News for hackers. Stuff that matters.

A comment on Engineering for Slow Internet. Parent comment

Re:Engineering for Slow Internet by LtWorf on Friday May 31, 2024 @10:36AM (#40535759)
Surely you're aware of gzip encoding on the wire for http right?
[ Parent | on HN ]
Re:Engineering for Slow Internet by RGamma on Friday May 31, 2024 @11:45AM (#40536612)
Sure, would be interesting to know how it would fare against purpose-made compression under real world conditions still...
[ Parent | on HN ]
Re:Engineering for Slow Internet by LtWorf on Friday May 31, 2024 @09:24PM (#40542058)
gzip is fast. And it was made for real world conditions.
[ Parent | on HN ]
Re:Engineering for Slow Internet by floam on Saturday June 01, 2024 @01:28PM (#40547544)
Brotli is a better example, as it was literally purpose made for HTML/CSS/JS. It is now supported basically everywhere for HTTP compression, and uses a huge custom dictionary (about 120KB) that was trained on simulated web traffic.

You can even swap in your own shared dictionary with a HTTP header, but trying to make your own dictionary specific to your content is a fool’s errand, you’ll never amortize the cost in total bits.

What you CAN do with shared dictionaries though, is delta updates.

https://developer.chrome.com/blog/shared-dictionary-compress... https://news.ycombinator.com/item?id=39615198

[ Parent | on HN ]
Re:Engineering for Slow Internet by LtWorf on Sunday June 02, 2024 @05:52PM (#40557441)
Is it supported by curl or any browser or no?
[ Parent | on HN ]
Re:Engineering for Slow Internet by floam on Monday June 03, 2024 @12:13AM (#40559261)
Every browser, curl yes.
[ Parent | on HN ]