Servo 2025 Stats
82 points by brendan
82 points by brendan
I've been following servo for years and it warms my heart to see it getting very noticeably better every time I download it and try it out. Of all the new web browser engine projects, I feel the most hope for Servo over the long term.
Very glad to see the alternative browser space heating up in 2025 and also really pleased to see how good Servo is coming along since Igalia started spearheading development!
as an outsider, what does it mean to have a pass rate of 93.4% for WPT?
Isn't that good enough as a daily driver? Is the other 7% really hard to get ?
That number is arbitrary, as it only includes the tests that Servo has "enabled"
Here's a link to the full WPT test results, with all the major browsers + Servo/Ladybird/Flow included:
One could argue that these numbers are still misleading, though, as there are a huge number of CJK-encoding-specific tests that inflate the results across the board. Ladybird often filters out those tests when showing WPT results (but I'm not sure exactly how to do that on wpt.fyi).
When I last worked on my personal site I got the feeling that Firefox was increasingly the one browser that didn't implement what I wanted to use, over and over, way more than I've experienced in the past. I guess that feeling is just an actual trend in reality and that Browser-specific failures (BSF) graph shows it really well:
Firefox's % of BSF failures was consistently in the teens in 2019 and is now consistently in the 30s. Here's the average % of total bsf's for each browser by year:
| Year | chrome | firefox | safari |
|---|---|---|---|
| 2018 | 11% | 24% | 65% |
| 2019 | 11% | 17% | 72% |
| 2020 | 13% | 18% | 69% |
| 2021 | 12% | 22% | 66% |
| 2022 | 12% | 24% | 64% |
| 2023 | 13% | 30% | 57% |
| 2024 | 13% | 29% | 58% |
| 2025 | 13% | 32% | 55% |
| 2026 | 12% | 38% | 50% |
There are thousands of generated tests for some APIs that virtually nobody needs and maybe only a handful for how iframe works. The ratio is distorted so a relative pass rate is unfortunately meaningless.
The stat I want to see is the number of serious dependent projects. I recently tried to use servo in a custom documentation browser application. There are so many little quirks to the process of integrating it into an application that I have to think the project assumes most people are still working on it either for the sake of working on it or for eventual inclusion in a single "mothership" project like Firefox. I ended up using webkitgtk instead.