Not logged in
Log in now
Create an account
Subscribe to LWN
LWN.net Weekly Edition for May 23, 2013
An "enum" for Python 3
An unexpected perf feature
LWN.net Weekly Edition for May 16, 2013
A look at the PyPy 2.0 release
Wayland and Weston 1.0 released
Posted Oct 24, 2012 16:06 UTC (Wed) by nix (subscriber, #2304)
Most users of X do not want or need network transparency.
I know I couldn't get my daily work done without X transparency. I don't presume to know whether this is true of others, but the people who claim that I am a very rare case presume to know that with just as little data to back them up.
Posted Oct 24, 2012 16:11 UTC (Wed) by mpr22 (subscriber, #60784)
Posted Oct 24, 2012 16:42 UTC (Wed) by raven667 (subscriber, #5198)
Posted Oct 24, 2012 17:35 UTC (Wed) by daniels (subscriber, #16193)
To me, having a proxy compositor which forwards everything on for you and takes care of the details (including deciding which compression method to use) sounds a lot more appealing than that.
Posted Oct 31, 2012 17:19 UTC (Wed) by renox (subscriber, #23785)
Because if you add in the toolkit/application optimizations for remote display on top of a protocol designed for local use only, it is very likely that there will be interoperability issues (toolkit X with display server Y..), even if 'tookit X with display X' can be better optimized than a generic protocol.
Then again KH has said that he optimized Wayland to reduce the number of RTT, so we'll see, maybe a Wayland proxy will work well even in remote display..
Posted Oct 25, 2012 10:48 UTC (Thu) by dgm (subscriber, #49227)
Even you know people besides yourself, don't you? Come on, just and count the people around you (co-workers, family, friends) that actually need a remote X display. If you do it honestly you will see that the number is diminutive compared with the amount that just run their code in the same machine used for display. And even those who do need a remote display, usually do not have very sophisticate needs.
Posted Oct 25, 2012 21:24 UTC (Thu) by njs (guest, #40338)
I'm amused that xpra has become the feasibility proof for wayland remoting, and here your argument for X is the entire reason xpra exists in the first place -- the X protocol is so unbearably sucky at this exact use case. (Home DSL -> university compute cluster = nasty latency, university wifi is flaky and compute jobs are long, screen/tmux are insufficient because you need visualizations, and PhDs have to be procrastinated on *somehow*...)
Posted Oct 26, 2012 2:51 UTC (Fri) by drag (subscriber, #31333)
I agree, sorta.
Remote desktop and remote applications are now _MAINSTREAM_.
It really opened my eyes working in a large scale corporate environment that HEAVILY depends on remote applications, remote desktops, virtualized desktops, thin clients, think clients, and everything in between for ALL users for pretty much everything. Between remote applications and html-based applications pretty much everything I use except my browser is a remote application of one thing or another and that is the way for almost everybody else around me.
But you know what?
X11 has nothing to do with it at all and would be nearly useless if somebody tried to be use it in the same manner. After seeing what people are now doing with remote applications the idea that something like X11 and it's idea of 'network transparency' is competitive or unique is silly. Microsoft Windows, especially combined with third party solutions like those provided by Citrix, have surpassed it.
OS X and Microsoft Windows has very good X Windows support. OS X has it nearly-native and Microsoft Windows, when combined with third party applications, has pretty much any and all X11 features you'd care to use, including GLX support. And, of course, you see none of those users clamoring for X11 support for any of the applications. What is more they have no problems when they feel the need to do things remotely in a relatively effortless and secure manner.
I am confident that if Wayland developers decided to incorporate some form network support in their system, even if it's just primitive as pixel pushing with jpeg compression (and they take security serious) that few people will find that lacking compared to X11. In addition to that you are not giving up X11 support by using Wayland. So if you need it: it's there.
If X11 'network transparency' is wanted so much then providing the option of NOT using X11 should not be a threat at all. If it's desirable then people will continue to develop and use X11 exclusively even if they can bypass it completely for their windows api needs.
In addition if they ever decide to be competitive with Microsoft Windows and Citrix in terms of network transparency and create a X12 X Windows protocol then Wayland could, using it's compositing approach, incorporate X12 support much easier then if we were still using our X11 X Servers to twiddle bits on our PCI buses in the traditional manner. :)
Posted Oct 26, 2012 3:15 UTC (Fri) by dlang (✭ supporter ✭, #313)
This is only true in the short term
As soon as some graphics toolkit decides that supporting both X11 and Wayland is 'too hard' you no longer have the option.
just look at the flap over GNOME maintainers deciding that supporting both systemd and non-systemd linux systems is 'too hard'
Posted Oct 26, 2012 3:35 UTC (Fri) by rahulsundaram (subscriber, #21946)
Posted Oct 26, 2012 4:44 UTC (Fri) by dlang (✭ supporter ✭, #313)
But we are concerned that the ability to run an X based desktop is going to go away
compatibility with wayland is a one-way street.
if the app is written for wayland it can talk to wayland
if the app is written for X it can talk to wayland
this is being excused by claiming that the major graphics toolkits support both, so you will never find an app that is wayland-only
I'm saying that this is only the case until some toolkit maintainer decides that it's "too hard" to continue to support X, after that, all software that uses that toolkit will be wayland-only.
If they were to make a shim layer that would let a wayland-only app show up on an X desktop, they would go a long way towards relieving this concern.
Posted Oct 26, 2012 6:12 UTC (Fri) by mjg59 (subscriber, #23239)
Posted Oct 26, 2012 11:34 UTC (Fri) by daniels (subscriber, #16193)
Weston has had, pretty much since day one, a really competent nested-under-X11 compositor backend. It works pretty much like Xephyr, in that your desktop is rooted rather than being window-by-window, but it works great.
If the clamour of people who genuinely need X11 and cannot live with anything else is so great, a solution for this will definitely emerge.
Posted Oct 27, 2012 8:41 UTC (Sat) by paulj (subscriber, #341)
Apps in the future will continue, overwhelmingly, to be written to one of a small number of toolkits. At least some of these toolkits, including the 2 most popular, *already* support multiple rendering outputs (from X11, to Windows, to direct framebuffers, even to HTML!). Hell, they *already* support Wayland!
Your fear is unfounded. You can relax, your apps will, overwhelmingly, continue to work with X11, Wayland, whatever, without you or your apps having to care. X11 will continue to work, even if fewer people use it.
Posted Oct 26, 2012 8:25 UTC (Fri) by dgm (subscriber, #49227)
That would be true if you decided to base your work on a closed toolkit. Good luck then, anyway.
Posted Oct 26, 2012 13:34 UTC (Fri) by drag (subscriber, #31333)
Like I said before:
If X11 networking is valuable then desktop programmers and users will continue to insist on it's usage.
Claiming that having the _option_ to not use X11 will lead to it's demise on the Linux desktop is effectively just admitting that the majority of people not only do not really on X's features, but would actually try to avoid using it because better alternatives exist for programming GUI applications.
The way I look at it is that X11 networking is inadequate at a best and if people would like to see it's continued existence then it's up to the X folks to make it more attractive then it's competition.
In addition to that due to it's API agnostic nature Wayland can actually facilitate the usage of a 'Improved X' then the current X Server-as-display-manager approach to things. However, honestly, I think that in the long term a improved X is not going to appear and people will use other methods for their remote desktop/application needs.
Mac OS X suppot is just about awful
Posted Oct 26, 2012 12:04 UTC (Fri) by dps (subscriber, #5725)
The killer failure of me is that I can *not* run mac os X application remotely, so I can't that M$ office as part of my desktop environment (which has a much bigger display). Unless you use relatively rare X extensions, most obviously xshm. that does not apply if you use X11 applications.
I am not arguing for the same protocol but I anything that does not allow me to mix normal applications running on multiple systems in a single desktop is several steps backward. Mac OS X fails the "can I run local applications on a remote display?" test. I hope wayland will not.
Posted Oct 26, 2012 16:23 UTC (Fri) by daniels (subscriber, #16193)
'Relatively rare'?! MIT-SHM is used for pretty all image transfer (which is most everything these days) by every major toolkit, and has been shipped with X.Org/XFree86 servers since 1991. It's actually of legal drinking age in the US.
Posted Oct 28, 2012 15:54 UTC (Sun) by nix (subscriber, #2304)
Posted Oct 25, 2012 23:53 UTC (Thu) by jschrod (subscriber, #1646)
Can you please post a link to reviewed research, beyond personal anecdotes, that support that claim?
Posted Oct 26, 2012 8:14 UTC (Fri) by dgm (subscriber, #49227)
Posted Oct 26, 2012 16:15 UTC (Fri) by cortana (subscriber, #24596)
Copyright © 2013, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds