://demozoo.org/sceners/35903/

Public Key

npub1yxvlxqtpdy5plkt6fwtcy05hm8u36tlj8tksuevenswjp0vxajysunlgas

NIP-05 Address

[email protected]

Profile Code

nprofile1qqszrx0nq9skj2qlm9ayh9uz86tan7ga9ler4mgwvkvec8fqhkrwezgprpmhxue69uhhyetvv9ujumt0d4hhxarj9ecxjmntqxwaqg

Publishing to

relay.momostr.pink

Author Public Key

npub1yxvlxqtpdy5plkt6fwtcy05hm8u36tlj8tksuevenswjp0vxajysunlgas

Show more details

Published at

2026-05-24T11:03:44Z

Kind type

0 Metadata

Event JSON

{ "id": "e10ee2379a5116021a8e0099bfb9f9fccb5b9522935761dd02f116f8a3689ab7", "pubkey": "2199f3016169281fd97a4b97823e97d9f91d2ff23aed0e65999c1d20bd86ec89", "created_at": 1779620624, "kind": 0, "tags": [ [ "t", "photography" ], [ "t", "painting" ], [ "t", "hiking" ], [ "t", "knitting" ], [ "t", "crochet" ], [ "t", "games" ], [ "t", "PixelArt" ], [ "t", "RetroComputing" ], [ "t", "astronomy" ], [ "t", "biking" ], [ "t", "weaving" ], [ "t", "crossStitch" ], [ "t", "demoScene" ], [ "t", "hsp" ], [ "t", "skiing" ], [ "t", "blackwork" ], [ "t", "beadweaving" ], [ "proxy", "https://floss.social/users/jbqueru", "activitypub" ], [ "L", "pink.momostr" ], [ "l", "pink.momostr.activitypub:https://floss.social/users/jbqueru", "pink.momostr" ], [ "-" ] ], "content": "{\"name\":\"Jean-Baptiste \\\"JBQ\\\" Quéru\",\"about\":\"I'm JBQ / Djaybee, Husband, Immigrant, Veteran, Highly Sensitive Person #HSP. He/Him. I write about tech and other things. I'm fluent in French and English, I live in Greece, previously France, UK, USA.\\n\\nI like:\\n\\n#skiing, #hiking, #biking, #games, #photography, visual #astronomy, #RetroComputing, #DemoScene, #PixelArt, #painting, #knitting, #weaving, #crochet, #BeadWeaving, #CrossStitch and #BlackWork embroidery.\\n\\nProfile: photo of me\\nHeader: logo of the MegaBusters\\nAuto-delete: 1 month\\n\\nCodeBerg: https://codeberg.org/jbqueru\\nLinkedIn: https://www.linkedin.com/in/jbqueru/\\nMastodon in French: https://mastodon.top/@jbqueru\\nDemozoo: https://demozoo.org/sceners/35903/\\n\",\"website\":\"https://floss.social/@jbqueru\",\"picture\":\"https://cdn.masto.host/floss/accounts/avatars/114/407/568/810/730/622/original/cb2f94b5fcf3c04c.jpg\",\"banner\":\"https://cdn.masto.host/floss/accounts/headers/114/407/568/810/730/622/original/0aa4a4082037fe13.png\",\"nip05\":\"[email protected]\"}", "sig": "1063e990f51eb39a0ded1f9b6b0044eadf250a5617086700e41538ec04980c5acfea2609f160a599610a50643aa8dba331519dde737086e713484ac72c92b16a" }

Last Notes

2026-01-29 16:35:54 GMT

- reply

And, in case you want to see broader examples, companies use that approach against employees. Disallow something that everybody does, that everybody is comfortable doing because they see everybody else doing it... and that's convenient grounds for firing employees you don't like.

2026-01-29 15:42:40 GMT

- reply

Unless you can make the multiplexing really transparent to the underlying code, you'll probably find that the induced constraints make the rest of the code harder to design, to write, to test, to iterate, to enrich, and that the resulting game might suffer as a result.

2026-01-17 07:34:17 GMT

- reply

What if the science and engineering are invisible when done well, and only shine by their absence?
I know systems that don't scale, I know systems that collapse when you look at them sideways, for respective lack of science and engineering in the process that created them.

2026-01-16 20:01:09 GMT

- reply

Oh, I'm sorry, I'm still trying not to get left behind the blockchain bandwagon.

2025-12-28 17:30:50 GMT

- reply

Doors on the HHGTTG Heart of Gold.

2025-12-19 04:04:25 GMT

- reply

The restrictions on Y make it difficult to inc/dec the YX pair, in turn implying that memory is organized in 256 segments of 256 bytes, with a notion of near and far pointers like on x86. If you could at least swap X and Y (through register renaming) the issue would disappear.
What flags are set by INC X?

2025-10-28 05:34:23 GMT

Now that major distros have released upgrades, it's time to ask again:
I'm looking for a RISC-V board that works with Debian 13 or Ubuntu 25.10.
-really supported by the distro in question, not by some random downstream fork.
-installable via USB, not through some custom unobtainium cable.
-with display support, at a decent enough performance to be usable as a 1080p desktop.
#RiscV #Distro #OpenHardware #SBC

2025-10-10 19:58:21 GMT

Sore throat. Negative covid, flu, strep. Isolation, hoping that I didn't contaminate my wife. The hardest part is waking up with a dry throat, it's difficult to swallow anything at that point.

2025-10-08 16:25:31 GMT

And, just like that, a ninth cat in the house.
This is Cloud, he's probably about 2 months old, he lost his mom yesterday, he's scared, he's tired, but he's now safe and loved. He's incredibly kind and loving, he didn't even try to scratch us or to escape when we were giving him a bath. He's already adopted me, he's been cuddling on me, even falling asleep on me less than 2 hours after coming home.
It took a series of oddly unlikely events for us to get him, we feel blessed.
#CatsOfMastodon
https://cdn.masto.host/floss/media_attachments/files/115/339/447/500/117/605/original/dfa90500cbbb768e.jpg

2025-09-19 06:52:03 GMT

All the tidbits of news about Android make me sad.
I used to be part of the Android team.
When I worked there, making the application ecosystem as open as the web was a goal. Releasing the Android source code as soon as something hit end-user devices was a goal. Being able to run your own build on actual consumer hardware was a goal.
For a while after I left, there continued to be some momentum behind what I had pushed for.
But, now, 12 years later, this seems to have all died.
I am sad...

2025-08-30 14:53:12 GMT

- reply

Old soviet joke:
"I'm looking for an obituary"
"But, sir, the obituaries aren't on the front page"
"The obituary I'm looking for will be on the front page"

2025-08-27 14:09:47 GMT

Our little Mandalena, 3 1/2 months old, after 10 days in our house, has been making herself very comfortable. She sleeps in the open, she even spent some of last night in our bedroom. She plays a lot with her stepsister, she's learning to accept and enjoy getting petted. She doesn't sit on us quite yet and doesn't let us pick her up, but it's only a matter of time.
Her big stepsister seems to be very happy about the situation, the two are becoming inseparable.
#CatsOfMastodon
https://cdn.masto.host/floss/media_attachments/files/115/101/083/268/882/228/original/0a34029ac1642a26.jpg

2025-08-11 04:32:03 GMT

- reply

And 8008 is more microcontroller than microprocessor, its stack doesn't store arbitrary data, so you can't just push the jump address and RET to it.

2025-08-11 04:28:31 GMT

- reply

That's definitely an interesting approach, though one that might be hurting branch prediction on modern processors.
8008 didn't have it, 8080 has PCHL, named JP (HL) in Z80, JMP (An) in 68000... and RISC-V has JALR, which I believe ends up being used a lot because that's the way it does jumps longer than 1MB as well as the equivalent of RET/RTS.

2025-08-10 18:49:34 GMT

Me, looking at the 8008/8080/Z80 instruction sets: "bah, why do you need conditional calls and conditional returns when the conditional jumps are already more than enough?"
Me, coding in 68000: "I really wish I had conditional calls and conditional returns, relying on conditional jumps makes the code harder to read than necessary."
I am never happy.
#RetroComputing #Assembly

2025-08-08 06:24:19 GMT

My dad, in his 70s, bought a book on Amazon, published by Amazon, to try to better learn how to use his latest camera.
Within a few pages, the book started to explain features that don't exist on the camera.
My dad, on his own from what I taught him, figured out that the book was AI slop.
Amazon refunded immediately, no questions asked, and did not even ask to get the book shipped back.

2025-08-07 18:21:53 GMT

- reply

Yeah, that's maddening, because it shows that They don't even understand what accessibility is; that they, at most, think of it as a matter of compliance.
The thing about compliance is that it stops right at that point where you're doing something so bad that any worse would be literally illegal, so bad that no competitor can legally do worse, i.e. building the worst possible product that's not illegal. The goal should be excellence!

2025-08-07 18:06:16 GMT

- reply

What scares me the most in your description is that there seem to be no designers involved, when in my experience at least 90% of accessibility issues are design issues.

2025-07-28 14:23:10 GMT

Most programmers: "Who wrote that shitty code?" then, after checking git history, "I wrote that shitty code 8 years ago, i had forgotten."
Me: "Who wrote that shitty code?" then, after checking git history, "I'm the only person who ever committed to that repository," then, after checking a bit more, "I wrote that shitty code yesterday morning, I don't remember ever doing it."

2025-07-20 12:07:13 GMT

- reply

Very deep inside, we need to remember that those systems were quite explicitly designed to fool humans, by being trained on data that was classified based on the way humans react to that data.
To me, a major reality check is to read through the term of use from OpenAI: in every single way they try to implicitly or explicitly absolve themselves from responsibility, they know that their systems are actually catastrophically untrustworthy.

2025-07-20 08:25:12 GMT

- reply

While I do reject generative AI by default, because I can't convince myself that the value outweighs the costs (and that's by a very large margin,) I struggle to reject all forms of AI, and especially transformative AI.
As examples, speech synthesis, speech recognition, text recognition and machine translation are a godsend for those who need it, including for purposes of accessibility; I do believe that AI brings more value into this domains than it costs.

2025-07-10 14:17:14 GMT

- reply

It's moving even further from being a social network (finding content by users like you, i.e. a C2C network) to being a social media site (i.e. finding content from a small number of large creators, i.e. a B2C network.)

2025-07-10 13:43:06 GMT

- reply

"If you had kept your CO2 within your borders, you wouldn't have to worry about us keeping our smoke within ours."

2025-07-08 07:03:21 GMT

Once again, I'm working on code from the beach. That's a benefit of living in a coastal town in Greece, I'm just a few minutes away, it takes longer to set things up than to drive.
I prefer to do that early in the morning, because it's a bit less hot. It was 26 degrees C (79F) when I got here at 8am, it's now 30 degrees C (86F) which is still OK in the shade with the coastal breeze.
Even if that's technically not as productive as sitting at my desk, it surely is more pleasant.
#Greece #Beach
https://cdn.masto.host/floss/media_attachments/files/114/816/294/409/553/130/original/6c9f2298e3160214.jpg

2025-06-28 07:01:37 GMT

- reply

The Cathedral and the Bazaar.
Yup, we've had those words since 1997.
Essentially, proprietary development is a cathedral model by nature. Therefore, that's what all non-FLOSS corporate development looks like, even when such development relies of Open Source code.
What we're seeing is an attempt by such corporate developers to force a cathedral model onto Open Source maintainers, when we now have long documented evidence that the bazaar mode is very credible.
(4/n)

2025-06-28 06:58:26 GMT

- reply

I was about to settle on "secretive" and "transparent" development, wanting to say that libxml2 bug reporters are requesting a secretive development model, while the maintainer instead explicitly chose a transparent development model.
But, then, it hit me: we already have words for those models, and we have had them for decades. We even have a long writing, almost historical at this point, that documents the point.
(3/n)

2025-06-28 06:55:50 GMT

- reply

I don't want to go (yet) into a lengthy discussion about Open Source and Free Software, this'll be for another time.
Here's the deal: by requesting embargoes, reporters of those bugs are requesting that development of libxml2 happen behind closed doors. The exact model that Google used 15 years ago for its Android contributions, that the enthusiast community didn't like (to say the least).
I've been looking for words to define such practices, trying to avoid overloading existing words.
(2/n)

2025-06-28 06:51:18 GMT

Recent discussions about libxml2, and the maintainer's decision to make all bug reports public, remind me of a sitaution I've been in.
For 4 years, 2009-2013, I was the Google Tech Lead for the Open Source aspects of Android.
One of the discussions that came back often was whether Android was truly Open given that its development happened primarily between closed doors.
Back then already, we thought we didn't have good words to talk about that development model.
(1/n)

2025-06-28 05:22:29 GMT

Hahaha, I asked ChatGPT to write a tiny piece of code for me. I had written that code by hand a few days ago, in a few minutes, and it worked the first time I ran it. The ChatGPT version has 13 instructions and 3 directives, i.e. 16 lines of code.
I have to say, the ChatGPT code look credible at first glance, if you don't think about the details involved. The kind that would get a CEO to say "that can replace my coders."
So, let's look at the code. A thread.
(1/6)

2025-06-10 05:40:39 GMT

- reply

if you distribute containers, you've just become a distro maintainer, plain and simple.

2025-05-18 13:11:49 GMT

- reply

I've had a question in the back of my mind that I haven't been able to get a proper answer for, I'd love to hear your thoughts.
Does it make more sense to build such plants in generations of identical installations, or it is better to evolve each new construction a little bit from the previous one?
(I have thoughts about what makes either option better, but I can't weigh those against each other).

2025-05-08 16:15:29 GMT

- reply

My wife and I theorize that D+ intentionally had an unusually high number of reviewers, and gave those reviewers all 12 episodes, which is almost unheard of, so that those episodes can be leaked in case of censorship.

2025-05-08 14:20:47 GMT

- reply

Awesome. It's always good when there's a recommended toolchain.

2025-05-08 14:17:59 GMT

- reply

I'd probably use the minimum amount of C that allows me to switch to assembly and Open Source tools.

2025-05-08 14:10:51 GMT

- reply

You really make me want to get one once you have them ready.

2025-05-08 13:02:29 GMT

- reply

Sounds great. Don't overdesign, just go ahead.
Keep in mind that we tend to overindex on compatibility, and we often fall into a trap of freezing things too early, hoping to maintain compatibility with a relatively tiny number of existing use cases and hurting a large number of future use cases in the process.
Break early, break often, until it becomes truly impossible to maintain all of the apps that you absolutely want to be maintaining.

2025-05-08 12:56:02 GMT

- reply

Exactly. That's precisely what I'm doing for the Atari ST literally as we chat. It wasn't preemptive out of the box, and I write my preemptive stuff bare-metal, without caring for what the original ROM offered.

2025-05-08 12:53:17 GMT

- reply

Specifically, here's where it's especially messy on something like 65816: in single-task you'll obviously want the task to have access to all the RAM, while in a multitasking environment you'll probably be dynamically allocating 64kB banks and making apps deal with the fact that they don't always get the same banks. Going from dynamic allocations to static is easy, but the other way is hard.
(that's MS-DOS and far pointers vs 286 protected mode).

2025-05-08 12:48:24 GMT

- reply

Good to know about 65816 (not very familiar with it, never did SNES or IIGS), at a glance in Wikipedia there are some subtleties I don't know (256 segments of 64kiB each, is it?)
In my experience, don't try to build single-task that is binary-forward-compatible with multitask. If you want that, you need multitask right away. Link statically for now, deal with evolution later...

2025-05-08 12:39:37 GMT

- reply

Not much experience, but I'll throw some thoughts in your direction:
-can your ABI support a memcpy from any address to any address, of any size? If not, it's not enough.
-6502 is honestly a mess for any kind of multitaking, zeropage and stack are too hard to share in practice. You can assume you'll only have one program running at a time.
-build your ABI so that a second one can be implemented later (i.e. don't consume all possible resources for the first version).

2025-05-04 09:35:53 GMT

- reply

That's a good point, I'll have to try. It feels a bit weak and clunky for that, but the proof is in the pudding.

2025-05-04 09:08:08 GMT

- reply

Yeah. But browsers offer a portability that apps don't have. E.g. nobody on the list seems to have a linux ARM app.
I don't need E2E, though I'll use it if it doesn't get in the way.

2025-05-04 08:56:56 GMT

- reply

Yup. That's where I started, but the IM category is quite sparse, even more so around treating web like a first-class client.
There are more options for email, for sure.

2025-05-04 04:47:51 GMT

I'm currently considering switching to EU-based (or EEA or EFTA) providers for some of my communications.
I have the following requirements:
-For email, a segmented Inbox like that of Gmail, that sorts my emails in categories without having to set up filters, and can learn from my preferences.
-For instant-messaging, desktop browser + Android app, and being easily able to use multiple clients and to see my conversation history in all of those clients.
I'm looking for recommendations.

2025-05-02 14:55:52 GMT

- reply

It happens to me every time I am not 100% sure that I can get that code to work right away, because I don't trust my own debugging abilities. And, more often than not, it works the first time, because my brain has been thinking about that specific issue a lot by the time I actually write the code.

2025-04-30 06:42:24 GMT

- reply

My instinct is option 3, recognize the packet format in the interrupt handler and switch tasks when a whole packet has been received. I feel that'll keep the interrupt handler short and simple, but also save 2/3 of task switches when the mouse moves, but I'm willing to be convinced otherwise.
Thoughts?
(4/4)

2025-04-30 06:37:00 GMT

- reply

(3) I could recognize the packet lengths in the interrupt handler, so that I buffer all the bytes that make up a packet and wake up the task only when there's a whole packet.
(4) I could process the semantics of the packets in the interrupt handler, e.g. clipping the mouse coordinates to the screen size, and only wake up the task when something interesting happens. That's a lot of complexity in the interrupt handler.
(3/n)

2025-04-30 06:34:49 GMT

- reply

(1) I could have the interrupt handler simply wake up the task, and handle the serial chip from there. It makes the interrupt simplest and shortest, but causes a task switch for each byte (plus for each transmission error).
(2) I could handle the details of the serial controller in the interrupt handler, to the point of reading a byte without interpreting it, and waking up the task to process the byte. That saves a task-switch when there's a transmission error.
(2/n)

2025-04-30 06:31:17 GMT

Question for folks familiar with interrupt-level kernel programming: I am coding the interrupt handling for the keyboard/mouse/joystick controller on the Atari ST.
On my side, it's a typical serial controller. The mouse protocol, which is the one that sends the most data, needs 3 bytes for one packet.
I am trying to figure out the balance between how much to process in the interrupt handler and how much in a regular task, knowing that task-switching is quite expensive.
(1/n)

2025-04-28 20:35:30 GMT

- reply

Specifically, I've learned that "no politics" is actually a strong political statement: we don't need any discussion for the government to keep doing what it already does, but we need discussions to make the government go in a different direction. So, "no politics" is a way of advocating that the government keep doing what it does, that we maintain the status quo, even when that status quo includes an unfair power structure.

2025-04-28 20:31:32 GMT

All right, apology time: I apologize for being naïve in the way I had picked fosstodon as my Mastodon instance. In restrospect, yes, it might have been possible to predict the future from reading between the lines of the code of conduct there, but my privilege had probably shielded me from having to learn to recognize those signs. Lesson learned. I'll do better next time... and I hope that there's no need for a next time.