Gregory's ServerThis happened exactly 40 years ago.
 75 comments
Ah, back then they switched from W (Wayland) to X and are slowly realizing their mistake and are migrating back to W now?   @Hunterrules0_o @hanser @nblr X11 was developed and Xorg carries a legacy from a time where computers where resource constraint and you had to be mindful about how you'd spend CPU cycles and memory. Wayland is the result of a mindset that developed in a time where screen resolutions were still small (XGA, and memory "plenty"; ~1GiB), and you could easily fit 512 screen-fulls of pixels into VRAM. 1/ These days we have 4k HDR screens. Such a screen's framebuffer comes in at 2×32MiB (double buffering). And in a compositor based graphics system such as Wayland, each every window will hog up to 32 MiB, even if minimized or otherwise hidden, so that quick preview will work. For 8k screens you've multiply those numbers ×4. Assuming a GPU with 8GiB VRAM that's just 250 windows you can fit in there. And of course you've to bear the needed memory bandwidth, too. 2/ But there's more: For a long time GPU accelerated font rendering was elusive. Eventually Eric Lengyel solved the problem with his SLUG library. However the algorithm is patented to FOSS can't use it. In 2015 I did work on GPU accelerated font rendering, too. However when SLUG got released I was wary that its algorithm might be similar to mine (I'm now confident, that my algo is sufficiently different to not infringe patents). /3 Bottom line is: FOSS has to do it's glyp rendering on CPU and either transfer the whole rendered text for each content change (more resource efficient) or has to upload humongous font atlases – good look if you want to render logographic script. So that's inducing a bottleneck. The lesson here is: The graphics related resource constraints of 40 years ago never went away. They just bode their time, and now are rearing their ugly heads again. 4/ So in conclusion: Of course a graphics system developed with the constraints in mind (because it wouldn't have worked otherwise) will perform a lot better on modern systems, than graphics systems designed toward the needs of 15 years ago. /fin @datenwolf @hanser @nblr and is still more stable than modern wayland. in fact when wayland crashes it takes all your programs down with it, The devolpers who designed wayland cant do what x did 20 years ago.   @nblr Cool. On the other hand, nowadays I've to explicitly look under the hood in order to know which window system is running. For at least the last ten years I just emerged the WM/DE of choice and it pulled a window system (most likely X.org, perhaps Wayland on my new gaming rig). Without conf editing. So, to me, X came a long way and matured fast in the last 20 years. Happy birthday!   @nblr "Anyone who wants the code can come by with a tape." We've come a long, long way!   @nblr "There is no documentation yet; anyone crazy enough to volunteer? I may get around to it eventually." 🎶 Tale as old as time… 🎶        @anatoliyl It wasn't protocol version 11 yet by a long shot. The protocol changed very rapidly in the first few years and didn't really stabilize until version 10 in 1986, followed shortly thereafter by version 11 where commercial vendors (mostly DEC IIRC) had contributed enough production-quality code to make it more than just a research prototype, which I think was in late 1988, where it's been frozen ever since -- first by broad adoption and more recently by abandonment.  @nblr I was researching distributed window systems at the time, but mine went nowhere and everyone started using X on Unix.   @nblr I had no idea that before X was W. How far back in the alphabet does it go, or was W where it started?      @nblr Did not know that the first API was in CLU. One of my favorite languages! @jef @nblr Both came out of Barbara Liskov's research group. In 1994, although Scheifler had moved on, Barbara and her group were still there, around the corner and down the hall from me. The 5th floor of 545 Tech Square was an astonishingly productive place: X, the End-to-End Principle, library operating systems, packet audio and video over wide-area networks, "soft state", RSVP, mesh networks, content-addressable file systems, TCP time-sequence plots, ...             Interesting history. Not just about W and X, but the email. They were likely on bleeding edge SMTP and POP. The use of the X logo is interesting also.  |
@nblr Anyone who wants the code can come by with a tape.