Learn about Linux
Download Linux
Get Linux helpSpecial Offers
Feature
Ubuntu Developer Summit report: X.org improvements, driver controversy, and bling
Share
Print
CommentsRight now, Ubuntu "just works" on a lot of hardware. However, when X.org fails, it's possible for users to run into a unrecoverable failure. Developers discussed how to fall back to a mode that's likely to work on all hardware, and talked about providing a GRUB boot option to allow users to activate a "safe" mode -- like VESA 800x600 at 256 colors -- that will allow them to try to configure X.org in a more pleasing mode.
One of the possible configuration tools that was given serious consideration was SUSE's SaX2, which was acknowledged as a comprehensive tool for munging X configurations -- particularly dual-head configs.
What's coming in X.org
However, that may only be an interim solution. X.org's Keith Packard turned up for several of the X.org discussions, and he indicated that X.org will be solving some of these problems upstream in the 7.3 timeframe. According to Packard, X.org 7.3 shouldn't even require an xorg.conf -- everything should be autodetected at run time. X.org will receive information about new devices dynamically from D-Bus. X.org will also be moving to fontconfig to make fonts more dynamic.
| |
| Keith Packard talks about upcoming X.org releases - click to view video |
If all goes as planned, X.org will be able to deal with configuration changes on the fly by the time Feisty+1 is released. X.org's development cycle is out of sync with Ubuntu's. Ubuntu releases typically land in April and October, while X.org's biannual cycle puts its releases almost squarely between Ubuntu releases.
This means that the Ubuntu developers only need implement a stopgap solution for Feisty, as 7.3 is expected to be released between Feisty and Feisty+1. There's little sense in Ubuntu developers putting a lot of effort into a tool that will be made obsolete in less than a year.
Packard says that 7.2 will be a "modest" release, mostly driver fixes and portability work. At this point, because 7.2 is close to release, Packard says that they're more focused on the 7.3 release.
In addition to better dynamic configuration of input devices, Packard says that X.org X11R7.3 should also feature better output support. He says that X.org is in discussions with card vendors on getting better hotplug support for adding monitors and such on the fly. This will be a nice addition for Linux users who give presentations and need to be able to connect a projector to a laptop and have it "just work" while X.org is running.
Binary drivers
The announcement that Ubuntu will ship binary drivers by default in Feisty is getting a lot of negative commentary from users and Ubuntu members alike. Of course, there's also a vocal contingent that complains that Ubuntu and other distros are unsuitable for general users because they don't ship with Nvidia or other binary drivers enabled. There's no position here that will satisfy all users.
At any rate, Shuttleworth pointed out that this is not a new policy. Ubuntu has shipped binary drivers, enabled by default, for wireless devices. For those users who are set up to get online via a Wi-Fi connection, this is the only way to allow them to get online and get updates at install time.
However, Shuttleworth says that "Feisty will actually warn you that you're running proprietary drivers and point you to alternative hardware ... so you're better educated for your next hardware decision. I don't expect this to be very popular with Nvidia or ATI, or with manufacturers of proprietary-only Wi-Fi cards, but of course the easiest solution for them is to open source their drivers."
Easiest? Perhaps not. One thing that many users forget is that a company like ATI may not own all of the intellectual property for its drivers outright. In one of the sessions to discuss X.org in Feisty, Packard mentioned that it was unlikely that ATI would be able to open source its drivers, but, he did say that they were trying to find ways to open source the documentation for the devices to allow the open source community to write better drivers for ATI cards.
|
| |
| Quinn Storm talks about X.org - click to view |
Owners of Nvidia cards may have a better option for 3-D in the not-too-distant future. The open source nv driver is suitable for 2-D but not 3-D work, and the proprietary Nvidia drivers are widely considered a Bad Thing™ for licensing reasons as well as practical concerns about security. Therefore, the Nouveau project is developing an open source 3-D driver for Nvidia cards.
The project is still in its early stages, but Packard suggested that it might be ready in time for the X.org 7.3 release. If that happens, the general consensus seems to be that Nouveau would not be on par with the binary Nvidia drivers, but would be able to provide sufficient performance for 3-D desktop effects.
Shuttleworth says that he thinks that shipping proprietary drivers is the best way to allow users to experience the rest of the free software stack. "Is the 3-D functionality a necessity, or a nice to have? My view is that it's a fundamental chunk of the hardware that we haven't been taking advantage of.... Now we'll enable them [binary drivers] and let the free software guys innovate on top of that."
You ain't got a thing, if you ain't got that bling
The innovation that Shuttleworth is referring to is, of course, the work going on with Compiz and the Beryl project.
Ubuntu Feisty will almost certainly ship with desktop effects available by default, but the question is whether Feisty will ship with Beryl or Compiz. It seemed that developers were leaning toward Beryl, but that remains to be decided at the closed meetings being held by Canonical employees and Ubuntu developers this week to go through all of the specs from the UDS and decide what will and won't be included in Feisty.
Beryl was originally a fork of Compiz, but lead developer Quinn Storm was clear on the fact that the Beryl project members bear no animosity toward the Compiz project, and that they're not trying to "kill" Compiz. However, the Beryl developers felt that Compiz was not open enough to accepting work from outside developers, and they felt it was necessary to fork Compiz to be able to innovate at their own pace.
Storm says that Beryl has a similar relationship now to Compiz as Ubuntu does to Debian. "We have gone our own way because we want to do things that are a bit more fresh, that are a bit more, perhaps, exciting. Compiz has their own goals and their own path, and so do we."
One of the biggest goals for Beryl, according to Storm, is to listen to the wishes of users and keep a "fuzzy line between those who are involved with development and those who aren't, so anyone who wants to come and help out our project can."
The Beryl developers put on a demo of their desktop effects on Wednesday morning. In addition to the usual wobbly window and spinning cube effects, they showed a "burn" feature that would burn down a window when it was minimized (Shuttleworth joked he'd like to make that the default on April 1) and accessibility tools like full and partial screen magnification, color inversion, and other features to make the screen easier to read -- which is a fairly compelling argument for enabling "bling" on the desktop, if it also provides a demonstrable benefit for accessibility.
Storm says that the 0.2 release of Beryl is scheduled for February, and even if it is not the default compositing and window manager, it should be available in the Ubuntu repositories.
Comments
on Ubuntu Developer Summit report: X.org improvements, driver controversy, and blingNote: Comments are owned by the poster. We are not responsible for their content.
Reliability: This will remove several big issues people have with Linux. A common complaint seems to be that when using Firefox in Linux, there is tearing as you move windows over Firefox. Using a compositing manager removes this problem. Another great example is minimizing. It makes the system feel smoother and more reliable.
Ease of configuration: Enabling important accessibility features, such as magnification, is much easier with a compositing manager. Additionally, configuring some more advanced features, such as multiple desktops, can be made more clear and concise by using effects to demonstrate what is happening. While some of the effects probably are crack (such as the wobbly windows, although thats useful for showing that windows stick to other windows, the wobbliness probably needs to be reduced so its hardly noticeable or not there unless the window is sticking to something else)
Speed: More CPU time will be available as the CPU will no longer be rendering the graphics. The GPU, which is usually idling on most computers, will now be rendering the graphics. As the GPU is usually idling, the wait time for it to become available is usually nothing, whereas the CPU is usually a little longer then nothing, as its also running the program.
1.) Reliability: When I do my apt-get updates, yes, those binary blob video drivers have stopped working. This isn't a bug in Debian, Fedora, or Ubuntu. It's a bug in the binary drivers--the very fact that *you* can't update them to account for your needed kernel/library updates. On top of that, they're rife with security holes. Oops, so much for reliability. The fix for all these issues? Go back to the 2-D open-source drivers from stock X11.
2.) Ease of configuration: those proprietary drivers make it a royal PITA to configure my system to do what *I* want...which is...actually function without fuss. These binary blob drivers have not allowed me to set, say, 1600x1200 at 32-bit color, and yes, I've got the video RAM to do it (128MB). I had to go to 16-bit color. On the other hand, the stock X11 drivers let me do whatever I want, be it from the GUI or the command line.
3.) Speed: Unless you're referring to "speed at locking up X11," I can't agree with this one, either. Those binary blob drivers have locked X11 on me more than once. It is only because I have sshd running on my boxes that I was able to remotely get into them and do an init 3 to put me at the command line. Then I had to hack<nobr> <wbr></nobr>/etc/X11/xorgconfig to put things back to the stock, open source X11 drivers...which worked perfectly. Oops, we're back to the "Reliability" problem. What does it matter how fast you think your video board will go...if your X11 subsystem gets hung up by the binary blob driver, thus preventing you from using it in the first place?
Just to be safe and make sure that this proprietary crap that crashed my GUI was GONE, I re-installed the OS. I wanted to be absolutely sure, as this is the only true cure for a rootkit...which I consider these blobs to be.
Just use the Intel video, and things will simply work. Intel's drivers are now Free Software. To hell with ATBlind and nVidious.
Anonymous
You have to realize that your statement is a personal opinion and your only speaking for yourself.
Why not give users the choice? An informed choice. You can shove open source and free ideals down peoples throats. Give everyone the drivers, let them know that they are not free and move on. If a user does not want to use them, they won't.
Also would be nice to be able to increase the size. YouTube and Google has much better flash media players.
Sorry for going off-topic.
Also, please make "matrix code rain" effect, that would be geek++.<nobr> <wbr></nobr>:D
Or make so instead of the desktop wallpaper be an still picture, it be an real-time rendered spiffy jiffy screensaver like cool thing.<nobr> <wbr></nobr>:D
By the way, 7.3 sounds really exciting!
Windows had this several years ago. I recall something doing it on win98 which was written by one of the people that used to do winamp visual plugins. 6+ years later, I've still not seen anything like it for X.
for example i want to run the helios screensaver<nobr> <wbr></nobr>... i would do this
helios --root
if you want it to run on X startup just adding
helios --root & to your<nobr> <wbr></nobr>.xinitrc would enable it.
There are a few issues like the fact that X doesn't draw anything which is obscured by other windows, meaning that some effects can get screwed up when windows are placed on top of them (Composite is meant to get around this, but I haven't tried it with a screensaver hack before)
Last time I tried Compiz though I couldn't get root-window-drawing stuff to work in it (specifically ChBg), so I don't know if this technique will work in a 3D X server (XGL/AiGLX)
* <a href="http://www.geisswerks.com/drempels/index.html" title="geisswerks.com">http://www.geisswerks.com/drempels/index.html</a geisswerks.com>
Yeah, thats what I mean, something like that!
rio is the 9P protocol which makes it network transparent even though it doesn't include any network-aware code. rio can run inside rio.
I think Plan 9 use the P9 protocol for graphics, audio and pretty much everything, and its all network-transparent, and all use one protocol instead of 15 different protocols for 15 different things. Sounds cool.
How does it compare to Twin, XGGI, Y Window System, Fresco, XDirectFB, etc?
Not really sure. I've never spent much time evaluating these simply because it seems very unlikely that they'll be adopted. If the Y page is any indication, that project is dead, or at least comatose. The last release announced was in 2004. The latest news on the Fresco page was from 2003.
AFAIK, none of these were discussed in any detail at the Ubuntu summit, so they're really not germane to the article.
They can call it Ubuntu Pure.
Then those people who want to use Ubuntu without non-free components could download this version which may have problems on some hardware.
Looks like Canonical's got the Vista Virus - they can't keep their product plans focused on functionality, compatibility and security. They still can't get the concept of simplicity, and, for those who want all that crap, to point them to some supervised offsite repository... No, they have to build it in to present to the clueless masses that the product is "cool" or "hip" or "word" or "bling" or whatever is the appropriate phrase these days.
I hear that even some of the pimply-faced uber-Linux-geeks over there are saying the same things; that they need to get all the basic tasks of an OS complete before worrying about the color of the box...
Oh well, "software for humanity"? Only if you pay no attention to the man behind the curtain...
MRK
Maybe it doesn't, but it makes it prettier. And that is a positive value. If it keeps the stability etc. of the software the same, but makes it prettier, why NOT it shouldn't be done?
How do desktop-effects make the system better (instead of just prettier)? Well, you can have visual cues that help the user (think of the genie-effect in OS X). You can make existing features more tangible, which makes it easier for users to get the hang of it (for example, virtual desktops being different sides of a cube). They can also make the system faster, when the GUI is handled by the GPU, instead by CPU. Today, the GPU sits mostly un-used. It can also make the GUI more pleasant to use, when you can elimiante the jerkiness and jaggies.
"they can't keep their product plans focused on functionality, compatibility and security."
So, if they add one feature that makes the system prettier, it means that they have abandoned everything else? It doesn't quite work that way. You CAN have a system that is pretty, and usable, stable and secure at the same time.
You are basically complaining because in the near future we will have the Ubuntu we have today, with bunch of improvements all over it. And among those improvements we have a feature that makes the system prettier. Answer me this: why shouldn't they do it? Are you saying that they should only do things that make the system "more stable, functional, productive and simple"? Every feature that does not directly and tangibly add to those qualities should be explisitly banned. Is THAT what you are saying?
Last time I checked, Ubuntu ships with games. How do games make the system "more productive"?
Lots of people who are not "pimple-faced uber-Linux-geeks" as you call it actually do like this eye-candy stuff.
You could run an default ugly theme from 95, or you could use a nice good looking desktop. You could use a simple window manager like Ratpoison, ION, LarsWM, etc or you could use a bigger desktop environment like KDE or GNOME.
I for one welcome our graphics hardware accelerated desktop overlords.
I would argue that a lot of the desktop effects are just eye candy. However, I find the thumbnail/expose effect to be particularly useful.
Also, I watched one of the Beryl demos where they showed off the screen magnification features that were enabled as part of the desktop effects package. That technology is particularly useful as an accessibility feature for users who have impaired eyesight. (Also probably useful for some graphics work.)
Ubuntu is hardly the only Linux distro that's trying to get the desktop effects stuff smoothed out -- it's already shipping in Fedora and SUSE Linux, and I expect it will take hold in most distros over time.
Why? Because you hate Microsoft? Not everything that microsoft creates is bad, and C# /<nobr> <wbr></nobr>.net "could" (conservatively speaking) be better then java. I dont want to stop the flame war up here, but why drop something which consumes negligible amount of disk space and no system resources until you run it.
out of java, c, c++ and<nobr> <wbr></nobr>.net I prefer to write code in<nobr> <wbr></nobr>.net / mono...
- lariva
I dual boot Windows, so I have no problem with binary stuff touching my machine, although I would still like to see ATI and Nvidia open up their drivers. There's no need to penalize myself for their shortsightedness.
Good news (mostly)
Posted by: Anonymous Coward on November 14, 2006 07:53 PM"Shuttleworth says that he thinks that shipping proprietary drivers is the best way to allow users to experience the rest of the free software stack."
Unfortunately I have to disagree with Shuttleworth on this point. Most Linux users (with a graphics card that supports 3D acceleration) will want to try the 3D desktop effects but it's also a feature that most users will turn off once the novelty fades out (and it will fade very quickly). These 3D desktop effects are not worth compromising the open source spirit of Linux.
Still, it's good to hear that Ubuntu actively explores all the available open source alternatives. Maybe we'll get the open source Nvidia drivers by default in Feisty+1?
For the Linux desktop to win new users, I think the following three features are far more important than the 3D effects: reliability, ease of configuration & speed. Continually evolving X.org and good open source drivers for graphics cards should make sure that Linux on the desktop has a bright future!
#