.png)
3 days ago
2
I contacted Enrico a while ago — before the whole drama about the fork exploded — and he told me about his plans to fork and a lot of things about the future of Xorg.
A lot is being said online about Enrico, but the reason I contacted him is that I looked around for developers committed to keep working on Xorg, and Enrico was the only one that replied.
Here’s my interview with Enrico so you can make up your own mind.
When did you start developing for xorg?
I once used to be there back when honorable old Keith did the fork (not much coding, just build fixes, testing, etc).
Came back over a year ago (frankly, don’t remember exactly 😲). My original goal was developing the Xnamespace extension (that’s now in upcoming Xlibre release), but found out there’s yet a lot of old technical debt to clean up.
The Xnamespace extension btw is yet another building block for a planned future mobile OS, which is pretty much an average gnu/linux distro, optimized for touch devices and is putting untrusted applications into containers (no need to write new apps, just use existing packages).
What’s your motivation for developing for xorg?
Many different things.
First of all, it’s battle tested, reliable tech, while the alleged alternative (Wayland – actually it’s Wayland plus a huge stack of extra stuff) just isn’t suitable at all for a wide range of use cases. For example, not long before (re)joining Xorg, I’ve been working for an automotive vendor whose control center tech heavily relies on X features (eg. I’ve customized xfwm4 for them, because they need some specific window management policies, due safety constraints)
I’ve soon found out that Xorg is undermaintained and somebody has to step in and do something about that.
Another aspect is BigTech pushing so hard for the really unfinished Wayland stuff, that I’ve became suspicious. Something’s not right here. And by digging deeper I slowly got a more complete picture: it’s classic “embrance, extend, extinguish” strategy. Quite similar to systemd and several other things. Also lets not forget how Redhat tried to proprietarize lots of FOSS code (including mine, eg in the kernel). [I’ve been one of those who were pretty close of terminating all license contracts we’ve given to them – GPL’s explicitly prepared for this]
What’s your setup? (distro, desktop environment, terminal, monitors, editor, etc.)
Pretty boring notebooks. Main one is an X1.
Do you play games on linux? If so, which ones?
Rarely have time for that. Occasionally some old DOS games, or some FOSS stuff (newest proprietary one I own is some CoC clone back from the 90s, that an old friend once gave me as birthday present 😉).
I don’t run any proprietary code. (except for efi firmware, I couldn’t replace yet 😉).
What’s your take on the Xorg versus Wayland debate?
Well, both have their place, depending on use case.
Wayland is great for (semi-)embedded systems, eg. settop boxes, game consoles, etc – where you don’t wanna have a full desktop, but just a few programs that aren’t interacting much with each other and only need something to show their graphics on screen. It could also serve as a driver layer for an Xserver (which actually was the plan once)
Note that Wayland itself is only about surface composition, nothing more (plus a little bit input routing). It was created as an experiment to explore how future composition component in the Unix/X11 stack could perhaps look like – the idea of building whole desktop directly ontop of it (without X) came much later.
Today when people talking about Wayland, they’re actually talking about a huge stack of several things, where Wayland is just a tiny part of. (so yes, Wayland itself is very small, but it also just does a very small part of the whole job). Most stuff is done by various other protocols, eg desktop portals. And that all is completely incompatible to X, so a lot of stuff needs to be rewritten, retested, recertified. It’s building the whole desktop stack from scratch. And it’s much more like Windows or MacOS than traditional Unix.
And still it lacks lots of vital X features, that many use cases just can’t live without, eg. network transparency. (and no, lossy video streaming a la waypipe isn’t sufficient here).
If you had unlimited resources, what would you change in the linux display infrastructure?
huh, tuff question.
Few things that are in the planning:
- gallium pipe: recording the gallium calls as a pipe driver (just like for an actual gpu) and sending it to remote side, eg. an Xserver or some compute node.
- replace the whole (old) GLX code by glamor/EGL. replace most of the xf86-video-* drivers by generic modesetting + glamor (for all recent enough GPUs).
- native gallium-pipe X11 extension for remote-capable 3D rendering (and optional local shortcuts, eg. shared memory buffers), as alternative for GLX and DRI.
- fully multithreaded Xserver
- some more X11 extensions (eg. server side video decoding)
- true wide-gamut display support (HDR is just an ugly hack :p)
- multi-tier display support in Xserver and remote forwarding: replacing Xinerama and DMX.
…
For the kernel side: unifying drm kernel ABIs (instead of entirely different ioctls per GPU type)
In a typical linux desktop, how much code is a) exclusively for Wayland, b) exclusively for Xorg, c) shared between display servers?
Shared are most parts of widget toolkits, mesa, etc, (and most of the applications). But the whole portals stuff is extra code, just for Wayland-world only.
What do you think about 3D graphics APIs? (OpenGL, Vulkan, Direct X, Proton, etc.)
They’re just different APIs. For mesa, they’re just frontends (terminus technicus: state trackers). Shall everybody choose what one likes best for the job. With mesa, everything’s quite smooth (except CUDA) – the only actual trouble makers are proprietary drivers –> NVidia.
What’s your opinion on NVIDIA?
Problematic corporation, ever since. Recently they’re starting to learn, opened up pieces of their kernel drivers. But the X and GL side is still a mess.
Their major problems (regarding Unix/Linux support) are:
- bad upper level management –> secrecy
- obviously understaffed
- very hackish engineering practices
- pretty problematic communication (didn’t answer any of my emails yet – need to confront them in their own user forums in order to get some responses)
I’m trying my best to get along with them, but supporting this stuff is really hard. I really wish them becoming a bit more open and cooperative.
How important is Steam for the future of linux display infrastructure?
No idea. Never used it, never looked at it. This is a special niche thats not relevant to me. The company seems to be quite some good FOSS work, that’s great. But not actually practically relevant for X11 and the professional workstation world.
Why was a fork needed?
Because Redhat doesn’t want X to continue.
Why do you think you were banned from freedesktop.org?
See above. And because I dared to talk to the journalist whose name must not be spoken 😉 And because I’m not caring at all about “DEI”.
The fork is described as “non-DEI”, can you clarify your stance on contributions from people of different political leanings?
“non-DEI” means there is no such thing as “DEI” here. We don’t need it. We’re all adult people, we don’t need a kindergarten teacher who’s lecturing us on how to get along with each other.
Anybody is welcome – political views or whether somebody belong’s to some minority or personal taste, etc, etc – don’t matter at all. All that matters that we’re coming together and create great software.
Is this the year of the Linux desktop?
For me it already is since the 90s. Don’t recall when I’ve deleted my last Windows instance – must be somewhere in mid 90s.
So that’s Enrico, the only Xorg developer that is committed to Xorg (now XLibre). I did ask on the xorg mailing list if any developer was committed to the Xorg server and no one raised their hand. Many users explained why they will keep using Xorg, but no developer stepped up. They also did not bother to explain why Enrico was banned despite multiple people from the community worried about the precedent and asking for an explanation.
I’m not going to insert my value judgement regarding the ban here, all I’m going to say is that I’ve running the XLibre server without problems (after I sent my first contribution).
Talk is cheap. Show me the code.
Linus TorvaldsI find this quote from Linus very insightful, and regardless of what Enrico’s detractors say of him, his code speaks volumes. Nobody else from Xorg is stepping up, and the new fork has already attracted new developers (which I might interview later on). In the end it’s the code that will tell the story, just like it did when Xorg replaced XFree86 more than twenty years ago.
