Original note from: MuyLinux
To read Eduardo Medina’s original in Spanish, click here
The transition to Wayland has gone from promising to one of the more winding paths on the GNU / Linux desktop. The graphical protocol has encountered many obstacles, of which the complexity of its implementation and the poorly defined framework that it has drawn to face certain situations can be highlighted. However, not all the obstacles have been internal, as Wayland has also been weighed down by the despicable arts of the biggest graphics processing giant: NVIDIA.
Those who have followed Wayland’s trajectory know that NVIDIA’s relationship with the protocol has always been somewhat special, especially because of the company’s tendency not to adopt the standards agreed upon by Intel, AMD, GNOME, KDE, and the developers of composers. While x86 processor manufacturers and composers and desktop developers decided to adopt GBM as a buffer, NVIDIA opted for EGLStreams, thereby breaking the consensus that the rest had reached.
NVIDIA motivated its decision that EGLStreams is an open specification from Khronos Group (the same institution behind Vulkan and OpenGL) that offered good performance, in addition to apologizing that ‘libgbm’ (the GBM library) formed part of Mesa, an essential component of the GNU / Linux graphics stack that the company has always seen as a rival, despite being released under the permissive MIT license. Obviously, that did not appease the criticism from the community, but on the contrary, seeing the way NVIDIA acts.
EGLStreams ended up being implemented at least early stage on GNOME and KDE Plasma, but NVIDIA found the results to be a disaster. In the case of KDE Plasma, there is evidence of the company’s personal involvement, but composer developers encountered a complex panorama as they had to adapt to a very specific NVIDIA technology while the rest of the drivers supported GBM.
With the failure of the EGLStreams implementation, the pressure to set Wayland as the default in GNOME, the abandonment of Xorg, and the fact that it was lagging far behind AMD and Intel, NVIDIA had no choice but to take steps. backwards to better conform to the standards agreed upon by the rest. The green giant announced last January support for hardware-accelerated XWayland and DMA-BUF, and now we echo that the company has apparently given up on adopting GBM.
James Jones, an NVIDIA employee involved in supporting Wayland, has submitted a merge request to allow Mesa to discover and load alternative GBM backends to the built-in Direct Rendering Infrastructure (DRI) backend by default. This move can only be sustained by the fact that NVIDIA has a real intention of supporting GBM for its proprietary official driver, which would make it easier for existing composers to handle.
It is important to note that the GBM (Generic Buffer Management) interface is a means of allocating graphical buffers and can be used with EGL. This is because GBM offers an abstraction layer capable of working with all graphics drivers. Taking these data into account, everything seems to indicate that NVIDIA’s stubbornness with EGLStreams has had a lot of political and little practicality.
In short, and if NVIDIA really ends up offering what it promises with the aforementioned Mesa merger request, it means that in the future the company’s official driver should work without major complications in a Wayland session and without requiring major modifications. on composers.
For now, the NVIDIA driver version 470 promises to be much friendlier with Wayland, but looking at the background, the sensible thing to do would be to demand that the company actually perform real results in real terrain (that is, that end users really see that NVIDIA is usable with Wayland ).