no sound in glx since installed new ubuntu

[divVerent]

Maybe there were legal reasons for that. fglrx is not GPL.

[Alien]

This is not a reason to release in such state. So patch Xorg if you can't patch fglrx.

Talking about alsa:

application -> alsa-lib pulseaudio backend-> pulseaudio server-> alsa hardware layer
instead of application -> alsa-lib -> alsa hardware layer. Of course, it's a matter or personal preference.

[divVerent]

There are good reasons NOT to use alsa-lib directly, and to have a sound server in between, namely ALSA's bad support for multiple sound sources on one sound device. "dmix" is known broken, and what else is there? Nothing.

So they made a sound server, hoping it does that in a better way.

Also, the sound server can be accessed by apps directly. This has the advantage that apps do not NEED to be specifically written for ALSA, which is for Linux only. A pulseaudio app can be used on many other operating systems too.

Now for non-pulseaudio apps, the path must be alsa-lib -> pulseaudio -> alsa-lib -> alsa-kernel and NOT alsa-lib -> alsa-kernel, as only one app can use the alsa sound output in the kernel at once for many sound cards.

Or do you think

alsa-lib -> pulseaudio -> alsa-lib -> dmix -> alsa-kernel
alsa-lib -> dmix -> alsa-kernel

is any better? That way, pulseaudio-using apps have to go through TWO sound mixers. That would be REALLY insane...

[divVerent]

This is not a reason to release in such state. So patch Xorg if you can't patch fglrx.

Who says it was even POSSIBLE to patch X.org for this? What if the change was a larger change in the architecture of X.org, and nothing that can be easily reverted? Then you'd be able to EITHER support old fglrx, OR support new drivers for new hardware.

No, in this case, ATI was clearly at fault for dropping support for old hardware. Halting X.org development just for an old legacy driver made by a company which refuses to update that driver is certainly not a solution.

Here you rather see why closed source drivers are bad.

[Alien]

This WAS not architecture related. It was just stupid driver X -version check, which WAS FIXED later. Everything happened because xorg changed numbering scheme and nobody at ati fixed their abandoned drivers. Far easier change was to change version output instead of blaming ati. End users do not care about this, they want working system (my windows were working, blah blah).

[divVerent]

I still see no reason for Ubuntu or ANY OTHER sane distro to mess with stuff like version number reporting (which IS important for when X.org have to deal bug reports by Ubuntu users) because some third party driver for outdated hardware is broken.

And what if another driver (e.g. the nvidia driver) checks for the NEW version number scheme, and will fail with the old one? What now?

[Alien]

Ok, so it seems that the best thing is to leave users for their own fate. Sorry, I don't agree and don't want to argue more.

[divVerent]

?

A working driver for these cards is contained in X.org, which was why ATI discontinued support. Note that ATI discontinued support. Not Ubuntu.

So where is your problem? "BUT I WANT TO USE THAT DRIVER FROM ATI INSTEAD!!!!!!!!!eleven"?

BTW, same happened with nvidia. IIRC their legacy driver for Geforce 1 cards no longer compiles in current kernels (but I may be wrong with this). Are you expecting distributions to stay at old kernel versions so that legacy driver works?

Or, a case I have. I have an eeepc. To change the CPU speed, I need a special "eee" kernel module from http://code.google.com/p/eeepc-linux/ - but this module does not compile with Fedora's current kernel. Now is Fedora at fault for this? No, the authors of this module are, as well as the Linux Kernel developers for not keeping a stable API for drivers.

[Alien]

At that time there was no 3d acceleration in oss driver (r300 is experimental).

[divVerent]

At that time there was no 3d acceleration in oss driver (r300 is experimental).

You are saying that it's not ATI/Nvidia's fault but write "cause ATI stopped the drivers development for older cards." The reasoning behind them (=ATI) stopping fglrx-development was that there are open-source drivers available already (with 3D support), so why not let user use those free drivers?

Now what? The open source drivers being "experimental" is no reason against, because, what software you use is NOT experimental?