On Sun, 2018-01-14 at 11:02 -0800, Howard Howell wrote: > 3. Wyland!!??!!! I liked X. It worked! Wyland has some > quirks, including the inability to run some kinds of video cards, like > Nvidia, and while it was brutal before, at least you could get it > working. Now???? Wayland works about as well as X with the nouveau driver, as far as I'm aware. If you want to run the binary driver, X is still there, you can still use it. There are (relatively easy) technical and (relatively hard) political issues with running wayland on the binary driver, and we're working both with upstreams and with NVIDIA to find solutions. In order to know what those remaining issues are, we needed to change the default away from X sometime, so we did. > I know it is not your fault when manufacturers > (those idiots) refuse to reveal their instruction set or communications > architecture, but shouldn't some form of industry standards > organization be created to set up some standards so new systems could > be used across OS's and platforms? To the extent such an organization already exists, it's called Khronos (of which Red Hat is a member). But one does not just get to invent a standard and expect people will adopt it, the standard must both solve a problem they're having and save them money (or time or other resource). Neither does defining a standard make its implementation magically appear. The kind of standard that would enable us to support any given GPU as soon as it hits the market would look something like AHCI, essentially a register file definition for every possible feature you might want to drive. This would not solve a problem any GPU vendor is having, it would create a new one, because their existing programming model is not going to match that standard. Nor would it save them any time or money or effort, it would cost them significant amounts of each. API standards like OpenGL _do_ solve a problem those vendors have, which is how to keep high-value applications running while retaining the freedom to improve how the hardware works. Even a much lower-level standard like Vulkan still abstracts the hardware away, though to a different set of primitive operations. In order to keep pushing that abstraction layer down towards the hardware, we also work on projects like Mesa for the "save them money" part of the incentive structure. And as a result we've seen vendors start to simply adopt Mesa rather than support their own GL stack; at some point Mesa becomes cheaper. So I wouldn't call those vendors "idiots", myself. They might be self- centered, but they're not stupid. They're doing the best they can to serve the markets they have and want. If we want to change their behavior, it's incumbent on us to provide the incentive. - ajax _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx
