My latest article looks at the performance best practices for all three parties in the Revive Adserver ecosystem: the ad server, the publisher and the advertiser. #webperf

revive-adserver.com/how-to/display…

Excited (and a little nervous!) to be giving my first live talk at Google I/O (gulp!) in three weeks time on Instant Loading.

io.google/2024/explore/2…

Looking forward to meeting up with so many people and having some great chats!

Although I mainly focus on my P98, it was a pleasant surprise to see my P75 drop below the magical 500ms mark.

#webperf

📜 WLS a besoin de votre expertise !🌟 Faites rayonner votre passion pour la webperf.
🚀 Proposez vos idées :
- Fondamentaux à explorer
- Détails techniques à approfondir
- Observations transversales à discuter

📅 Avant le 26 mai à 23h59 (heure de Paris).
docs.google.com/forms/d/e/1FAI…

Over 8,000 WordPress sites are using this plugin now!

wordpress.org/plugins/specul…

If you've a WordPress site, that you wanna make faster, then you should check it out.

Networking decompression (gzip, brotli) have been moved off-main-thread as part of ongoing efforts to reduce main thread contention. This work has delivered huge performance wins on our high level page load metrics, reducing FCP and LCP by 10%. See bugzil.la/1356686

Luckily I spotted Chrome/124 started rolling out before I deployed some performance improvements.

Never deploy the same day as Chrome to keep your RUM analysis clean and honest. #webperf

ActivationTime tells you how much lead time the browser has to prerender a page. (higher=better)
Here you can see improvements to ActivationTime after I switched from eagerness conservative to moderate:
🟪P75: 150 » 800ms
🟦P50: 110 » 300ms

#webperf #speculationRules

Now testing speculation rules with eagerness=moderate instead of conservative. This is possible after making some changes to FOSS Revive Adserver. Once tested I will make a PR #webperf

This partial rendering in browsers keeps driving me nuts. In the screenshot you see the last <option> is cut off because of buffering. Why even attempt to render and cause a layout shift. #webperf

Some extreme tuning to remove extra bytes have reduced the problem on most of my pages. The initial rendering now has enough information to avoid CLS of due to the search form being incomplete. #webperf

After starting an origin trial I saw an unexpected increase in CLS. Why?
1⃣ The OT header adds 280 extra bytes
2⃣ My search bar is now split in 2 batches (buffer)
3⃣ Incomplete <form> is rendered
4⃣ Rest of <form> arrives and triggers a Layout Shift

#webperf #buffering

Happy Friday everyone!

The RUM Archive dataset has been updated through Mar 31, 2024:

* Page Loads table
* Third-Party Resources table
* Insights

Visit us at: rumarchive.com

116!

Plus a smaller one of 103 in front of it, but I can't fit all that in one screenshot 🥵

Launching a new site...

We're making the performance ecosystem better. It's a must read for anyone looking for services to improve their site speed.

performance.shopify.com/blogs/blog/don…

RUM noise and compression dictionaries were some of the best learnings we got at performance.now() conference. Thanks to Tim Vereecke and Patrick Meenan for their amazing talks on this

Read more about it from Matt on our blog tech.loveholidays.com/when-stroopwaf…

Earlier this month Chrome shipped support for the zStandard compression. I wrote a blog post about determining which compression is best for your content, and what levels to use. paulcalvano.com/2024-03-19-cho…

Tonight is Web Performance NY's 'Meet for Speed' session. We'll be taking requests from attendees for sites to do a performance analysis on!

Join us virtually at 6:30pm EDT - meetup.com/web-performanc…

Also, feel free to submit sites to analyze here - bit.ly/meet4speed-sub…

We've also added a Browser Versions Over Time chart, shows some really interesting trends!

rumarchive.com/insights/