Wayland - Beyond X (The H)

Posted Feb 22, 2012 0:31 UTC (Wed) by bloopletech (guest, #71203)
In reply to: Wayland - Beyond X (The H) by jschrod
Parent article: Wayland - Beyond X (The H)

I am honestly sorry if this comes across as trolling.

My laptop has an AMD chipset. This has necessitated a lot of messing about with drivers and broken upgrades, and basically a lot of spent time and effort. At the time this was really, really, really frustrating and angering. I often had a broken, or not-working-very-well system.

I have generally not complained about this online, because the difficulties were inherent to the status quo - ubuntu moves really fast which leaves proprietary drivers behind, the open source drivers guys were working like mad to get the drivers out, but only had limited resources.

I have also generally not complained because I don't like reading entitled whinging, so I try not to produce it myself (sorry, that was snarky ;).

But now the status quo appears to be changeable, in favour of easier/better driver development (that is only an assumption, I have no idea yet if it is true). At least, upgrading your drivers [as a user] will be much easier (this part I'm fairly sure of).

My real frustration is that there seems to be a group of people who do not understand that there are more things pulling on the steering wheel than just their needs. I do realise this applies to me too, but afaict their use case will not be massively affected, whereas my use case *might* be massively improved.

I do still think there is a disruptive sense of entitlement; but I've already written enough inflammatory material on that topic.

to post comments

Wayland - Beyond X (The H)

Posted Feb 22, 2012 0:44 UTC (Wed) by dlang (guest, #313) [Link] (1 responses)

How will Wayland change this? it will still require that the video drivers be written for the particular chip.

I really don't think that the network transparency part of X is a significant factor in the problem of getting working drivers.

At least AMD provides documentation, so the folks working on opensource drivers for their cards don't have to reverse engineer everything like the nVidia folks do.

Wayland - Beyond X (The H)

Posted Feb 22, 2012 1:47 UTC (Wed) by bloopletech (guest, #71203) [Link]

Apologies; I'll try to be more specific.

I'm sorry this information is vague, it's been a while since I've had to deal with these driver issue. My drivers work pretty well these days actually; my concern is more that new drivers for new series (oversimplifying it, AMD redoes their chips every few years, and you have to build new drivers for these new chips) will be as difficult to develop, and as difficult to install use and troubleshoot, as they were in 2010/2011.

My understanding is that driver development at the moment is intertwined with X excessively, and that this makes driver development more difficult. The X.org wiki, freedesktop project wiki and more specifically http://www.x.org/wiki/RadeonFeature have more detail on this I think; these were the resources I used mainly when trying to get my graphics to work.

I know that the driver I have installed depends intimately on your kernel version and also you Xorg version, because some parts are in the kernel, and some parts are in X, and these all have to agree etc etc.

Another more nebulous issue is that with more (and more complex) layers in the graphics stack, it becomes much more difficult to troubleshoot issues that you have. I used to have a lot of VSYNC issues on my machine, until I learned a lot of estoeric detail about how to get it to work. It also becomes more difficult to make sure the user's desire at the surface gets through to the bottom of the stack.

My understanding is that by having a smaller and clearer target to develop and test against, driver dev will be easier; and that removing layers from the stack it also becomes easier to test and configure etc.

Of course it's not X network transparency itself that makes the development more difficult; it's developing against X in the first place.

My understanding could be completely wrong; if there are any driver devs, it would be really good to hear what they think of wayland and whether it would be easier for them.

I am massively massively grateful for the work of the open source driver developers, and I hope I haven't put words into their mouths here.

Wayland - Beyond X (The H)

Posted Feb 23, 2012 23:05 UTC (Thu) by jschrod (subscriber, #1646) [Link] (1 responses)

But the post that I answered to had nothing about your difficulties with X per se, but was only about X network transparency.

We should start with the fact that when you use X on a local desktop, X network transparency is not involved at all. Since quite a long time, local graphics output in X access the graphics card directly. This was first done by the X software itself, and recently some/most of that hardware access was moved to the Linux kernel by means of Kernel Mode Setting (KMS).

The problem for Linux is that the graphics card API for fast 3D graphics is often not published by the hardware manufacturer. I fully agree with you that the state of graphic support is not satisfactory. In fact, I'm more pessimistic than you; some things that worked in my Unix/Linux environments since 1990 is more instable, since that move to KMS. That nowadays the main recommendation for non-working graphic setup is "boot with nomodeset" speaks volumes, IMHO. But I don't blame the developers, hardware has changed too much, and the road they took on the kernel side -- i.e., the hardware drivers -- seems to be a sensible one. Nevertheless I'm still not able to use VMware with 3D acceleration on standard hardware because some OpenGL primitives are not supported. And then I read here that new developments demand full GL support on notebooks, and I shudder.

Thus, I think your conception of problems in the Linux graphics stack is right on spot, it might be even worse than you see it. But that has nothing to do at all with X transparency, and Wayland won't change the situation with non-sufficient hardware documentation/drivers. So you might want to address your rants to issues where they are appropriate and leave network transparency to us who need it daily.

Cheer, and apologies for calling you a troll.

Joachim

PS: Disclosure: I am biased in favor of X, if that isn't clear by now. When I started to use X10 in 1987, and helped to port X11R4 to several architectures in 1988, I always regarded it as a flexible and trustworthy design, and still think an evolutionary X12 would be better than Wayland. We see in the Linux kernel year after year that evolutionary developments are better than big-bang approaches. This is an important lesson that many players in the graphics stack development community should heed, IMNSHO. It doesn't help that I have friends who were active in the X development community for many years, nor does it help that my own open source "home-base" is the TeX community who is equally conservative concerning user-space changes...

Wayland - Beyond X (The H)

Posted Feb 23, 2012 23:56 UTC (Thu) by khim (subscriber, #9252) [Link]

We see in the Linux kernel year after year that evolutionary developments are better than big-bang approaches.

What are you talking about? Linux kernel certainly is developed using "big-bang approach". For example O(1) scheduler was replaced with CFS. USB, WiFi, Networking Stack - everything was replaced quite a few times.

True, linux kernel developers rarely replace everything at once - but so do Wayland developers: at first only some applications should be changed (Window Managers and compositors), later other pieces will be replaced as well. It'll be quite a few years till X11 itself can be removed from the picture!