tech-x11 archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Xorg vs Wayland (and MIR?) - future for NetBSD X ?



On Tue, 27 Dec 2016, Mouse wrote:
Some thoughts from someone who has done X hackery in the past, but has felt shoved to the margin by the directions X is going....

I'd also describe myself that way, sans the X hackery. The most I've done is to fix onesy-twosy bugs in Motif and played around in the open source CDE project. I know you've gone all the way down to the driver level.

Well, as someone who's done two ddx layers (MouseX, X for 2bpp NeXT hardware, and a cgfourteen ddx that supports both 8bpp and 24bpp on the same screen),

Heh, you don't need to qualify, I know the Mouse has a long and hard-fought list of street cred and hard-code pedigrees. :-)

my answer is twofold: (1) "because the off-the-wall platforms have documented video hardware"

Isn't it a cryin' shame that so much logic goes into the driver these days that all the vendors consider it super-secret? Also it's disappointing that graphics vendors see zero advantage in publishing specs or even working very hard with external projects. As you point out, they used to.

Then again, I used to buy stereo gear that came with schematics. I'm still butthurt about that going away, too.

and (2) "because for x86 you care about running the same X Linux does".

Yeah, they have all the momentum on the graphics-drivers front. That's undeniable.

"Modern" X has lost a lot of what gave its roots such staying power. For example, it appears to no longer care about anything but 24bpp TrueColor hardware.

I didn't realize that. What is the current fashion? Doing everything at 16BPP?

It cares so much about x86 that it's willing to impose braindamage driven by the x86 video disaster on completely non-x86 machines (I saw SPARC servers doing sbus enumeration in userland, apparently because x86 braindamage makes it want/need to do PCI enumeration in userland on x86).

Ewwww. Gross. I hate when x86 leaks onto something else.

At least back in the 1990s when off-the-wall folks like Number Nine or Matrox would make cards for the Alpha, SPARC, or MIPS platforms they'd usually at least customize the BIOS and make the PCB layout hardware-friendly to the platform.

   I'm just ignorant of these dynamics.
I'm ignorant of (most of) the politics, too.

Well, we are at least both happier for that, I'm sure.

[...] the demands placed on them are lower. For example, on x86 people castigate you if your server can't use undocumented 3D rendering hardware via vendor binary blobs. You don't get that on sparc or amiga or vax or etc.

Very true. I'm sure that they don't hear from people like me who are jumping for joy just to get a decent 2D driver that works at all. VESA modes are helpful, but further acceleration there seems unlikely in the face of so many "native" drivers.

It's hard luck that the OpenGraphics guys never really seemed to lift off. That would have provided an "out" for all this, IMHO. With fully open specs, each OS could decide on the most "correct" implementation for themselves. It'd also put additional pressure on AMD & nVidia since if the alternative was cheap enough it'd probably be embraced by FOSS folks, arcade hacks, and on servers. Of course, as you point out, if it doesn't do 8 quadrillion polygons-per-second and run 3D dancing reindeer screensavers/games fast enough, lots of people would be crying like old women at a funeral.

-Swift


Home | Main Index | Thread Index | Old Index