Bref, nous avons besoin de vous pour pouvoir assurer à la fois l’organisation effective du calendrier et son équilibre éditorial. Quel que soit votre profil, n’attendez plus pour nous soumettre une proposition d’article.
framaforms.org/appel-a-suje...
The `unload` event no longer fires in Chrome by default!
Since the advent of the mobile web this has become increasingly unreliable and we think that's a footgun for devs who might think it works.
We've spent the last 3 years carefully removing it from Chrome.
developer.chrome.com/blog/unload-...
The pitch for single page apps was always “it’s faster after the first load”.
Google’s Core Web Vitals, used in search ranking, now measure soft navigations.
It’s true: LCP drops by about a third.
But, I don't think SPAs are ever really worth all the compromises and extra code.
Remember Internet Explorer’s “Page Transitions”?
Well, they’re back, thanks to the View Transition API!
→ Blog post: brm.us/page-transit...
→ Demo site: page-transitions.style
The project leverages IE’s meta tags so browser support is not limited to Modern Browsers™ but also includes IE 5.5–8.0 😆
⚠️ PSA for React 19 SSR apps: React automatically preloads ALL server rendered images. Without hints, early bandwidth is needlessly wasted.
Fix it:
* LCP image → fetchPriority="high"
* Never lazy load the LCP image
* Below-the-fold images → loading="lazy" (this also stops React preloading them)
How far ahead browsers fetch lazy images:
• Chrome: 1,250px on 4G, 2,500px on 3G, 6,000px (!!) on 2G
• Firefox: 600px, on any connection
• Safari: one viewport. Lazy iframes: 0px, they wait until on screen
This is a great article. I've always been sus about the hype around JPEG XL, as it never stacked up when I actually tested it with web use-cases.
giannirosato.com/blog/post/ca...
Empreinte eau des data centers: de quoi parle t'on ? quand ? comment ? combien ? 💧
Cet article est là pour vous aider à vous repèrer, dans le but de permettre un débat qui s'appuie sur des points de repère partagés.
👉 hubblo.org/fr/articles/...
Atelier #ParisWeb : le 24 septembre, @jpvincent.bsky.social vous montrera que même à l’heure de l’IA, il est encore utile de maîtriser tes outils les outils de performance classiques.
www.paris-web.fr/2026/atelier...#performance#outils#méthodologie
Pour rappel, une analyse à l'échelle mondiale montre que chaque kg de CO2 émis (principalement par l'utilisation d'énergies fossiles, pétrole, charbon et gaz) entraîne la fonte d'environ 15 kg de glace de glaciers
www.nature.com/articles/s41...
- Fin
⏰ Il ne vous reste plus que cinq jours pour profiter du tarif réduit sur les conférences et ateliers.
Si vous n’avez pas encore sauté le pas, il est encore temps !
inscriptions.paris-web.fr
Jean-Pierre Vincent - #webperf & #greenit 🚀♻️🌱 reposted
Il ne vous reste littéralement plus qu’une seule semaine pour profiter du tarif réduit sur les deux journées de conférences et d’ateliers de #ParisWeb 2026. Une semaine !
www.paris-web.fr/actualites/2...
Inscrivez-vous à moindre coût tant qu’il est encore temps.
In my audits, the question I hear most after a deploy is "why hasn't CrUX moved yet?"
The window is always 28 days; daily, weekly and monthly are just cadences over the same data. And p75 resists change by design
I explain it with two interactive demos
joanleon.dev/en/crux-28-d...#WebPerf#CrUX
We've just launched a new free tool: the Website Bottleneck Finder detects which resource requests hold back the FCP, LCP, and H1 rendering time!
www.debugbear.com/website-bott...
If you make websites for a living, stop what you're doing and read this research by @tammyeverts.com; it shows what many of us have been saying for a long time: even if there is such a thing as "fast enough", it's generally *much* faster than you are targeting:
embrace.io/blog/research-co…
Did you know that H3 has a built-in `writeEarlyHints` utility that sends early hints?
(Thanks to the initial contribution from @danielroe.dev, we’ve had it since the early days!)
There's an HTTP status code for performance.
103 Early Hints sends your preload and preconnect headers before the real response, so the browser downloads your CSS and hero image while your backend queries the database.
The first part of Declarative Partial Updates (`<template for>` out-of-order HTML) is now available, unflagged, in Chrome 150!!
developer.chrome.com/release-note...
The second part (the new JS setting APIs, include the streaming ones) we're still working on the spec side, but maybe for Chrome 152?
What you're seeing today in computing price hikes means longer service lives and stagnant performance, which is what the *majority* of the market for computers that load websites have felt for a decade. The privilege bubble popping is only news to bubble denizens:
infrequently.org/2025/11/perf...
If your content is served over HTTP/3, make sure your DNS record advertises it, otherwise you're missing out on free performance.
Check your site here: savearoundtrip.com
💡 L’Arcep publie son nouveau rapport “Intelligence artificielle générative : quels défis environnementaux ?”
Il souligne le manque de transparence de l'industrie de l'IA et propose des leviers d'action.
Découvrez-le ici : www.arcep.fr/actualites/a...
De l'importance de savoir comment chaque métrique est calculée : ici, aucun changement du point de vue utilisateur, mais un TBT qui explose (x10). Le TTI a été repoussé, parcequ'un preload a été ajouté.
(c'est pour ça qu'en formation je passe bien 3-4H juste sur les 10 métriques de la perf)
Un bel article écrit par Arthur Russias pour AEF Info, basé sur les données du projet #DCWatchwww.aefinfo.fr/depeche/7484...#datacenters#opendata#osint
One of the most important pieces about web performance that I've ever worked on is finally live. After a few months of work, my analysis of the correlation between site speed and conversion across the entire @Shopify platform is out: www.shopify.com/enterprise/b...
A Gaussian blur in Canvas 2D takes 280ms. The same filter with WebGPU: 8ms.
I wrote the second article in the WebGPU series: image processing in the browser with WGSL shaders and real benchmarks.
joanleon.dev/en/webgpu-im...#WebPerf#WebGPU#JavaScript#Performance
For years, the performance ceiling in the browser has been clear: the CPU.
WebGPU changes that. 🧵
It’s not a library or a framework. It’s a browser API that gives direct access to the device’s GPU, enabling a level of parallelism the CPU cannot match.
#WebPerf#WebGPU
Nouvel outil de visualisation (si vous avez déjà des traces) : waterfall-tools.com, par le créateur de Webpagetest lui-même.
Ça lit tout : des HAR produits par chrome et firefox aux JSON en sortie de webpagetest en passant par tcpdump.
Post du créateur : blog.patrickmeenan.com/2026/04/12/i...
There's something quite exciting happening with AVIF and progressive rendering.
Some recent patches let you provide custom 'frames' as progressive passes, so you can provide e.g. smaller blurry versions of the full image to use as a progressive pass.
Source maps are a vital part of modern web development, but the lack of a standard made it impossible for us to add new features or improve the debugging experience.
This is the story of how we standardized source maps and how we continue to ship features through standards 📜
🧵
Suite des travaux de l'analyse environnementale (Analyse de Cycle de Vie) des #GPU pour l'#IA auxquels #hubblo participe, le rapport complet de l'étude via AltImpact est sorti : librairie.ademe.fr/3426-economi... 1/3
Good blog post from Carlos (a colleague) about a new proposed addition to Compression Dictionaries allowing short-lived HTML pages to continue to be used as compression dictionaries for longer:
dev.to/carlosmateom...#webperf
Pourquoi et comment réduire l'INP ? Use case sur l'équivalent italien de Le Bon Coin :
dev.to/subito/how-w...
* -100ms d'INP = +5% d'engagement
* méthode empirique à base de tests A/B pour trouver les coupables
* GTM bien sûr, mais aussi tiktok et facebook
* React compiler fait ses preuves en prod
How do you measure when a component is “ready”? ⏱️
LCP is too global; Element Timing is too specific.
🛠️ I’ve been working with Bloomberg on a new #webperf API: Container Timing. Think of it as “LCP for a web component”, providing aggregated paint timing for a subtree.
Un SVG complexe alourdit il la page ? Comparaison avec le PNG
En résumé :
- jusqu'à 2 Mo de code, le SVG est plus léger que sa conversion en PNG.
- l'INP avec un SVG dans la page est équivalent au PNG, jusqu'à 400 Ko de code SVG.
From Chrome 145 (on general release next week!), DevTools we will start to show so called "soft" navigations and "Soft LCP" in the Performance Panel traces.
These are for SPAs which don't do a full page load, but instead "fake it" by updating the current page and pushing a new history entry.
1/5 🧵
Un an de recherche
352 datacenters en activité, 65 projets identifiés
des centaines de vues satellite analysées
8,16 TWh de consommation en 2024, hors salles IT dans les batiments tertiaires
(also available in english)
Étude sur ce que voient les crawlers des chatbots IA, notamment des infos sur le temps max d'attente et l'interprétation de JS, mais aussi des frames ou du shadow DOM
think.resoneo.com/sentinel/geo...
Morale habituelle : être performant, HTML est la base + des surprises
Web Performance Calendar'25 day 1 with @mb21.bsky.social on HTTP streaming in the world of HTTP2/3 and Node. Mauro shows us how to start rendering earlier, turning a slow (full) page load into a progressive experience using db.stream() in JavaScript
calendar.perfplanet.com/2025/improve...
#JavaScript: When it comes to breaking up long tasks, the scheduler API seems quite convenient:
await scheduler.yield();
Alas, not supported by Safari yet: caniuse.com/mdn-api_sche...
Thankfully, providing your own fallback implementation is easy: web.dev/articles/opt...
I co-wrote a post on the Etsy engineering blog about implementing Speculation Rules API (SRA) to prefetch product listing pages from search results. And many thanks to @tunetheweb.com and @cdaveross.bsky.social for the help and support!
www.etsy.com/codeascraft/...