This is roughly what we’ve been doing with some internal Rust work as well.

https://blog.sulami.xyz/posts/sync-core-async-shell

https://quic.video/blog/never-use-datagrams

https://quic.video/blog/quic-powers

https://phys.org/news/2024-05-carbon-pricing-major-meta.html

This is a good article on TCP_NODELAY. The associated HN thread also has some interesting comments.

For another discussion of this topic, see the recent Oxide and Friends episode, also available in podcast form.

https://blog.cloudflare.com/aim-database-for-internet-quality

Worth the read

https://cacm.acm.org/research/talking-about-large-language-models/

The AT Protocol that powers BlueSky (distributed Twitter) is interesting.

https://steveklabnik.com/writing/how-does-bluesky-work

Especially the ability to use domain names as identifiers and the ability to move between servers while maintaining that identifier.

https://bsky.app/profile/coverfire.com

I’m not sure I have need for micro-blogging any more but perhaps a non-centralized version could be better.

We just published a new Podcast episode featuring David “Tubes” Tuber and Carlos Rodrigues from Cloudflare. We talked about Cloudflare’s speedtest and more generally, Internet experience.

Google Podcasts

Apple Podcasts

Spotify

This is worth reading.

https://tailscale.com/blog/quic-udp-throughput/

A few interesting things:

• More and more important traffic uses UDP
• The shift from TCP is driving optimizations that weren’t a priority before
• There’s an RFC for a QUIC perf test that includes requests per second
  • https://datatracker.ietf.org/doc/html/draft-banks-quic-performance-00
• CPU power states always matter when benchmarking