finds.dev

We-Amp/pagespeed-optimizer

mod_pagespeed optimization worker, nginx module (container images), management console and ASP.NET Core middleware: WebP/AVIF image, CSS/JS and critical-CSS optimization with a shared zero-copy cache. Apache-2.0.

View on GitHubHomepage ↗

★ 1CApache-2.0updated Oct 2026

// the take

A revival of Google's abandoned mod_pagespeed: a C++ optimization worker plus thin nginx/Apache modules that transcode images to WebP/AVIF, minify CSS/JS, and inline critical CSS, with a shared mmap'd cache between the web server and worker. It's for ops teams who want PageSpeed-style automatic optimization at the server/edge layer without re-architecting their origin, and who are comfortable running C++ in the request path.

The zero-copy mmap cache shared between the web-server module and the worker process is a genuinely good design — it avoids IPC/socket overhead on cache hits, and graceful degradation (serve original content if the worker is down) means a crashed optimizer can't take the site down. Capability-based variant caching (WebP vs AVIF, Save-Data, mobile vs desktop) handles a real and usually poorly-solved problem correctly. The test setup is unusually serious for a 1-star repo: ASan+UBSan+LeakSan+TSan sanitizer runs plus full E2E in Docker Compose, which matters a lot for a C++ process parsing untrusted content.

This is a C++ process parsing arbitrary third-party HTML, CSS, JS, SVG and images sitting directly in the request path — that's the exact attack surface that made the original mod_pagespeed a recurring source of parser-level CVEs, and reviving it doesn't remove that risk. The build is locked to clang-20 + libc++-20 specifically (older libc++ lacks std::atomic_ref, GCC trips abseil constexpr bugs), which is real friction on any LTS distro's default toolchain. The headless-browser component (CDP client, Chrome process management) for critical-CSS extraction adds a heavyweight, failure-prone dependency for a feature usually done statically. At 1 star / 2 forks with no production track record, and given Google walked away from the original project's maintenance burden, there's no signal yet that this fork can sustain the C++ security patching cadence this kind of software needs.

We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →

// at a glance

Stars
1
Forks
2
Language
C
License
Apache-2.0
Last commit
23 Oct 2026

// nearby

If pagespeed-optimizer isn't quite it.

All repos →