Mesa Will Not Be Dropping Its Older GPU Drivers

[phoronix]

Phoronix: Mesa Will Not Be Dropping Its Older GPU Drivers

While last week was the ambitious proposal to drop older GPU drivers from Mesa including the likes of i915 and R300g -- and possibly branching them off to their own Git branch for continued maintenance by interested individuals -- that proposal isn't going to fly...

http://www.phoronix.com/scan.php?page=news_item&px=Mesa-Not-Dropping-Old-Drivers

[stan]

That's great! Dropping drivers or splitting off the old ones is most of the time an anti-user, selfish move on behalf of developers. Proponents first they say it's better for older drivers to be split off so that they don't get broken by updates (aww.... how touching, they really care... NOT!). Then when the older drivers bit rot and miss out on security-related code restructuring, they use that as an excuse to call for the deletion of the older drivers altogether because they have security vulnerabilities. The writing and agenda is on the wall, and I, as a user, am glad it was thwarted.

One of the coolest things about open source software is that planned obsolescence (Apple forcing you to buy new hardware because they don't port newer versions of their OS to the older hardware) doesn't fly. It's a point of pride to see how well Linux works on 20-year old computers, and shows the dedication of the community! There's nothing cooler as a developer than seeing your hardware-agnostic code optimizations in subsystems like the scheduler, file system or graphics framework/GTK improve performance and give new life to older computers. Not to mention being good for the environment, etc.

[Adarion]

I feel kinda relieved. Okay, I do not have that much old HW that actually uses something in mesa (SiS, VIA, AMD Geode,... don't really have something working in mesa and are bound to be 2D only) but e.g. some soldered R100 on an embedded machine. I'm happy it is a R100 so you can actually use that machine and have some (limited) fun.
I also see that a lot of devs like to play with recent HW and gimmicks, but from a user perspective keeping the older drivers and giving them some occasional love would be neat. Putting them in an extra block might sound okay but possibly might also lead to them falling off the world silently (e.g. does no longer compile with newer main code of mesa, shows bugs like freezes, blank screens, corruption,...).
If the older drivers actually were feature complete... that would be a little different, but they aren't.

[FishPls]

That's great! Dropping drivers or splitting off the old ones is most of the time an anti-user, selfish move on behalf of developers. Proponents first they say it's better for older drivers to be split off so that they don't get broken by updates (aww.... how touching, they really care... NOT!). Then when the older drivers bit rot and miss out on security-related code restructuring, they use that as an excuse to call for the deletion of the older drivers altogether because they have security vulnerabilities. The writing and agenda is on the wall, and I, as a user, am glad it was thwarted.

One of the coolest things about open source software is that planned obsolescence (Apple forcing you to buy new hardware because they don't port newer versions of their OS to the older hardware) doesn't fly. It's a point of pride to see how well Linux works on 20-year old computers, and shows the dedication of the community! There's nothing cooler as a developer than seeing your hardware-agnostic code optimizations improve performance and give new life to older computers. Not to mention being good for the environment, etc.

I'm quite frankly completely fine with dropping support for older hardware if having them in-tree hinders development speed with modern hardware.

What value does holding completely broken drivers in-tree provide to you as the end user? You try to install them and find out they don't work. Are you somehow happy because of this? Nobody will step up and fix them, the hardware is ancient and no sane person will want to be using them.

[starshipeleven]

That's great! Dropping drivers or splitting off the old ones is most of the time an anti-user, selfish move on behalf of developers. Proponents first they say it's better for older drivers to be split off so that they don't get broken by updates (aww.... how touching, they really care... NOT!).

Don't worry, apart from r300 those drivers are already rotten and limping around, and the situation won't improve in any case.

Then when the older drivers bit rot and miss out on security-related code restructuring, they use that as an excuse to call for the deletion of the older drivers altogether because they have security vulnerabilities.

Even if they did this dick move (it isn't the usual way of acting), you can't "delete" something that is opensource. Distros have their clones and can keep packaging that as long as they want.

One of the coolest things about open source software is that planned obsolescence (Apple forcing you to buy new hardware because they don't port newer versions of their OS to the older hardware) doesn't fly. It's a point of pride to see how well Linux works on 20-year old computers, and shows the dedication of the community! There's nothing cooler as a developer than seeing your hardware-agnostic code optimizations improve performance and give new life to older computers. Not to mention being good for the environment, etc.

People need to fucking start to learn the difference between obsolescence and planned obsolescence. We are talking of ancient shitty GPUs whose 3D component is woefully inadequate even if it was 100% working (again excluding r300, although it's kind of meh too)

That's obsolete stuff, not planned obsolescence, just fucking obsolete.

Also please start learning how drivers are dealt with in Linux. Mesa is for 3D acceleration, Xorg has 2D acceleration drivers, Kernel has actual hardware support, crappy video and normal text mode.

Dropping Mesa drivers for them would mean very little for their users as in 100% of the cases they are using 2D acceleration and the driver for that is in Xorg, not in Mesa.

[Tuxee]

That's great! Dropping drivers or splitting off the old ones is most of the time an anti-user, selfish move on behalf of developers. Proponents first they say it's better for older drivers to be split off so that they don't get broken by updates (aww.... how touching, they really care... NOT!). Then when the older drivers bit rot...

They are bit rotting, even if they stay in mainline Mesa.

and miss out on security-related code restructuring, they use that as an excuse to call for the deletion of the older drivers altogether because they have security vulnerabilities. The writing and agenda is on the wall, and I, as a user, am glad it was thwarted.

One of the coolest things about open source software is that planned obsolescence (Apple forcing you to buy new hardware because they don't port newer versions of their OS to the older hardware) doesn't fly. It's a point of pride to see how well Linux works on 20-year old computers

You are aware, that we are talking about graphics drivers (and therefore desktop use), right? Because on 20-year old computers (1997 - that's when I had a Pentium 100 and considered upgrading to a AMD-K6 running at 233MHz) contemporary distributions which use recent graphics drivers won't work at all.

...and shows the dedication of the community! There's nothing cooler as a developer than seeing your hardware-agnostic code optimizations improve performance and give new life to older computers. Not to mention being good for the environment, etc.

Ah, the environment. Well, using an abacus (prefereably all wood recycled from even older abacuses) is even better for the environment.

[brent]

Splitting up old drivers into a legacy branch isn't anti-user at all, quite the opposite. I don't understand how some people arrive at this conclusion. I can only assume that most of these people are not developers.

Right now, most drivers for old GPUs (and that starts with everything older than 5-6 years) are bitrotting as the Mesa code around them changes. Since there are relatively few users still relying on these drivers and even fewer developers that still care about them, it's unlikely to change. Splitting up these drivers into a legacy branch would stop all the Mesa infrastructure changes, so there would be no further regressions. Some dedicated developers that care about the old hardware could then fix up the drivers and keep them working for some time (it's unlikely to happen, but much more likely to happen compared to working against a moving target on the master branch). On the other hand, the master branch could progress faster because there's less baggage that needs to be kept around.

[Luke_Wolf]

That's great! Dropping drivers or splitting off the old ones is most of the time an anti-user, selfish move on behalf of developers. Proponents first they say it's better for older drivers to be split off so that they don't get broken by updates (aww.... how touching, they really care... NOT!). Then when the older drivers bit rot and miss out on security-related code restructuring, they use that as an excuse to call for the deletion of the older drivers altogether because they have security vulnerabilities. The writing and agenda is on the wall, and I, as a user, am glad it was thwarted.

One of the coolest things about open source software is that planned obsolescence (Apple forcing you to buy new hardware because they don't port newer versions of their OS to the older hardware) doesn't fly. It's a point of pride to see how well Linux works on 20-year old computers, and shows the dedication of the community! There's nothing cooler as a developer than seeing your hardware-agnostic code optimizations improve performance and give new life to older computers. Not to mention being good for the environment, etc.

Quite frankly, you're an asshole.

There's no such thing as planned obsolescence in the tech industry outside of Apple, and especially not in Open Source. Yes companies drop hardware support in drivers over time, but it is absolutely not out of some malicious intent to force you to pay them more money. The reason they drop hardware support is because it isn't worth their time to work on it because it's not a current or future product (and quite often it's a completely obsolete crap product that they're dropping support for) , and continuing to support it is getting in the way and is only of benefit to a rapidly diminishing population anyway.

In Open Source however things are a bit different, because if you and I do mean you, want support for your 20 year old obsolete crap, then you are free to go in and maintain and update the code yourself. If that support gets dropped then you have no one to blame but yourself. And the only answer to "Oh but I don't know C" is you go learn it or... pay someone to support it for you. You are not entitled to anyone's support or anyone's labour. If you want something then it is on you to either do it yourself or pay for someone to do it. No excuses. There's plenty of free resources online. If you can't do that... then tough luck

People's asinine assumption that just because their hardware still somehow manages to run means that they're entitled to free software updates forever gets in my way and everyone else's who has to deal with the nonsense of people like you. If you want to run that hardware until it no longer works... fine... do that, and if you want to put in the effort to keep it working... then great... otherwise you get updates until the people who are actually doing the work decide you don't anymore, and you have no right to anything beyond that point. Got it?

[schmidtbag]

For the sake of anyone who would use hardware classified as "legacy", I think it's a little unfair that Arceri has given up so easily (it affects current-gen users too). How many times has he actually asked the question with a strict "yes or no" poll? If other devs don't give a straight-cut answer (even if it's slightly in favor) then that answer should be ignored. Arceri was being democratic about this and the doubters weren't participating; therefore, their answers don't matter. If the results were mostly positive, then clearly the majority has spoken - go forth with the change.

It's not like the drivers have to cease to exist. It's not like this is affecting a large percentage of users. If holding onto these older drivers adds bloat and causes regressions or developmental difficulties (for either legacy or current models) then that alone is enough of a reason to separate them, regardless of skepticism.

[cj.wijtmans]

I will say it again monolithic tree is bad. But it does simplify some things. Imagine this wont ever be a probpem if all the drivers had their own tree on top of base mesa. Also the ceiling would be lower to push some patches for a driver. Just my thoughts.