[ale] Is ATI Linux compatibility good?

[Richard Faulkner]

This is where I think I'm bumping-up against NVIDIA, a *clash* between
the driver and Xorg.  My next step in working this issue is swapping
GPU's and see what happens.  But that is another thread isn't it...



On Mon, 2010-03-29 at 14:04 -0400, Michael B. Trausch wrote:
> On 03/29/2010 08:33 AM, Charles Shapiro wrote:
> > Uh, I don't think they're dropping it. They're just supporting
> > **only** their closed-source driver, with no published stuff to allow
> > other people to write drivers.
> 
> Correct.
> 
> It seems that older versions of the NVIDIA drivers, IME, work just fine 
> (well, that is, unless you get a card that the older drivers don't 
> support).  The problem lately I think is that there have been a lot of 
> changes in the Xorg system and the way that it handles various things, 
> and of course the free software drivers that are open for modification 
> can keep up with these sorts of changes more quickly.  It also helps 
> when the kernel-half of the driver can be updated to match API changes 
> in the kernel (nobody in the kernel world cares about ABI compatibility 
> for drivers, since all the in-tree drivers that are needed are built 
> against the kernel when it is rebuilt anyway).
> 
> This is, I think, a key part of the argument for free software that many 
> people seem to ignore.  For example, the Free Software Foundation would 
> have everyone decide to use free software for the sake of using software 
> that isn't proprietary, but this is not very practical for most people. 
>   But there are some practical arguments that make sense for those who 
> aren't completely fanatical in nature.
> 
> As an example, due to the open development process that the Linux kernel 
> uses, and the fact that majority of the drivers for the kernel are 
> in-tree drivers, they feel (and rightly so, I think) that they can 
> change the API and the ABI any time they please.  This presents a bit of 
> a problem for binary-only drivers (or even out-of-tree drivers) since 
> they are not maintained in lock-step with the current kernel.  It's an 
> even larger problem for binary-only drivers in that there can never be 
> backports without action from the sole producers of the binary-only drivers.
> 
> I've never really been able to use the open-source "nv" driver with any 
> of the chipsets that I have owned, and the nvidia driver works well on 
> most (though abysmally on two of the ones that I have, and one of those 
> is, I think, a hardware issue and not a driver one).  I have to say that 
> I am glad that there is a free software driver for the ATI chipsets that 
> works as well as it seems to work.  I just hope that the hardware + 
> driver combination in my laptop (and its exceptional performance and 
> reliability) isn't a fluke of some sort.
> 
> 	--- Mike
>