>>105893092
> wayland secure
Don’t send wayland a 4097 byte message though!

>>105871402
Wayland is an extreme security risk.
They just “fixed” this DOS attack that makes it burn CPU forever in an infinite poll loop by hardcoding the buffer size, and now the dynamic buffer resizing code is dead.

In the 90s, on X on sun machines, we were already using 8k packets over loopback (just memory with access controls, really).

The problem is the wayland perpetrators really don’t seem have any experience in software development, computer architecture, networking, graphics, etc, so you’ve got this thing where if you send it a 4097-byte message, it’s now permanently burning a core 100%

>>105824491
> security?
Let’s put it thus way, for the first 10 years of wayland’s life, it was subject to DOSing CPU infinite loops if anything sent it a message bigger than 4096 bytes causing a (random, hard-coded) core protocol change last year.
This is despite the fact it handles dynamic buffer resizing. Well, maybe no longer… code is there though. Dead code.
This is in the fundamental message loop.
They’re literally nowhere.