Latency Matters

I'm sick of software feeling slow, and waiting for all of these web animations, and all of this.

[Pavel Fatin » Blog Archive » Typing with pleasure - Blog about human and tech...] is an elaborate discussion of why, exactly, latency matters for modern systems, and why we might be losing it. [Pavel Fatin » Blog Archive » Scrolling with pleasure - Blog about human and t...] has the same outlook but tackles scrolling.

[The compositor is evil | Raph Levien’s blog]: a take on compositing, or how guaranteeing a full render - among other things - leads to demonstrably slower user interaction overall when X can often get you window management abstraction for free. It's an overview of lots of compositors and their interactions with the screen.

All of Raph's articles are incredibly detailed and informative, but [Advice for the next dozen Rust GUIs | Raph Levien’s blog] in particular lays plain all of the issues with developing a GUI framework today. [Fast 2D rendering on GPU | Raph Levien’s blog] also might be useful... I will need to figure out how to do this to write the manager.

As noted in that article: it looks like there is some room to "strongarm" the ecosystem, insist that a particular compositor architecture is the way to go, and develop applications fit exclusively to and for this architecture. The window manager boundary feels like a critical point at which Linux infrastructure can evolve - so it might be best to build a window manager, damn other abstractions, and fit solutions to this specific WM.

We can't really get rid of Linux - but we can provide a cohesive, high level computing environment around it - and to do this we rewrite the WM.

[Computer latency: 1977-2017] from Dan Luu is a measurement of input lag over the years.

[]: I can send an IP packet to Europe faster than I can send a pixel to the screen.

[Latency numbers every programmer should know · GitHub]

2022-11-01 35e6efb
2022-10-31 eed1984
2022-10-31 2577ddb
2022-10-30 281447e