Colas.Nahaboo.net - Blog

Invent your own HTML tags!

2026-08-31T00:00:00+00:00

The HTML spec has had a proper way to invent your own tags since 2019. Most web developers seem to have missed it.

Want a simple way to have a word in blue? just use <ink-blue>word</ink-blue>

My migration of this site to Zola Tera 2 forced me to rethink shortcodes and ad-hoc styling, which is what prompted me to dig into §4.13 of the HTML standard for the first time. I found three pieces, all now supported by every current engine.

What?🔗

Custom elements are an old idea — Chrome shipped a v0 API in 2014, the cross-browser v1 standard followed a few years later. The pattern is straightforward: extend HTMLElement, register the class via customElements.define(). The simplest example, lifted from the spec, is a country-flag icon:

class FlagIcon extends HTMLElement {
  constructor() {
    super();
    this._countryCode = null;
  }
  static observedAttributes = ["country"];

  attributeChangedCallback(name, oldValue, newValue) {
    this._countryCode = newValue;
    this._updateRendering();
  }
  connectedCallback() {
    this._updateRendering();
  }
  get country() { return this._countryCode; }
  set country(v) { this.setAttribute("country", v); }
}
customElements.define("flag-icon", FlagIcon);

And use it as:

<flag-icon country="nl"></flag-icon>

The only rule I had to learn the hard way: the name must contain a hyphen. <flag-icon> works; <flagicon> does not. The hyphen is the spec's way of ensuring your tag can never collide with a future native one.

Three lifecycle hooks do most of the heavy lifting — connectedCallback, disconnectedCallback, attributeChangedCallback. There is also adoptedCallback (when your element moves documents) and the newer connectedMoveCallback (added so the element keeps its state when moved inside the DOM without being disconnected). That is all there is to it.

ElementInternals🔗

This is the piece that surprised me the most. Set one static field and call one method, and your custom element behaves like a real form control — it appears in form.elements, submits values, has validity, willValidate, and labels. It also carries default accessibility semantics, so screen readers treat your tag like the proper role without you attaching ARIA manually:

class MyCheckbox extends HTMLElement {
  static formAssociated = true;
  static observedAttributes = ["checked"];

  constructor() {
    super();
    this._internals = this.attachInternals();
    this._internals.role = "checkbox";
    this._internals.ariaChecked = "false";
    this.addEventListener("click", this._onClick.bind(this));
  }

  attributeChangedCallback(name, oldValue, newValue) {
    this._internals.setFormValue(this.checked ? "on" : null);
    this._internals.ariaChecked = this.checked;
  }

  get checked() { return this.hasAttribute("checked"); }
  set checked(flag) { this.toggleAttribute("checked", Boolean(flag)); }
  _onClick() { this.checked = !this.checked; }
}
customElements.define("my-checkbox", MyCheckbox);

A nice extra: this._internals.states.add("checked") plus the CSS :state(checked) pseudo-class lets you flip states from JavaScript and style them with CSS, no attribute-flipping dance.

ElementInternals has been Baseline widely available since March 2023.

Declarative Shadow DOM🔗

The piece I missed most. <template shadowrootmode="open"> inside your custom element lets the shadow tree be rendered from pure HTML — no JavaScript required for the initial render:

<my-component>
  <template shadowrootmode="open">
    <style>p { color: tomato }</style>
    <p>Hello from a shadow root</p>
  </template>
</my-component>

For static sites this is a game changer — Zola, Hugo, plain Markdown — your shadow tree ships as part of the HTML, the browser wires it up at parse time. No flash of unstyled component, no JS bundle needed.

When upgrading from HTML that contains a declarative shadow root, check ElementInternals.shadowRoot first — if it is already populated, your markup already shipped the tree and you should not call attachShadow() again.

Browser support landed between March 2023 and February 2024 — Chrome 111+, Safari 16.4+, Firefox 123+. There is also shadowrootclonable and shadowrootdelegatesfocus for cloning and focus delegation. See the Declarative Shadow DOM guide on web.dev for the details.

Closing🔗

Three pieces, all in the HTML spec, all shipped in every current browser. The only thing missing was awareness. So, awareness raised — we no longer have any excuse to keep reinventing component frameworks. :-)

KXF: The Koala XML Form

2026-08-30T00:00:00+00:00

Legacy

A blast from the past... In the last century, when we wrote XML by hand, I designed a format to get rid of one of XML multiple annoyances: as it used ordinary characters as deleimters, such as <, >, &, it means often quoting hell.

The idea: Use non-ascii-7bits characters as markup!

Use «foo» to delimit tags, instead of <foo></foo>
Use these regular elements as escape sequences instead of entities. Escape « by «-». Not really readable — but more than entities — but natural and consistent to parse.

I wrote the KXF <-> XML converter in Klone, my lisp scripting language, that I definitely need to publish here as part of these legacy series.

Here is the doc of the format:

KXF: The Koala XML Form

Purpose

The Koala XML Form XML avatar has been defined with the following design goals:

Ease writing by hand
Ease human readability
Solve only syntaxic problems
- no entities
- no DTD
- no CDATA
- straightforward whitespace handling
- get rid of most quoting hell: use less special chars, and use non ascii ones

Design

Here are some examples:

KXF	XML
«tag»	<tag/>
«tag foo <> bar»	<tag>foo < > bar</tag>
«tag(x=y z="1 2" t='gp') gee»	<tag x='y' y='1 2' t='gp'>gee</tag>
«! comments»	<!-- comments -->
«? PIs»	<?PIs?>

Quoting special chars:

Char	Quoted form	Where
«	«-»	character data contents
»	«+»	character data contents
'	«!»	attribute values
'	«/»	attribute values
"	«#»	attribute values
"	«*»	attribute values

»«/tag» optional ending forms can be used to check nesting as in
«foo a very big contents... »«/foo»
- ending contents is ignored and can be used for comments, as in:
  «foo a very big contents... »«/foo the main loop»
- attributes if present must match the starting tag attributes:
  «foo(x=1) a very big contents... »«/foo(x=1)» matches, but
  «foo a very big contents... »«/foo(x=2)» raises an error
Whitespace around sub-elements are discarded, but not around character data.
You can keep them by postfixing element name with = (it sets the xml:space attribute to "preserve"), such as
«pre=(a=b) «b x» «i y»» or by enclosing them in the pseudo node «= »
«x «= » «y» «z»» keeps space before y, but not between y and z
For comfortable use, it is recommended to assign «» to shift-keypad-minus and shift-keypad-plus respectively
Encodings are raw: either ISO-Latin-1, or UCS2, which can be autodetected as fields must begin with «<letter> (ISO-Latin-1), \0« (UCS2), «\0 (MS UCS2)

Benefits

More readable
Most text editors know how to balance « and » better than <foo> and </foo>
It is self sufficient: whitespace handling is explicit
No need to quote the common chars <, >, &

Drawbacks

You cannot use XML Tools on this format. Not really a problem, as you can convert to and from XML

Tools

For now, there exists only a parser/writer in Klone (xml-ksf.kl, used to implement Klone scripts (xml2kxf, kxf2xml) that converts back and forth to XML.

A java SAX parser is planned

Author

Colas Nahaboo

Back to XML Avatars

Magnets, magnets everywhere

2026-08-27T00:00:00+00:00

I have started putting magnets on small things around my desk. A cable that was wandering around now stays where I need it, and a USB hub now sits on the top of my PC case instead of taking space on the desk.

Cables🔗

For cables, I use round magnets with a hole in the middle. A small tie goes through the hole and loops over the cable. The cable rests on top of the magnet, and the magnet holds the whole thing to a magnetic surface.

I can move the cable, pull it away when I need it, and put it back without fighting a fixed clip. The magnet stays with the cable, so I do not have another small part to lose.

The photo below also shows another version. I bolted a double-sided Velcro band through the central hole of a round magnet. It makes a reusable strap for a thicker cable or a small bundle.

USB hubs🔗

I glued square magnets (the leftlmost ones on the photo above) made for cabinet doors to the sides of the USB hub. I used MS polymer glue and waited for it to cure. The magnets now attach the hub to the steel side of my PC case.

The hub remains removable, but it stays put. The same trick should work with metallic desk legs, provided that the metal is magnetic.

One small detail matters: a metal surface is not necessarily magnetic. I test it with one of the magnets first. I also let the glue cure fully before hanging the hub and keep strong magnets away from cards and magnetic storage.

This is the kind of DIY I like: cheap parts, no drilling, and an arrangement I can change later.

Buzz was right. Magnets, magnets everywhere. :-)

All my posts, on one page

2026-08-26T00:00:00+00:00

At last, I added the page I was missing: all my posts, ordered by date, in the same style as the blog, code and surf lists but for everything at once.

Nothing fancy technically: the RSS feed already did this for feed readers, so I just made a human-readable version of it. French twin posts are listed once, with their flags — newest first, from my first AI-assisted pull request down to my musings of 2007.

Enjoy!

My Salli saddle chair

2026-08-21T00:00:00+00:00

I am an old programmer, and a game and surf addict. I have been spending hours and hours a day in a chair for decades: my work, then my evenings of computer games, are all done sitting. My sports — surfing, windsurfing, standup paddling — are wonderful for the body in the water, but alas, they do not fix the static hours spent at a desk. Over time, my hips started to stiffen, and I realized my sessions were not enough to compensate for all that sitting.

So I decided to fix the problem where it happens: at the desk itself.

Why a saddle chair?🔗

The idea of a saddle chair is to make you sit like you stand. On a normal chair your thighs are horizontal, at a 90° angle to your body, which presses your vertebrae and disks together at the front and strains the back muscles. A saddle chair opens the hip angle to about 135°, tilts your pelvis forward, and restores the natural S-curve of the spine instead of the slouched C of a regular seat. The weight goes on your sitting bones rather than your thighs, which is what your hips and knees like. It is a bit like sitting on a bicycle saddle, but on a proper chair.

Sitting like standing: the pelvis forward, the spine in its natural S-curve.

The Salli🔗

I bought a Salli, the Finnish pioneer of the saddle chair, with its patented two-part seat and an adjustable gap in the middle — which also keeps things dry and pressure-free down there, a nice bonus.

The first days were promising for my hips. The main practical points:

You must raise your desk by 15 to 20 cm. The saddle sits you higher than a normal chair. Conveniently, the tabletop height for saddle sitting works for standing too, so a standing desk handles both.
No backrest needed. Once the pelvis is tilted forward, your back holds itself in its natural upright position and does not strain as it does slouched. I did not miss the backrest at all.
It is tiring at first. The muscles that hold you upright are suddenly working, so I could only manage about an hour the first days. But you progress fast: after two weeks it was genuinely fine, and my hips clearly appreciated it.

(my setup — a bit of a surprise to visitors, but worth it)

A note for the desk-bound🔗

If you work seated and have back or hip pain, it is worth knowing that in some places such a chair can be bought on a doctor's prescription, like mine was — so check with your doctor or employer before paying full price yourself.

Which prompted me, as usual, to think: if I am going to spend so many hours a day in a chair, the chair had better be a good one. My walking pad fights the sitting, and now the Salli makes the sitting itself far less harmful. Two sides of the same problem.

Long live not sitting like a lump!

Why a walkpad, and which one

2026-08-18T00:00:00+00:00

I never liked exercising. I love the physical effort required by surfing, windsurfing and standup paddling, but I was never able to run or swim as workouts because I get bored out of my mind so quickly.

However, over 60, I do not have the motivation anymore to SUP or surf when the conditions are not fun (chop, onshore winds, rain, too big, rips...) so I cannot rely on my sessions to stay healthy. And sitting down all day to program or play games is literally lethal.

The solution I decided on is a walking pad — a bare, mini treadmill that you can slide under your couch when not in use and walk while working at a standing desk. I can thus watch YouTube videos on my PC screen right in front of me. I now look forward eagerly to my workouts (walkouts?) as I love watching science videos (Sabine Hossenfelder, PBS Space Time, Anton Petrov, or various geek, programming and AI topics). All this for less than 200€.

(not me in the image, obviously :-) )

The name of the game is to spend less time sitting. The studies agree: prolonged sitting is bad for the metabolism and the heart, and even light walking during desk work improves the numbers (Mayo Clinic NEAT study, Cleveland Clinic). A walking pad is the lazy, home-friendly way to get some of that.

Why a walkpad, and not a treadmill?🔗

Because you do not run on a walking pad. You walk, at speeds up to jogging pace. And for walking, a full treadmill is overkill: it is bulky, heavy, very expensive, noisy, and its console full of buttons serves no purpose. A walking pad is cheaper, quieter, weighs 30-50 lbs, and fits under the couch. That is the whole point: it is there when you want it, out of sight when you do not.

And why not just walk outside? Because outside I get bored instantly, and become a meal for tiger mosquitoes. My walking pad lets me combine a workout for both mind and body, independent of the weather.

Also, the pad setting the pace for you is what makes it great compared to indoor bikes, manual treadmills, rowing machines, etc. You do not have to think about your workout.

Which one to buy?🔗

There are many brands selling what are essentially the same machine (Urevo, Sperax, DeerRun, Costway, Kingsmith, GoPlus...), and the prices range from cheap to silly. A lot of advice exists, and this reddit thread on cheap walking desks vs expensive ones gives a good feel of the issues — but read also the comments: they report feedback after months or years of use. There is also a handy summary of the offers made by a redditor.

My own advice, after research and months of daily use:

Brushless motor for reliability and silence. This is the first thing to check.
Standard Bluetooth FTMS protocol, so that any app can drive it. This is crucial for geeks like me: I wrote my own app, walkpadspeed, that drives my pad through my own routines. Without the standard protocol, you are locked into the vendor's app — often with subscriptions and bugs. With it, any good free app works.
High weight capacity is an indication of sturdiness. Check the listed max weight.
Incline matters. Being able to incline the pad is great: you get the same workout at a lower speed, so less noise, less stress on the gear. Manual incline is OK — automatic systems are heavier and bulkier to store, and not worth the tradeoff in my opinion.
6 support points beat 4 Having more support points brings greater reliability, as there is less flexing of the frame. Prioritize a solid, low-flex frame with firm rubber shock pads over a soft, springy walking surface for hardware longevity.
Good shoes and a mat give you a lot of shock absorption in any case, but is especially important with a low-flex frame. Slip-in shoes (e.g. Skechers) are especially nice if you have health issues.
On cheap models, prefer many short sessions instead of a long one, to avoid motor overheating.

A note about the "slides under the couch" selling point: sliding is actually a bit tricky, as you have to tilt most pads to use their wheels, and tilted, it does not fit under the couch. My solution was to create a sliding mat by gluing (with double-sided tape) a cheap skating slideboard under a cardboard sheet, and now the pad slides effortlessly under the couch.
Some walkpads can fold in half and do not need to be tilted to use the rollers, so they can slide under a couch more easily, but some reviews say the folding can be felt underfoot while running, and is not pleasant.

My choice🔗

For what it is worth, I am quite satisfied with my Fousae ZX-390, a simple entry-level pad. I bought it favoring mechanical qualities over sophisticated features, and it has everything I need: brushless motor, FTMS protocol, manual incline. Other brands sell similar pads.

Plus points:

it has 6 feet instead of 4, making it more durable — the deck does not flex — as the middle feet are positioned just where your feet land on the pad.
it has a lubrication hole to pour the silicone oil into the belt, instead of having to deal with the messy operation of spreading it by hand.
the motor housing at the front limits the length of your stride, compared to the rare models without it.

Minus points:

its belt is thinner than a high-end treadmill's, so walking barefoot will give you blisters fast
no automatic incline: either 0% or 9%.
its speed is limited to 6 km/h. To get to 12 km/h you have to raise the handlebar.
to work at a desk, you cannot remove the handlebar; you have to fold it, and you can hit your heels on it.

And of course, I wrote walkpadspeed to drive it, which I just released as v1 — see my announcement post. It loads routines from simple text files, and can even adjust the speed automatically to keep my heart rate in a chosen zone.

My advice🔗

You need to get used to it. Make short sessions at first, 0% incline, at very slow speeds, and raise the speed progressively each day, e.g. by 0.2 km/h.
Once comfortable with it, you can set the incline to 9% and keep it there.
Not all kinds of work are realistically doable on a walking pad. Avoid tasks that need precise mouse movements.
Maintenance is essential. Oil it regularly and avoid dust and carpet lint.
I use shoes, but walk barefoot once daily on one of my 15mn post-meal sessions. It keeps my feet in shape for surfing, where I walk a lot barefoot to get in and out of the water, but it is not enough to develop blisters or joint pains. However, you should not attempt to walk barefoot in your first month!
I do not use socks for convenience (I only wear slaps at home), I just put my shoes on a gentle air blower I use to dry my wetsuit boots, to keep them sane.

My routines🔗

Since the pad is always available, I built a small weekly program around it. Here are the routines I use the most:

Short Digestive: a 15mn relaxed walk, done 15mn after each meal (thus 3 times a day), to lower the glycemic charge — the spike in sugar, and thus insulin, in the blood. It keeps the heart rate in "Zone 0", because walking faster triggers the liver to release sugar in the blood, anticipating a need. It is also called a... Fart Walk! You can stop after just 5mn with most of the benefits, or lengthen it if you are watching a captivating video.
Z2 40mn: the recommended brisk walk, the most efficient to burn fat. Must be done on an empty stomach. Staying in Zone 2 mobilizes fat burning without the stress of a faster walk or jog that triggers a cortisol spike and prevents fat burning. It boosts endurance and metabolism. I do it 2 to 3 times a week.
Japan IWT 38mn: the famous Japanese Interval Walking Training has proven extremely effective at enhancing cardio and fitness, and is a practical solution for walking pads, which are not designed to resist actual running. I do it once a week.

Thus, my weekly schedule is 3 digestive walks after each meal, and per day of the week:

Japanese IWT
Zone 2
Resistance training
Zone 2
Resistance training
Zone 2
Rest, or resistance training if I feel like it

Resistance training for me is short (20mn) but at the highest intensity, as I am over 60.

Two hours of episodes, one workout. I would not go back.

Cooking tip: weigh all your utensils

2026-08-16T00:00:00+00:00

I have been cooking this way for years, and I am surprised it is not more common: I weigh all my pans, pots, bowls, colanders, plates, basically anything that can hold ingredients, and keep the list handy.

You know the dance. The recipe says 300g of flour, so I put the bowl on the scale and tare. Then I want to add the sugar in the same bowl, but the bowl is already on the scale with the flour in it — I do not know the tare anymore, so I do mental arithmetic, or I weigh each ingredient in its own bowl and wash more dishes than I cooked.

With the weights written down, I never need to tare: pan on the scale, read the total, subtract its weight. A 1.2kg pan showing 1.8kg total — 0.6kg of food in it, exactly. It also gives me total transparency on portions: I can weigh the whole pot of pasta and divide by the number of plates without empty plates sitting around the kitchen.

I keep the list on a post-it stuck on a shelf, but you can choose a note app (e.g. Google Keep) — any paper sheet or note app works. Just keep it up to date.

Bug reports: Favoring Prompt Requests over Pull Requests

2026-08-15T00:00:00+00:00

More and more often, I find a bug in some code that is not mine — often in a language or codebase I barely master. My reflex used to be to hack a fix anyway and send a pull request. But recently I spontaneously began to do the opposite: I let an AI agent fix the bug on a local copy, and instead of a pull request, I submitted the prompt I used — the one that seemed to fix the issue for me, passing the tests. An evolution of my AI without AI?

I did this twice, recently: for the Apollo Zola theme and the Openfox AI harness.

Which prompted me to wonder whether this practice already had a name. I found the CodeRabbit post on prompt requests, which was right on the spot, but nothing else. Hence this post, to draw attention to the matter.

What?🔗

The idea, in short: when you find a bug in a project, and you use an AI coding agent to fix it, do not submit the AI-generated pull request. Submit the prompt you used to fix the bug, with a comment that it seems to fix the bug on your setup, and passes the tests. Plus an overview of the useful info your AI investigation produced: the root cause analysis, the edge cases, the alternative paths tried.

The maintainers can then run the prompt themselves, in the harness and the LLM they are comfortable with — not the one you happened to have at hand — refine it, grill the AI with a grill-me style interrogation, and adjust it to their codebase. Instead of having the pain to review an unfamiliar PR.

This is not my invention, the name already exists: prompt requests. The CodeRabbit article traces the idea to a debate between Peter Steinberger, creator of the self-hosted AI agent Clawdbot, and Gergely Orosz of Pragmatic Engineer. At first I was going to call it "PS" for Prompt Suggest, but let's not reinvent the wheel.

Why?🔗

Maintainers are drowning in PR slop: AI-generated code that compiles, passes CI, and is utterly wrong for the codebase — style drift, redundant helpers, subtle architectural shortcuts. The asymmetric velocity problem: an agent generates code in seconds that takes a human hours to review.

The numbers are staggering: GitHub merged pull requests went from 25 million a month in January 2023 to 90 million a month in March 2026. A 3.6x increase, in an era where the number of human developers did not triple.

The most spectacular example was the 13,000-line AI-generated pull request on the OCaml compiler, rejected by the maintainers, who cited the lack of review resources. The author, Joel Reymont, had "carefully shepherded AI over the course of several days", and, asked why some files credited a maintainer as author, answered "Beats me. AI decided to do so and I didn't question it."

This is so bad that GitHub now lets maintainers disable pull requests entirely, or restrict them to collaborators. Open source projects are starting to close their front door rather than drown.

The root problem: code is now cheap. As Jeremiah Lowin writes in An Open-Source Maintainer's Guide to Saying No: "Historically, we could assume that since writing code is an expensive, high-effort activity, contributors would engage in discussion before doing the work. Today, LLMs have inverted this. Code is now cheap, and we see it offered in lieu of discourse."

Which prompted me to think: if the code is the cheap part, then the value of my contribution is not the diff — it is the investigation. The bug I found, the root cause my AI uncovered, the proof that a fix exists and passes the tests. And the prompt, which is the recipe to reproduce the investigation.

How?🔗

So here is what I think I will put from now on in a bug report instead of a pull request:

**Bug**: [short description, with a reproduction]
**Prompt** (tested, fixes the bug on my setup, tests pass):
  > [the exact prompt(s) I gave to the agent]
**Harness / model**: opencode / DeepSeek v4 flash (or Claude Code / ...)
**Root cause**: [what the AI found]
**Edge cases**: [what the AI's analysis surfaced]

The exact prompt is the key: like a minimal reproduction, it lets the maintainer run the fix in his environment — the harness and model he trusts, in the context of his codebase. He can refine it, ask the AI questions, interrogate it about edge cases. I have a grill-me skill in my own setup, I know it works :-)

Why do maintainers should like it?🔗

No trust question: the code is generated in the maintainer's own sandbox, by his own model, with his own system prompts. The question is no longer "do I trust this stranger's code?" but "does my setup produce a good fix for this prompt?"
Architecture control: the generated code follows the codebase conventions, because it is generated in the codebase.
The maintainer stays the author: he is not inheriting 800 lines of alien code, he is applying an insight to his own project.

And the contributor still gets the satisfaction of having found and fixed the bug — which is the valuable part anyway.

Am I reinventing the wheel?🔗

Partly. The CodeRabbit article is more cautious than me: it argues prompt requests and pull requests are not in conflict, and prompt requests work best before pull requests — review the intent before generating, then still review the code that ships. That makes sense within a team, where you will have to maintain the result, and where determinism, git blame and accountability matter.

My point is narrower, and for a different case: the drive-by contribution. When you fix a bug in a project that is not yours, and you have no intention to maintain the result — which is most bug reports — the prompt is a better unit of exchange than the diff. It is the AI without AI of bug fixing: I provide the prompt, you run it in your own wetware (or your own GPU).

There are cases where a PR is still better: the trivial one-line fix, the case where you will maintain the fix, or the project that explicitly asks for PRs. Use your judgment.

So, next time your AI agent fixes a bug in somebody else's project: resist the urge to git push. Open an issue, and give the maintainers the prompt. They will be grateful, and the code will be better for it.

Maybe a modern twist on the old saying "dont give fishes, show how to fish".

Remediate subpixel color fringing on Linux

2026-08-14T00:00:00+00:00

I plugged in my new MSI 491CQPX, a QD-OLED ultrawide, and the first thing I noticed was that text looked dirty: every letter bordered by a red fringe on one side, a green one on the other. Not blurry, but colored — as if my glasses had chromatic aberration.

That is subpixel color fringing. QD-OLED panels arrange their red, green, and blue subpixels in a triangle, while font rendering engines — ClearType, FreeType, all of them — assume the classic RGB stripes of LCD panels. Subpixel antialiasing is the culprit: it lights up individual subpixels to fake a higher resolution, and on a triangular layout the colors bleed sideways into a fringe.

PC Monitors has the full story, with macro photos (QD-OLED and WOLED both suffer). nuxx was so bothered that he returned his OLED — it made his eyes feel funny. I did not return mine: on Linux the fix is one setting.

The fix: grayscale antialiasing🔗

Subpixel rendering is the problem, so I stopped using it. Set your fonts to plain grayscale antialiasing and the fringes disappear. On GNOME:

Install GNOME Tweaks: sudo apt install gnome-tweaks
Open Tweaks → Fonts
Set Antialiasing to Grayscale — this is the whole fix

The letters lose a bit of the LCD-crisp sharpness, but I had to look hard to tell the difference. Hinting is a matter of taste, not part of the fix: I left it alone.

Not on GNOME?🔗

Any desktop works — the switch is in fontconfig, the config file every Linux font stack reads. Create ~/.config/fontconfig/fonts.conf with this:

<?xml version="1.0"?>
<!DOCTYPE fontconfig SYSTEM "fonts.dtd">
<fontconfig>
  <match target="font">
    <edit name="rgba" mode="assign"><const>none</const></edit>
  </match>
</fontconfig>

run fc-cache -f, log out and back in. Setting rgba to none is precisely what turns subpixel rendering off.

Black is too black🔗

The fringes are most visible at maximum contrast — white text on pure black. I keep my editor and terminal backgrounds at a very dark gray (#121212) instead of #000000, which calms things down further and looks just as "OLED" to me.

Do not try the ClearType Tuner under Wine🔗

Tempting, I know, but do not. The tuner writes Windows registry keys (Software\Microsoft\Avalon.Graphics), which FreeType and Wayland never read — zero effect. And on Windows it would not help either: the tuner cannot adapt ClearType to a triangular layout, which is why there is a long-running PowerToys issue begging Microsoft for OLED subpixel support.

Alas, the Linux advantage here is real: Windows users are still waiting for Microsoft, while I fixed my monitor with one toggle. Done.

... But still annoyed by the pixel refresh🔗

However, I still have to cope with the other trouble with OLED monitors: They just stop working for 10 minutes at least once a day... at the most inconvenmient times: the dreaded pixel refresh.

WoWLAN: waking a suspended Linux host over wifi

2026-08-11T00:00:00+00:00

What?🔗

I stumbled recently upon the fact that you can wake up a suspended Linux host over wifi, not just Ethernet. The feature is called WoWLAN — Wake on Wireless LAN — and, I must say, it took me by surprise.

For me wifi always needs some complex dance to re-connect, and I could not imagine it could work. But with WoWLAN the card keeps running in a low-power listening mode, still associated to the access point, and it can wake the whole host when it sees the right packet. It is the wireless sibling of Wake-on-LAN, with one obvious advantage: no cable.

Why?🔗

I am setting up two cheap Linux "AI servers" at home:

a discounted "Virtual Reality" Zotac "PC in a backpack" with a GPU I have been using for two years now to transcribe confcalls with Whisper.
a cheap PC with 3 GPUs of 12GB VRAM each made of spare parts and cheap second hand ones, to toy with the current ever-evolving crop of open source LLMs that are more powerful by the week.

I wanted to suspend them when not actively running AI, but wake them transparently on demand. And with wifi, this meant I could move their noise and heat easily away from my "work" room.

How?🔗

Linux exposes it through the iw tool, which is the least surprising part. Something like:

iw phy0 wowlan show
sudo iw phy0 wowlan enable magic-packet disconnect

You can wake on a magic packet, on a disconnect, on any packet, or a combination. Then put the machine to sleep, fire a magic packet from your phone or another machine, and it comes back. A phone is the natural partner here — wake your laptop from the sofa before you even walk to it.

Alas, it does not work everywhere. Not all wifi cards and drivers support it, so check the output of iw list first. Intel cards are generally well served. The card also has to stay powered during suspend, so a machine that cuts wifi power entirely in suspend will not cooperate. And some distributions or NetworkManager versions may reset the setting at boot, so you may need to re-enable it — a systemd service fixes that.

The tutorial I used, from cyberciti.biz: how to configure wireless wake-on-lan for your Linux wifi card, walks through the whole thing: finding your phy, checking support, enabling WoWLAN, and testing it with echo mem > /sys/power/state. For the low-level details, the official Linux wireless docs are the reference.

The idea that a machine with a sleeping CPU is still listening for a whispered packet over the radio — and will wake up at a shout — amazes me more than it probably should :-)

AI without AI

2026-08-08T00:00:00+00:00

Generative AI is currently morally bad, as most of it steals copyrighted works — but it would still be useful for people like me with no drawing skills. So I am increasingly using "AI without AI".

That is, I just post the prompt I would have used with a generative AI, and let the reader use his own wetware (aka "Brain") — the NI (Natural Imagination) — to visualize the picture.

As an example, in a reply to John Scalzi's post on Bluesky I described the book cover instead of generating it via AI:

And for some AI-without-AI, here is the prompt to generate the cover in your own wetware (aka "Brain"): «A book cover for the next Sci-Fi humoristic novel of the fictional author John Barefoot titled 'The Brother Laser Printer Aficionados'»

The nice thing about running prompts on your wetware is that the model is local, free, and the output stays in your head. No credits, no seed, no style references, infinite resolution, perfect privacy. It renders instantly — the only downside being that nobody else gets to see the result.

Also, it is something that has bothered me for a long time. As a boy, I was struck that during his famous 65-day survival drift across the Atlantic Ocean in 1952, narrated in his book Naufragé Volontaire, French physician and adventurer Alain Bombard — aboard his boat l'Hérétique — took only a few personal items, including a pocket score of Beethoven's string quartets (specifically the 15th String Quartet in A minor, Op. 132) alongside Bach's Passions. He recounted that studying and mentally playing this complex music helped keep him sane and motivated during extreme isolation and hardship.

I loved listening to Beethoven, but I could not fathom somebody actually "listening" by merely reading the score. It turns out he was doing AI-without-AI, ... decades before AI.

And now... start posting!

2026-08-07T00:00:00+00:00

I must say, I have been toying with the idea of posting more regularly here for a while. And now, at last, I decided myself to really start posting.

Let me reassure you right away: this is not a social network. I will not post reactions, reposts, or random thoughts the moment they cross my mind — no, I will not become yet another feed of instant takes. The last thing the world needs is my version of the Bluesky noise. What I will post is things that pertain to my actual activities, not my mood.

Which means: mostly coding and surfing, the two obsessions that occupy my days since I retired, and everything that surrounds them. The occasional silly waste of a full weekend on an obscure terminal command that only I will ever use. The existential satisfaction of a clean configuration file. A tool I built and decided to share openly.

Does this mean a post every single day? No. That would be a nice pledge, but I know myself too well to pretend I will hold that rhythm. I made that mistake once, in 2021, and I was right to be skeptical of it. So: no strict schedule, no guilt, just a sincere intention to write more, and to write about things that matter to me.

There is a quiet pleasure in it. Publishing on a corner you own, with zero algorithmic pressure and a total absence of pop-up newsletter modals, feels a bit vintage — like casting a message in a bottle into an ocean of infinite scroll. What changed is that now, this message may be found not by a human reader but an AI agent that will feed it to the giant automated ideas fisheries that has become Internet.

So, here is the promise I make to you, dear biologic or electronic — or imaginary - reader: One post every single day, until ... not.

Migration of this site to the new Zola Tera2

2026-08-06T00:00:00+00:00

I migrated this site to Zola 0.23, released yesterday with the new Tera 2 template engine (Well, 0.23.1 from today actually). When I chose Zola last March, I praised its creator for not chasing the latest fad. I stand by it: 0.23 breaks my templates, but it does it in a cleaner way.

Tera 2 removes the old macros and imports, and Zola 0.23 removes the shortcodes. Everything is now a namespaced component. A macro that I used to import and call like this:

{% import "macros.html" as utils %}
{% macro show_link(url) %}<a href="{{ url }}">{{ url }}</a>{% endmacro %}
...
{{ utils::show_link(url=link) }}

is now a global component with no import at all:

{% component site.show_link(url) %}<a href="{{ url }}">{{ url }}</a>{% endcomponent %}
...
{{<site.show_link url={link} />}}

The big philosophical change: components are global and no longer inherit the caller context. Every variable — page, section, config, current_path — must be passed explicitly as an argument. More verbose, but it makes dependencies visible. I like that more and more: one way to do it, and it is visible.

Tera 2 is also strict: a missing variable is now a hard error. My templates use optional chaining (?.) for anything optional, which catches typos at build time instead of silently rendering broken pages.

What concretely changed:

concat and slice filters are gone → array spread and slicing
get_env is gone → config.mode == "serve" for local editor links
include ... ignore missing is invalid → empty hook templates
get_page with fatal=false → canonical path with lang and allow_missing=true
get_taxonomy_url with name → term

The migration guide was my best map through all of this.

I use the Apollo theme as a git submodule, and as it does not support Zola 0.23 yet, I keep a small local template-only patch of it until upstream migrates. The design, CSS, and JavaScript are untouched: this was a syntax migration, not a redesign, and I did not rewrite my post contents either.

Use of AI🔗

I chose to have AI help me migrate my site. It worked flawlessly, so I encourage you to do the same. AI work especially well with Zola, as it is opinionated, and that means that there are clear ways of doing things so that AI agents do not get lost in rabbit holes trying to pursue all the possibilities of more flexible systems like Hugo.

For reference here is the first prompt I gave it to prepare the plan for the migration:

I want you to only create a document for agents ai/agents/tera2.md to summarize
how tera2 of Zola 0.23 differs from tera1 of Zola 0.22, and the plan to migrate
my site from Zola 0.23 to 0.22.

Examine the differences between Zola 0.22 and 0.23, not just tera.

Docs:
Tera2 https://keats.github.io/tera/
Migration guide Tera1 → Tera2 https://github.com/Keats/tera/blob/master/MIGRATION.md
Zola 0.23 docs https://www.getzola.org/documentation/getting-started/overview/

Take your time, be thourough, but do not modify yet any file of my site except
the ones in ai/agents/ and AGENTS.md

Then, plan the migration of this site to zola 0.23
1. not modifying the contents of the posts unless necessary for tera2 changes
2. not changing the look and feel, the UI of the site
3. adapting the various templates
4. changing my local copy of the Apollo theme I use if needed.
   These changes are a stopgap measure until upstream apollo is migrated to Zola 0.23

I use Deepseek v4 flash normally, but chose to use GPT-5.6 Luna for this migration as it is currently quite cheap and I trust it more to not go change unecessary files. And Deepseek as no vision, so it cannot check the actual visual results of changes. It helped me greatly for this migration, without any problems.

Summary of AI actions:🔗

Replaced Tera 1 macros, imports, self:: calls, and macro namespaces with namespaced Tera 2 components in the root templates and Apollo templates.
Converted the local language-listing, page, navigation, debug, file inclusion, and section-name helpers to components.
Converted the site shortcode implementations to components in templates/macros/shortcodes.html.
Converted active post calls from the removed Zola shortcode syntax to Tera 2 component syntax. This includes YouTube, snapshots, attachments, Mermaid, static-file inclusion, legacy overlays, obsolete notices, and SVG icons.
Passed explicit base_path values to components that used to receive page.path implicitly from Zola shortcodes.
Updated get_page calls to use the canonical path with lang="fr" and allow_missing=true.
Updated get_taxonomy_url calls from name to term.
Replaced the removed get_env call with config.mode == "serve" for the local editor links.
Replaced the removed concat and slice filters with Tera 2 array spread and slicing, and changed the object test to map.
Added empty Apollo hook templates because Tera 2 no longer accepts include ... ignore missing.
Updated dev/bin/site-check to recognize the new legacy component syntax.
Updated static/ZOLA to zola 0.23.0.

PDP: Publish, Don't Perish

2026-08-02T00:00:00+00:00

I stumbled recently upon Elizabeth Tai's POSSE and PESOS definitions:

POSSE: Publish (on your) Own Site, Syndicate Elsewhere
PESOS: Publish Elsewhere, Syndicate to Own Site

Which prompted me to think that after a first POSSE phase, toying with the emerging web landcape, I was doing mostly PESOS as I found that posting in topic-specific forums made reaching an audience much easier. But after some time, after seeing forums disappear suddenly, I began to POSSE-collect my scattered PESOS posts.

But now at 66, I am more and more focused on PDP: Publish, Don't Perish!

Why?🔗

POSSE and PESOS answer the question where you publish. They say nothing about the one that worries me most at my age: will what you published still be there when you are not?

The whole "own your content" movement assumes you will be around to keep owning it — to renew the domain, to pay the host, to answer the support tickets. But at 66, I am starting to think more about the opposite assumption: what happens to all this when I cannot maintain it anymore?

What?🔗

I started applying PDP years before I gave it a name. Five years ago, in one year retired I wrote:

It may sound ridiculous, but at now 61, setting things up so they do not disappear when I do is becoming an important consideration. I will thus progressively stop using self-hosted sites to publish on places that will survive me, as nobody in my family is tech-savvy enough to maintain a web site.

That was the switch from mercurial to git, and moving to GitHub. As we say in France, "the graveyards are full of indispensable men."

So, as a pun on the famous saying Publish or Perish I decided to call it PDP: Publish, Don't Perish!.

How?🔗

Over the years I have refined this into a simple triage: separate my content by how much I need it to survive me.

Transient — thoughts, rants, day-to-day musings. Keep them on this own self-hosted site. If it dies, no great loss.
For the family — photos, memories, things your children will want. Put them somewhere that outlives you and easy to use for non-technical wife and chidren, even if it costs a bit: this is where paying a reliable host is worth it. I use a Hetzner Box.
Perennial — code, hard-won knowledge, things the world might still want in 50 years. This goes into git repositories that I can thus move easily to git platorms that are the most probable to survive me, GitHub for now. I don't use facny GitHub features, to make moving or copying these repository as simple as possible.

And there is now a new, unexpected ally in the perennial column: AI. Even if I never get a single visitor on my pages on this site, I know that publishing on a perennial platform means my code will be ingested by AI assistants and become a tiny — very tiny — part of humanity's collective knowledge. I have no illusion that my contribution is a major one, but there is something quietly satisfying in the thought that my bash snippets will be digested forever by some distant neural network. The graveyards may be full of indispensable men, but their code lives on in the training data.

Am I reinventing the wheel?🔗

I did a bit of homework, and the honest answer is: no, the idea is not new. There is a whole movement around it, I just did not know the name.

The IndieWeb community has called it the Long Web for years, a concept Jeremy Keith articulated in 2008: the qualities that make a format good for the long term — simplicity, openness, standardisation — are the same qualities that make it good practice today. Brennan Brown's recent essay is the best modern treatment: the stats are sobering (a quarter of the web's pages from a decade ago are gone, two-thirds of links rot), and the conclusion is that the IndieWeb is accidentally a preservation ethic.

There is also Jeff Huang's "This Page is Designed to Last", a practical manifesto for building sites that survive 10 to 30 years — vanilla HTML, no hotlinking, static over dynamic. Much of what I did with this very site (plain Markdown files, no database, an SSG that outputs static pages) aligns with it, and the migration to Zola in 2026 was done with exactly these concerns in mind.

And if you want someone else to do it for you, the market exists: Posthaven promised to never shut down, and WordPress.com now sells a 100-Year Plan — $38,000 to keep your site online for a century. I wish them well, but it does raise the obvious question: what happens to the plan if Automattic itself does not make it to 2125?

So where does that leave PDP? I see it as the lazy man's version of all this. The Long Web is a movement, Jeff Huang's page is a manifesto, the 100-Year Plan is a product. PDP is just a reminder, for people like me who do not want to pay $38,000 or join a movement: choose your formats and your hosts with an eye on outliving you, and the rest takes care of itself. The good news is that the "correct" choice is also the simple one: plain files, open formats, and a host bigger than you are.

POSSE and PESOS assume you will be there to keep your site alive. PDP is for those of us who know we will not. :-)

Related thoughts: my single-binary fanaticism is the same reflex applied to code; the 2026 Zola migration and moving to GitHub were PDP before I had the name.

Fanatic of the single self-contained executable

2026-07-31T00:00:00+00:00

I am becoming more and more a fanatic of the single self-contained executable, with no runtime dependencies, that Go provides: one binary, forever. No npm, virtualenv, docker, flatpak, snap... no version hell.

The qualities I lived with and took for granted during my years of coding in C and shell.

To the point that when I find now a python/node/java/ruby/php piece of software I would like to install, my first reflex is to just recode the darn thing in Go. I know, I know, rewriting a perfectly working piece of software just to avoid a dependency is bordering on insanity, but the result is so satisfying... A single file that will still run in 20 years, with no upgrades, no breakage, no maintenance.

And with AI, now, it is becoming less and less insane!

(Bluepost )

Why is the US the only country to write dates as MM/DD/YYYY

2026-06-09T00:00:00+00:00

In short: chance, followed by isolation.

Before the late 19th century, there was no international standard for writing dates. In both Great Britain and the American colonies, people used two formats interchangeably.

You could write Month-Day-Year (e.g., "October 31st, 1776"), which matched the spoken phrase "October 31st."
Or, you could write Day-Month-Year (e.g., "the 31st of October, 1776"), matching the spoken phrase "the 31st of October."

Because the British Empire colonized America during a period when Month-Day-Year happened to be incredibly common, early Americans adopted it as their standard format.

However, during the late 19th and early 20th centuries, global trade, international shipping, and telegraph networks expanded. Great Britain sought to standardize its systems to align with continental Europe, which favored the Day/Month/Year format because it is strictly logical and hierarchical.

Ultimately, while Britain altered its habits to match Europe, the United States remained geographically isolated across the Atlantic and saw no pressing economic reason to change. The exact same thing happened with the adoption of the Imperial system (feet versus meters).

I recovered my Facebook account, read how

2026-04-28T00:00:00+00:00

What happened🔗

On 2024-10-24, my Facebook account was hacked. Alas, Facebook did not allow me to get back my account, or even report the account as compromised, without entering the current password that whad been changed by the hackers. So I had to create a new account, and moved on.

However, I recently created a test account on Instagram, to try to generate RSS feeds to follow Instagram accounts more comfortably via my RSS-Bridge. And I was surprised to see that my "new" account was in fact the account of another user that had been blocked, then unblocked.

This prompted me to try to re-log into my old Facebook account and... it worked! I was able to reset the password without knowing the current one. I thus will remove the new account after some time.

Nice, but why?🔗

But looking a bit (for instance about Risk-Based Authentication (RBA), it appears I was lucky to try to re-use my account just days after a 6 month delay: Facebook relaxes accounts the security measures 6 months after a detected suspicious activity!

The behavior I experienced — being blocked from recovery despite having valid contact info, followed by a sudden "opening" of access — is apparently a documented pattern in Facebook’s security ecosystem, notably their automated risk assessment procedures.

The "Security Freeze" (Why I was stuck)🔗

When a hijacker changes a password, Facebook’s "Trusted Device" and "Location" protocols often trigger a lock. Even if my email and phone were correct, Facebook may refuse a reset if:

I cannot provide the current password, aka The "Current Password" Requirement. This is a common defensive measure when Facebook detects a "conflict of ownership." If the system sees two different locations (mine and the hacker's) trying to claim the account, it often demands the current password to prevent the owner from being "kicked out" by someone who isn't actually the owner. Needless to say, this is a really bad design, as hackers first move is to change the password.
I use another browser. As I tried many ways to recover my account, I also tried to use another browser on another machine, to start afresh. But this failed Facebook IP Reputation detection: Seeing I was trying to recover the account from a network or device the system didn't "trust" at that moment, it applied the most restrictive recovery path to the account.

Why it suddenly worked (The 6 months "Cooldown" Period)🔗

Nothing is actually officially documented, but here is the likely scenario:

Risk Score Decay: After 6 months of inactivity (or if the hacker eventually triggered a "Security Lock" that made the account go dormant), the "Risk Score" associated with the account drops.
Account Dormancy: Facebook often flags accounts that have been compromised as "checkpointed." After a certain period of time without successful logins from the hijacker, the system may lower the threshold for recovery for the original owner, especially if he are using a device or IP address that was associated with the account for years before the hack.
The "Verification Reset": Facebook occasionally clears the "verification debt" on accounts. If the hacker was blocked by Facebook’s automated systems (e.g., for spamming), the account enters a state where the next person to provide valid 2FA or email confirmation is granted access without the "current password" hurdle.

All of this is totally automated, with no human intervention, by the system's "Identity Verification" logic resetting itself after a period of dormancy. Once the "Pirate" was no longer active, the system stopped viewing my recovery attempt as a "hostile takeover" and allowed the standard email/phone reset to work.

Summary of Known Behavior🔗

Stage	System Logic	Result
Initial Hack	High-conflict state; system protects the "active" user.	Recovery denied despite valid email.
The 6-Month Gap	Account goes dormant or hacker is "checkpointed."	Security tension on the account relaxes.
Recovery	System recognizes you are the owner IP/Location and valid contact info.	Password reset allowed without "Current Password."

What to do once the account is recovered🔗

Check the Settings for "Saved Login", "Where you're logged in", and "Apps and websites" immediately. Sometimes hackers leave a "backdoor" by authorizing a specific browser or app that doesn't require a password.
Change the password, enable 2FA (Two-factor authentication).
Remove all apps connected to the account.
Check other sites that authorized logins from this Facebook account, change their passwords andf look for suspicious activity.

Numéros de telephone arnaques

2026-04-13T00:00:00+00:00

Attention arnaque:

Le No de tel 0189628604 vous appelle en se faisant passer pour un conseiller bancaire (AXA), en ayant votre nom, adresse, en prétendant avoir remarqué des tentatives de paiement répétées depuis Amazon.

C'est une arnaque. J'ai raccroché de suite et appelé AXA pour vérifier.

La conseillère nous a d'ailleurs dit que elle-même avait failli se faire avoir récemment (un faux appel de sa mutuelle).

Je fais ce post simplement pour mettre 0189628604 (ou 33 189628604 ou 01 89 62 86 04) sur le web pour que une recherche de ce No tombe sur ce message.

A new version of the site for 2026, in Zola

2026-03-19T00:00:00+00:00

I said that I wanted to migrate this web site to a SSG (Static Site Generator). And, after 5 years of procrastination, Ta-Da! here is the new site!

Initially, I was planning to use Hugo, since I love coding in Go, and was not planning to learn Rust (Hugo is coded in Go and Zola in Rust), but I finally chose Zola over Hugo for many reasons:

Zola has basically all the performance advantages of Hugo, and is also a single binary, with no need for databases, docker, snap, flatpak...
Zola is opinionated. This is very important for me, I love more and more the systems were there is "only one way to do it". Needless to say, I hate perl, and I love Go.
Zola is stable, and its creator ("Keats") very committed to not chasing the latest fad. This means less risk of an upgrade breaking things.
Zola template engine, Tera, is both more powerful and more readable than the standard Go template engine used by Hugo.
Zola comes with more useful built-in features, but no plugins that in my opinion brings complexity, especially in docs, and upgrade hell.

As for the theme, I was looking for a clean responsive modern simple theme, with high scores in the various web performance benchmarks, with search and comments, and some support for multilingual (the site is in English with a few pages also translated in French). I tried the two leaders in the Zola theme ecosystem:

Tabi I liked the looks but it felt a bit too complex for my needs: as a Zola novice I was afraid to end up relying too much on Tabi-specific feature instead of pure Zola and CSS ones.
ABridge I liked the architecture, not so much the design, but I struggled to make it work, it seems the doc still advise settings that break on recent Zola versions.

I thus looked for a simpler theme, and chose:

Apollo which was perfect for my needs, and simple enough to be easily modified for any of my future needs. I also enabled its support for the Giscus commenting system.

Porting my contents was made easy because I had anticipated this migration, and used Markdown for my posts in the Foswiki engine, instead of the native Foswiki syntax. So converting was basically adding a frontmatter to the pages, and fixing some links. I converted older posts with pandoc.

I have made a quick first migration. I will continue working on the CSS architecture and the looks in the following weeks.

Update on 2026-03-22: I now am satisfied of the result. I plan to work a bit more to add a search for only the French pages (now the search ignores them), autodetection of the browser light/dark preferences, sort the tags alphabetically in the page info line.

Mon compte Facebook a changé (piraté)

2024-10-24T00:00:00+00:00

OBSOLETE

Warning: obsoleted by this new version.

Attention : le 2024-10-24, mon compte Facebook a été piraté. Hélas, Facebook ne m'a pas permis de récupérer mon compte, ni même de le signaler, et j'ai donc dû en créer un nouveau.

Je vous prie donc de ne plus utiliser mon ancien compte ( https://www.facebook.com/colas.nahaboo ) et de le bloquer sur Facebook pour être sûr, car les pirates vont l'utiliser pour pirater et hameçonner d'autres personnes, et d'utiliser mon nouveau compte: https://www.facebook.com/profile.php?id=61567836084230.

J'ai changé ma photo de profil (sur Facebook uniquement pour l'instant) pour différencier les deux comptes :

My Facebook account has changed (hacked)

2024-10-24T00:00:00+00:00

OBSOLETE

Warning: obsoleted by this new version.

Warning: on 2024-10-24, my Facebook account was hacked. Alas, Facebook did not allow me to get back my account, or even report it, so I had to create a new one.

So please never use my old account anymore ( https://www.facebook.com/colas.nahaboo ) and block it under Facebook to be safe, as hackers are going to use it to hack and phish others, and use my new account: https://www.facebook.com/profile.php?id=61567836084230

I changed my profile picture (on Facebook only for now) to differentiate the two accounts:

One year retired

2022-02-24T00:00:00+00:00

I have been retired for one year now, and what happened? Well...

I basically stopped coding and reading about anything computer-related for eight months. I didn't expect it, I was thinking I would start coding at once, going through my huge todo list of various projects.
I enjoyed being able to go SUP surfing at will, but I ended up actually less on the water than before, simply because i didn't feel pressured into going on the water if the conditions were not enjoyable: I have now all the time in the world, I can wait for better conditions.
I started what I should have done years ago: Stretching at least 30 minutes daily, and warming up a full 10 minutes before going surfing. Work and play for me were mainly sitting in front of a computer, and I had become awfully stiff. I will detail my routine in later posts, it is worth it.
I dived deep in modern physics: astrophysics, cosmology and quantum mechanics. This also will warrant a full separate post.
I started re-coding in the last four months:
- The first month I began working on various bits of code (shell scripts), progressively getting back to speed, and also deciding to learn the language Go. I was looking for a very longtime for an efficient language to replace lisp or C as compliment to scripting (I never liked Java), and was hesitating between Rust and Go. A discussion with Olivier Arsac convinced me to try Go.
- I switched to git (I was using mercurial) for version control, and moved my publicly available source codes from my personal web site to my GitHub. It may sound ridiculous, but at now 61, setting things up so they do not disappear when I do is becoming an important consideration. I will thus progressively stop using self-hosted sites to publish on places that will survive me, as nobody in my family is tech-savvy enough to maintain a web site.
- Then in December I stumbled upon the Advent of Code challenge that I decided to do in bash for the challenge, and it was unexpectedly productive? I learned more in these 25 puzzles than in my last 10 years of professional bash scripting. You can see my solutions in my post Bash lessons learned with AoC 2021.
- Thus, after finishing the Go tutorial, I decided to do the previous AoC years in Go. And I just finished the AoC 2015 in Go. This definitively made me a Go enthusiast, I realized that the designers of Go had exactly the same opinion of what should be a general programming language: They worshiped readability, simplicity, performance and maintainability while hating inheritance and the feature creep of most modern languages.

That was my first year. And now, embarking for the next one!

Pinboard pricing model

2021-03-01T00:00:00+00:00

I am a knowledge addict, and always has been. I was reading as many books as possible, and the advent of the web has been a marvelous gift for me. But I soon realized that I needed some tools to manage this knowledge and supplement my memory. It started with pen and papers, then I used a lot the PDAs (Personal Digital Assistants) such as the fabulous Psion Series 3 and Series 5, the various smartphones, and finally the bookmarks-in-the-cloud service Delicious

But as Delicious closed, I searched a replacement, and found Pinboard, and was hooked. Not only it offered the same functionalities as Delicious (and more), but if was fast and sleek, with a minimal no-nonsense design that I favor.

It also had an original pricing system: you paid a one-time fee of $0.001 per the number of current users to get access for life. E.g $9.26 if there was already 9263 paying users and you wanted to become the 9264th. The goal was to entice people to register now rather than waiting, even though it may not have scaled. (see Pinboard.in Pricing Model: Fixing Their Math).

Alas, it could not be sustained, so in 2015 it adopted a more mainstream yearly fee scheme: $11 / year. New Pricing Policy (Pinboard Blog)

But we, old timers, kept our lifetime access we bought before the change. We were basically free-riding the new wave of customers...

Of course it could not go on, so last week, Maciej, the Pinboard author, suggested we voluntarily switch to the yearly fee scheme. And I did it gladly, Pinboard is for me a model of what web sites should be: sleek, fast, and efficient (and not re-selling your private data). At the opposite of the bloated monstrosities that a lot of sites tend to be nowadays.

Long live Pinboard!

Retired, at last!

2021-02-25T00:00:00+00:00

At last, I am now retired!

I ceased all activity on 2021-02-17, and my work account will be destroyed on 2021-02-28. A new era of full time surfing, coding and playing opens up. For now it is mostly playing with the new Valheim game that is making a huge debut(*), as the waves are not good (winter storms) and I take a (small) coding break.

I will try to make a post per week on this site, starting now 2021-03-01. I know there is no way I will be able to keep the rhythm, but let's say it should be my long-term goal.

(*) I strongly recommend you read # Valheim: 7 marketing lessons from their amazing launch, which tries to shed some light on its phenomenal success.

A New version of the site for 2021

2020-12-27T00:00:00+00:00

A big change is planned for this site in 2021: I am going to abandon the Foswiki engine for it, to pursue my recent goal of going back to the web roots. Basically, it a quest that relies more on:

Fast rendering by no rendering at all: pre-compiling the site pages into static HTML+CSS+JS pages: the JAMstack approach.
What I call a LAUW approach (Linux Apache Unix-utilities Web Standards) instead of the traditional LAMP. Using the full strengths of Linux, Apache, the GNU Unix utilities, and the modern web standards (HTML5, CSS4, JS ES6) now that Internet Explorer is at last dead. This means not using software libraries (e.g: JQuery) and processors (e.g: Saas) that were designed to remedy to the poor state of some browsers.
Using Markdown for text editing
Using ultra stable and reliable technologies. I want to use things that will still work in 10 years, not the fad-of-the-day javascript framework that will be obsolete in 6 months, or that will break compatibility on each update.
Technical: And that do not try to be friendly to non-technical authors and admins, as I will be the sole admin and editor.

Now🔗

So now (December 2020), I am still using Foswiki, but progressively prototyping on it the features I will try to implement later:

A simplified skin, with all the "wiki-ism" (everything editable) removed. As you can see I got rid of the left toolbar, to keep only the top banner with its sections (Code/Surf/Blog) menu on the logo and minimal tools on the right
Editing contents in Markdown instead of the Foswiki syntax: I now do not use the Foswiki system to edit topics, but desktop-edit via marktext.
Having all the post sharing a common Form, so that the fields of this form can be easily converted to YAML as used as frontmatter (metadata) in modern Markdown-based systems.
But still using all the Foswiki power for the site admin and non-content topics (indexes,...)

The idea is to migrate progressively the contents into a form (Markdown + Frontmatter) usable in a modern static web generator like Hugo, while adapting Foswiki to use this content form, preparing a future migration.

I have developed the details of what this means in my post on the Foswiki site: Foswiki for non-wiki web sites

Spring 2021🔗

After I retire, at the end of February 2021, I will have more time to ditch Foswiki and implement the site on a new engine. It will probably be a static website generator, most probably Hugo, but maybe also Zola Eleventy, or Gasby, and incorporating ideas of the myriad of other ones.

Summer 2021🔗

I will then probably replace progressively various components of this system by ones that I will handcraft optimally to incorporate all the ideas I have been toying with during all these years: the HTML+CSS+JS layout, the tag/category system, the search engine, the comment engine, the editing process...

Later...🔗

And I may even at some time create a totally new web engine system, who knows?

Or code a wiki replacement: a collaborative editing site based on the concepts learned via this experiment.

A New Version of the site for 2020

2020-07-26T00:00:00+00:00

After 10 years of neglect, it is time for a new start of this site, because:

I am going to retire in early 2021 (in 6 months), and I will have now time to properly maintain a personal site
I am feeling the need to archive in a central place all the contents I scatter in various different places, if only for the selfish reason to be able to retreive them more easily
I was toying with the idea of going to a static web site, as it is now in vogue, and it has a kind of nostalgia for me as I started my web sites as static sites anyways. But by looking at the existing solutions, I was not convinced by them, as a lot seemed to naively solve the 80% easy part of the problem, and them struggle without any clear vision on how to solve the 20% hard part... a classic situation.

I thus decided to keep on using Foswiki (v2.1.6), as it is now quite mature, and thus stable: I like building on solid, time-tested fundations that will not change every month. And let's be honest, it is both extremely powerful and incredibly reliable. I am just tweaking it for using it as a public site (one author with all admin rights, and read-only viewers), rather than the standard wiki setup of many authors, all able to edit contents, with some able to also admin rights. My postulate is that a dynamic web site like a wiki, with proper caching, has all the advantages of a static web site, but is more practical to maintain (although more complex to implement).

I will detail my views on how to website-ize a wiki in future posts, but for now, on 2020-07-26 at noon French time, let go live!

Welcome to my (new) world!

Colas.

Migration to Foswiki

2015-10-18T00:00:00+00:00

At last, I found the motivation to work again on this site that I have left dormant since 2009. And I start with a long deserved migration of the engine from the old TWiki system to the superb Foswiki WIki engine.

Just installing a new Foswiki, and copying the data seem to provide a working base, I will then work on fixing the minor bugs and designing a new look in the following weeks.

Usability issues on a payment form

2009-03-03T00:00:00+00:00

I know, I know, it is easy to nitpick on other people design, but this morning I tried to pay my phone bill online, and... failed. I had to retry many times, because I kept making mistakes on a simple, run-of-the-mill form to enter my credit card data. I was dumbfounded. How can, in 2009, people manage to make such a trivial and essential part of online business go so wrong? As I guess you are as curious as me, here is what I found out: (See the picture for the 3 main pain points)

1 the expiration date example is ambiguous: does 0903 here means 2009-03 or 2003-09 ? Why didnt they disambiguate it by also adding a specification (like MMYY, or, as it is a french dialog, MMAA ?)
2 but today, it was not only ambiguous but plainly misleading! As we were in March 2009, I assumed that the example was built from the current month and year. I said cool, so it is YYMM. Wrong! it was actually the reverse, MMYY, resulting in one failed attempt
3 ok, I filled the fields, and hit the OK button... and it just reset the page to empty fields! Grumbling, I re-typed the fields, re-clic, and... another blank page! After these two more failures, I tried to think (which is what you do not want your users to do, trust me). I then read the button label and discovered that this big, prominent button alone at the bottom of the page was not the Submit button but the Cancel one. I should have hit the middle one "Valider", but this button was not in the correct place on the flow of the form.

So,after 3 failed attempts, I managed to pay my bill. All this on a simple form with no fancy verification code or Captcha. Well done Telefact, I wonder if you can find a worse example still in use today.

What is good code?

2008-10-22T00:00:00+00:00

I stumbled upon the blog post "Is it important to write good code?" the other day, and became more and more ill at ease as I realized that I thought that I preferred the original code, that the author was trying to ridicule, over his new "improved" object-oriented version. At first I guessed this was another manifestation of the the "Worst is best" scenario - enhancements are often not worth the added complexity - but I realized that it was perhaps a more profound factor:

The original code is very good because it ... is small. It fits on a teminal screen, so an human being can read it at once and have less items to maintain in his short term memory and also understand it easily because it follows a natural way of thinking with sentences using IF. This becomes obvious by reading the body of the blog post surrounding the code samples, where you can see that the author is using phrases such as "if I need this I do that", showing that in plain english, the if statement is the best way to make people understand what you mean. And making code that people understand is the best way to make debuggable and maintainable code.

At this moment I noticed the citation in the blog header: "Good programmers write code that humans can understand"

Indeed :-)

PS: I know I am a bit exaggerating the issues there, and that I unfairly nitpick on Fredrik Normé, but it is that it seems to me from my personal experience that the two changes I see most in my coding efficiency as I grow older is a decrease of my short term memory capacity, and that I make more and more typos where I realize I mix up totally words with totally different meanings but that sound the same, for instance writing "never" instead of "nether", making me suspect that our natural way of thinking may be much more language-based that I imagined...

Mercurial web templating

2008-10-09T00:00:00+00:00

For some time now, I have seen the light and I switched to the new wave of the Distributed Source Control management systems. Linus famous video of his Google talk decided me to try. I was a bit apprehensive at first, wary of engaging myself on a technology that would bring more problems than solutions, but after some days of use, the realisation dawned over me: Distributed Source Control may be one little step for a programmer, but it is a giant step for programming. Why? because, its mental model actually follows your first intuitions that most of us developed as young programmers before using any source control system. Suppose you want to try a feature? instead of just copying the directory, you clone it. You mess an operation? you just remove the directory, no embarrassing traces left to keep the burning scar of shame on you for the following decades of a central repository. You are used to think of your enhancements as patches, you can work with patches. I could go on and on, but many people have done it much better than me so I'll just say that you should definitely try it.

By the way, between the 3 main contenders, git, mercurial (aka HG), and bazaar, I chose mercurial because of its simplicity, its better support of windows (I work in Linux since 1995 but my coworkers currently use Windows), and because some people at ILOG started using git so I wanted to be able to try something else. I did not choose bazaar as I wanted to stay close enough of git, to be able to switch to it if ever my dream to work in a windows-free world materializes one day...

So I started to set up a public web repository of my Open Source personal work with mercurial, which gave me a simple way to publish my work in full detail, but, although mercurial is quite easy to use, but I had a bit of trouble figuring out how to customize its look & feel.

The rebirth of this site

2008-04-04T00:00:00+00:00

At last, I decided myself to redo my old website, which was a collection of static html and was based on many of the various tools I made in these past years. As I grew in love with wikis over these years, the natural tought was just to do it in a Wiki, and naturally in the Wiki I like the most, TWiki. So bear with me and keep posted, as this site will slowly emerge in this first half of 2008.

Browser-history

2008-03-21T00:00:00+00:00

Legacy

Browser-history started on February 15, 1996, out of a desire to overcome a Netscape bug: at the time, there was no global history. [Nowadays, in 2026, all browsers have one, of course.] Back then, if you closed a window, its entire history was lost.

For people visiting many sites, the ability to track where they had been meant they didn't have to clutter their bookmarks file. If you weren't sure a site was worth saving, you didn't have to bookmark it; if you needed it later, you could just browse your history files.

Later, it occurred to me that this could also be a valuable add-on for people developing experimental browsers [a lot at the time], sparing them from having to build the functionality themselves.

Browser-history was a small, efficient daemon written in C. Real user services could be built on top of its log files to provide more advanced features, such as graphical representations, advanced search, or collective histories. While it could be seen as a "quick-and-dirty hack" compared to the general solution of using a personal proxy for housekeeping, it was easy to use and it worked.

I created this legacy page to help people find the archived version of the project, specifically the latest version of 2000-07-20.

The code archive itself is browser-history-2.8.tgz

Here is a screenshot of the page:

Enjoy!

The manual page🔗

NAME🔗

browser-history - external history of web browsers, such as netscape

SYNOPSIS🔗

browser-history [ options ]

DESCRIPTION🔗

Browser-History is a client-side X daemon maintaining a browser-independent global history of all the web sites you visited .Its usage is very simple, just put the following line in your .xinitrc or your X start-up script:

browser-history &

And open the file ~/.browser-history/history-log.html and bookmark it.

Browser-history came from the will to overcome a Netscape bug: there was no global history, and if you close a window, its whole history is lost. For people browsing lots of sites, having a possibility to track back where one has been before means that you dont have to put everything in your bookmarks file. If you are not sure if a site may be worth remebering, don't add it in your bookmarks. If you need it later, just browse your history files.

Later, it came to our minds that this also could be a valuable add-on to people writing experimental browsers, so they dont have to add this functionality to their browser itself.

Browser-history is a small and efficient daemon. Real user services could be built on top of the log files it maintains for more possibilities (graphical representation, advanced search options, collective histories). It can be seen as a quick-and-dirty hack wrt to the general solution of using a personal proxy to provide this history and housekeeping facilities. But in the meantime, it is easy to use and it works.

IMPLEMENTATION🔗

Browser-History spies your web browser and logs in ~/.browser-history/history-log.html all the URLs you went through. You can then browse the log under Netscape or other browsers via the URL: file:~/.browser-history/history-log.html (replace the ~ by your home directory). It tracks automagically all already present browser windows, and all new ones created in the future. This program has no user interface. It just appens information to a log file in html format so you can browse it through a web browser. If more that one hour has passed since last entry, it draws an horizontal lines, and adds H1 headers to delimit new days. Each week (sunday mornings), it archives the week history, compresses it by gzip (that you must have in your path), and starts a new history with links to the older ones. To make room you can just remove the obsolete history files. You can search files for string XXX in shell via:

zgrep XXX ~/.browser-history/*

This version (2.6) works with Netscape, Arena and Amaya.

URLs can be excluded from logging by putting them, one per line in the file ~/.browser-history/history-log.exclude , then, if an URL begins with a line from this file, it is not logged. In this file, empty lines or lines beginning by # are comments This file is read once at startup, and re-read when receiving the signal 1. e.g:

# We exclude local files file: # Exclude search engines... http://home.netscape.com http://guide.infoseek.com

When browser-history is run, it looks if another one is running, and by default it kills the previous one if it is an older version. Otherwise, it the new one is the same version number or older, it just aborts.

OPTIONS🔗

All options can be given by their first letter: you can specify either -verbose or -v, but you cannot group options, e.g. you must say -v -k , but not -vk

-display display_name
Specifies X display, otherwise contents of $DISPLAY is used

-verbose
outputs information on what it is doing. useful for debug.

-Version
prints version number and exit.

-logdir directory
which directory to store files into? defaults to ~/.browser-history

-gzip gzip_filename
the complete path to the gzip compressor. Defaults to "gzip". E.g:

-gzip /usr/gnu/bin/gzip

-seconds delay
if two entries are made are more than delay seconds apart, an horizontal rule will separate them, else just a simple line break. Defaults to one hour (3600).

-replace
If there is an already running browser-history on the display, aborts. Default is to replace it only if the version is older than ours.

-noreplace
If there is an already running browser-history on the display, aborts. Default is to replace it only if the version is older than ours.

-kill
If there is an already running browser-history on the display, kills it, then terminates immediately in all cases..TP -DontGrab Never Grab the X Server, which might cause deadlocks while debugging, when browser-history or gdb tries to print on the grabbed xterm or emacs.

LOG FILE FORMAT🔗

A log file can have some decorative HTML to represent days, but each entry has the form: (on a single line since version 2.4)

separator
    <b>name</b>
    <a href="URL">URL</a>
    YYYY/MM/DD-HH:MN:SS <small>windowid</small>

where the following items are:

separator: either
or
name: the name of the document (window title) URL: its URL YYYY/MM/DD-HH:MN:SS: year, month number, day number, hour, minutes, seconds. (2-digit numbers) windowid: the X window ID of the browser window, in hexadecimal

Note Before version 2.4, the 4 sub-parts were separated by newlines, but since 2.4, they are only blank-separated to ease searching for URL in log files via "grep". Note Before version 2.5, the year was stored in 2 digits. Now it is stored in 4 (or more :-) digits, to fix this Y2K bug

HOME🔗

The latest version of browser-history can be found at:

ftp://koala.inria.fr/pub/browser-history

and its WWW home page, with full technical documentation is at:

http://koala.ilog.fr/colas/browser-history

AUTHOR🔗

Colas Nahaboo, http://koala.ilog.fr/colas

COPYRIGHT🔗

browser-history bears the same license as the X Window System: you can do everything with this code (selling it, modifying it), except suing me or using my name in your advertisements, or expecting any kind of support or guarantee.

Colas Nahaboo X mouse wheel scroll page

1997-12-30T00:00:00+00:00

Legacy

On December 30, 1997, I started this page to help developers manage the — then new — mouse wheel. My goal was to encourage a standard scrolling direction for the X Window System on Unix and Linux.

It quickly became the most-viewed page on my personal site, and even across all INRIA-hosted websites at the time. However, as mouse wheels became mainstream, the page lost its relevance. I let it quietly disappear in 2018 when I upgraded my site.

Seeing that this remains one of my most accessed URLs even in 2026, I decided to create this legacy page at the same legacy URL. It allows people to easily find the historical version on the Web Archive, specifically the latest version of 2015-09-06, from which is the screenshot:

Enjoy!