Xlibre [LWN.net]

Xlibre

Posted Jun 10, 2025 20:13 UTC (Tue) by pizza (subscriber, #46)
In reply to: Xlibre by ubhofmann
Parent article: Ubuntu 25.10 to drop support for GNOME on Xorg

The reason for this fork is that its creator was banned from further commits to the main xorg repo due to repeatedly submitting PRs that introduced pretty serious regressions -- of the sort that would have been prevented had the code been even _run_ once.

So by all means, fork away, but they're going to have a really hard time convincing anyone else to care.

to post comments

Xlibre

Posted Jun 10, 2025 21:20 UTC (Tue) by pizza (subscriber, #46) [Link] (19 responses)

> So by all means, fork away, but they're going to have a really hard time convincing anyone else to care.

...and when your fork has an incompatible binary ABI (+API?), you're not going to be compatible with existing proprietary drivers (eg NVIDIA) which has been the primary reason for folks needing to stick with a native xserver.

Xlibre

Posted Jun 11, 2025 8:21 UTC (Wed) by parametricpoly (subscriber, #143903) [Link] (18 responses)

It's also a bit late to the party. Distros are switching to Wayland exclusive DEs (Gnome, Ubuntu) and Wayland basically "works" for quite many use cases as a daily driver. The network effect is directing people to go forward.

If you do objective analysis, Xorg has some limitations and Wayland comes with another set of limitations. Extending and improving X will face some issues as other platforms using X still use the old APIs. At this point it's hard to justify moving back to X even if it will get fixes and updates. Also what would you do with a windowing system if the apps are abandoning it? I'm looking at something like lightdm or sddm. The repos are basically dead.

Xlibre

Posted Jun 12, 2025 19:53 UTC (Thu) by linuxrocks123 (subscriber, #34648) [Link] (17 responses)

I haven't ever used Wayland and don't have any plans to do so anytime soon, and I expect many other plan to continue using X indefinitely. It appears Wayland is still immature in areas such as running on headless servers, automating user input, taking screenshots, and copying and pasting from the clipboard.

Maybe one day they'll be able to reliably implement those cutting edge features.

Xlibre

Posted Jun 13, 2025 2:49 UTC (Fri) by numgmt (guest, #167446) [Link] (16 responses)

GNOME Remote Desktop does headless RDP way better than any Xorg solution I could find. It's GNOME-exclusive, but hey, it works for my use case, and that was the quality of solution I needed that nothing else (e.g. XRDP) met.

Xlibre

Posted Jun 13, 2025 10:26 UTC (Fri) by lindi (subscriber, #53135) [Link]

I’m also a big fan of headless gnome remote desktop. I’ve used it for several months now in debian testing and it has been very stable. I hope in the future we could get some public key based authentication as well.

Xlibre

Posted Jun 14, 2025 12:20 UTC (Sat) by eru (subscriber, #2753) [Link] (14 responses)

Can you use it to "remote" the window of a single application, instead of a whole desktop? (This is one killer feature of X11 that I use almost every day).

Xlibre

Posted Jun 14, 2025 13:07 UTC (Sat) by intelfx (subscriber, #130118) [Link] (12 responses)

> Can you use it to "remote" the window of a single application, instead of a whole desktop? (This is one killer feature of X11 that I use almost every day).

It's not really a killer feature of X11, not for a long time. Waypipe[1] exists and in practice works much better over mediocre links than X11 SSH forwarding, much less actual X11 network transparency, ever did.

[1]: https://gitlab.freedesktop.org/mstoeckl/waypipe

Xlibre

Posted Jun 16, 2025 10:30 UTC (Mon) by eru (subscriber, #2753) [Link] (11 responses)

Thanks, good to know for the dreaded day I cannot use X11 any more. But this seems to be a complex 3.party program that might not be available in all installations at both the client and server ends. With X11, you get the functionality out of the box.

Xlibre

Posted Jun 16, 2025 11:03 UTC (Mon) by intelfx (subscriber, #130118) [Link] (10 responses)

> But this seems to be a complex 3.party program that might not be available in all installations at both the client and server ends. With X11, you get the functionality out of the box.

The thing is, you don't.

`ssh -X` is as much of a third-party program as `waypipe` is.

And in the off-chance if you're **actually** talking about X11's **actual** network transparency, i.e., `DISPLAY=somehost:0`, then that stopped working satisfactorily even longer ago than `ssh -X` (unless you limit yourself to Motif and Tk, I guess).

Xlibre

Posted Jun 16, 2025 12:37 UTC (Mon) by eru (subscriber, #2753) [Link] (4 responses)

ssh is nowadays a common piece of infrastructure with multiple uses, so it can be assumed to be available. Rather like sh, ls, cat etc. Waypipe is specialised, complex, and has a ton of dependencies. X11 remoting has worked well for my needs, obviously I am not running games or video editors through it. Anyway, I'll stop grumbling now.

Xlibre

Posted Jun 16, 2025 13:48 UTC (Mon) by pizza (subscriber, #46) [Link] (2 responses)

> ssh is nowadays a common piece of infrastructure with multiple uses, so it can be assumed to be available. Rather like sh, ls, cat etc. Waypipe is specialised, complex, and has a ton of dependencies.

So... [random X11 application] isn't specialized, complex, with a ton of dependencies? I mean, if you're trying to run an X11 application remotely, you're going to need at minimum full client xlib+whatever else on one side, and an running xserver+whatever else on the other side. In other words, by definition, both sides need to have everything necessary to run said application, and *all of it* is outside the scope of what ssh provides (ie a forwarded TCP connection)

Along that line, nothing prevents [Portable] OpenSSH adding support for wrapping waypipe, eg by adding -W and associated configuration options. There's plenty of precedence, not just with X11 (via -X) and various authentication agents but also things like scp/sftp which work by forking off separate executables (on both sides) and shuffling data between the two.

(I would also point out that openssh's -X option claims to support interacting with X11 "Security extensions" by default; I don't know exactly what that entails under the hood but it's clearly more than just setting up a port forward and setting $DISPLAY on the remote side)

Xlibre

Posted Jun 17, 2025 1:50 UTC (Tue) by raven667 (subscriber, #5198) [Link] (1 responses)

> openssh's -X option claims to support interacting with X11 "Security extensions"

I think this refers to running xauth to set an MIT Magic Cookie value in ~/.Xauthority which is like an API token that prevents other users on the same system from just connecting to your forwarded X11 port on localhost and rickrolling (at best) your screen.

Xlibre

Posted Jun 23, 2025 8:29 UTC (Mon) by cortana (subscriber, #24596) [Link]

The SECURITY extension allows certain clients to be marked as "untrusted" which prevents them from being able to spy on input events, take screenshots of other clients and so on.

Unfortunately many clients break completely under such restrictions. For many years, Debian patched OpenSSH to disable the use of the SECURITY extension by deafult. Nowadays I think the situation is a bit better but I've not used X11 forwarding for a long time so haven't verified.

Xlibre

Posted Jun 16, 2025 16:11 UTC (Mon) by intelfx (subscriber, #130118) [Link]

Of course, OpenSSH (not just ssh, but very specifically OpenSSH) is going to be more widely available than Waypipe, simply by virtue of being an older and more established technology (as well as historical precedent). However, we weren't discussing that:

>> <...> With X11, you get the functionality out of the box. <...>

What we were discussing is that both are structurally *third-party software*. There is nothing conceptually "out-of-the-box" about `ssh -X`, not any more than Waypipe. Whether the latter is significantly more specialized and complex is in the eye of the beholder.

> X11 remoting has worked well for my needs

Nobody was disputing that. We were talking about whether it actually is a "killer feature".

Xlibre

Posted Jun 16, 2025 13:13 UTC (Mon) by paulj (subscriber, #341) [Link] (4 responses)

Is there a replacement for XPRA?

I still use ssh -X for some very graphically-limited apps - xterm mostly - but for anything using a modern toolkit you really need XPRA to get good performance. And the persistence / immunity to transient network issues/moves is a nice bonus in some cases; and absolutely essential in other cases. E.g., I use xpra to access the same instance of an app between work and home, without having to restart the app and my flow in it.

Xlibre

Posted Jun 16, 2025 16:13 UTC (Mon) by intelfx (subscriber, #130118) [Link]

> Is there a replacement for XPRA?
>
> I still use ssh -X for some very graphically-limited apps - xterm mostly - but for anything using a modern toolkit you really need XPRA to get good performance.

Waypipe does not need anything like Xpra to get good performance; you get it by default.

> And the persistence / immunity to transient network issues/moves is a nice bonus in some cases

There is no persistence, however. Earlier versions of Waypipe had rudimentary support for reconnection, but it was since dropped — not sure why, perhaps lack of interest on the sole developer's part.

Xlibre

Posted Jun 17, 2025 1:58 UTC (Tue) by raven667 (subscriber, #5198) [Link]

I think in other cases with Wayland the compositors just support RDP or VNC output, VNC being slow but simple and RDP working well over WAN links with various acceleration options available in the protocol (maybe including just encoding as a video stream, same as any modern WebRTC video conferencing and desktop sharing system) if the server takes advantage of them, degrading to something like VNC if not. I think Waypipe is more for the single-app forwarding use case (although I believe RDP protocol is capable of single-app forwarding), and I'd be super happy when it gets integrated directly into Portable OpenSSH or as a Linux-distro maintained extension for the popular platforms.

Xlibre

Posted Jun 17, 2025 8:50 UTC (Tue) by vasvir (subscriber, #92389) [Link] (1 responses)

Well there is wprs (xpra for wayland): https://github.com/wayland-transpositor/wprs

I use it as my daily driver and while it has its issues it is workable...

I have filed some bug reports and some of them have been fixed.

Xlibre

Posted Jun 17, 2025 9:50 UTC (Tue) by paulj (subscriber, #341) [Link]

Great news, thanks for the link!

Xlibre

Posted Jun 23, 2025 4:20 UTC (Mon) by wperkins (guest, #767) [Link]

Yeah! This is one feature that I have used every day for many years!

Xlibre

Posted Jun 12, 2025 7:25 UTC (Thu) by daenzer (subscriber, #7050) [Link] (8 responses)

> The reason for this fork is that its creator was banned from further commits to the main xorg repo due to repeatedly submitting PRs that introduced pretty serious regressions -- of the sort that would have been prevented had the code been even _run_ once.

That’s not quite accurate. He was banned from gitlab.freedesktop.org as a whole for violating the corresponding CoC, not for the low quality of his xserver MRs.

(As a member of the xserver project, I do wish the project itself had stood firmer against his flood of mostly-churn MRs though)

Xlibre

Posted Jun 12, 2025 19:46 UTC (Thu) by linuxrocks123 (subscriber, #34648) [Link] (7 responses)

Any details on this? He clearly has some weird and disturbing political views, but did he actually _do_ anything, or is this a thoughtcrime situation?

Xlibre

Posted Jun 13, 2025 0:45 UTC (Fri) by daenzer (subscriber, #7050) [Link] (6 responses)

I don’t know any more details than you. The reasons for CoC action are generally not shared publicly.

The freedesktop.org CoC committee generally bans users only after multiple infractions and unfruitful mediation attempts, so he most certainly has „done something“ though.

Xlibre

Posted Jun 13, 2025 9:35 UTC (Fri) by paulj (subscriber, #341) [Link] (5 responses)

"Justice must be done, and must be seen to be done".

Behind closed door enforcement may be problematic.

Xlibre

Posted Jun 13, 2025 10:42 UTC (Fri) by Wol (subscriber, #4433) [Link] (4 responses)

> Behind closed door enforcement may be problematic.

Open enforcement may be problematic.

Damned if you do, and damned if you don't.

Especially given that - in many cases - SWIFT enforcement may be necessary to protect the victims, and sometimes that can harm the perpetrators more than necessary. That sentiment has been expressed in these comments. And as a gentoo user I get the impression the project was seriously harmed (a good few years back) because some "bad actors" weren't dealt with quickly.

If you're in a position where you need to say "stop this NOW!", the last thing you need is a load of public bike-shedding. At the end of the day, you need a BDFL who you trust really is B. I don't think I'd do a particularly good job, but I like PJ's definition she used for Groklaw. "If I wouldn't have it in my living room, I won't have it on Groklaw". Simple, clear (pretty much), and there's no comeback.

Cheers,
Wol

Xlibre

Posted Jun 13, 2025 10:52 UTC (Fri) by paulj (subscriber, #341) [Link] (3 responses)

Didn't say it had to be slow, or had to be by committee or vote. It should be stated what for and why though, should it not?

Xlibre

Posted Jun 13, 2025 12:41 UTC (Fri) by daroc (editor, #160859) [Link] (2 responses)

While I agree that public decision-making is important for good community-building, I think this thread is wandering increasingly off topic. Let's leave it here.

Xlibre

Posted Jun 13, 2025 14:14 UTC (Fri) by paulj (subscriber, #341) [Link] (1 responses)

Course. I think at some point it could be useful to have an article in and around this topic, if there is some appropriate news to thread it around. There have been a number of cases of CoC enforcements that you've reported on where the grounds and process has been murky (and acknowledged as such in at least 1 case I think). Some way to distill the meta issue of "Justice should be /seen/ to be done" out of the various individual cases and see what different approaches there have been and pros/cons, might be useful. Comments will be fun too I'm sure. ;)

Xlibre

Posted Jun 23, 2025 13:32 UTC (Mon) by Wildfurangelplumes (guest, #177995) [Link]

I get the push for transparency, but projects like freedesktop.org aren't democracies. They're run by maintainers who have the right to manage their space, and that includes enforcing CoCs without turning it into a public spectacle. IMO open source doesn't mean every decision needs community consensus, behind-closed-doors processes are often the only way to avoid bikeshedding and harassment. As for Xlibre, the fork isn't the problem in itself. It's the fact that it breaks compatibility and the dev has a history of publicly toxic behavior, including anti-DEI rhetoric. You can fork, sure. But people are also allowed to say "no thanks."