Re: Fwd: Regarding EGLStreams support in KWin
Drew DeVault <[email protected]>
| Newsgroups | gmane.comp.kde.devel.kwin |
|---|---|
| Message-ID | <[email protected]> |
On 2019-02-21 10:19 AM, David Edmundson wrote: > I do not approve of this being linked on public inboxes for Reddit, it > doesn't feel like the goal was actually about discussing technical > issues with the relevant people. I didn't post it here so it could be linked on Reddit, I Cc'd this list because there are more wlroots people there than on the kwin list. The kwin list has public archives, too - but I don't really appreciate that this was posted on Reddit, either. > It is an open standard that happens to only have one vendor. > > We currently have a libhybris backend, which shims propreitory blobs > directly. That was done by the then kwin maintainer to no noise at > all. Having a different stance here would really be an unfair double > standard. > > A lot of the arguments apply there and overall and despite it being a > whole separate backend that's harder to even build it has proven to be > much more of an asset than a burden. I didn't know there was a libhybris backend, but I'm sad to hear that there is. Much of my concern applies here, too. > If we are going to discuss technical merits, we should include them > for EGLStreams too, such as the deferred texture decompression that > don't exist in GBM. Aye, I'm not saying that GBM as a standard is better or worse than EGLStreams. This is why I said that one "out" for Nvidia could be to implement EGLStreams in another driver, like nouveau. Alternatively, showing a promising plan for unifying Linux graphics under some other model would be acceptable. I'm not just saying "rah rah Nvidia doom and gloom" - there are tangible steps they could take to make these problems better. > I'll continue to work on these things, my priority is my userbase, and > overall I think this patch is a good thing. > > I want to welcome the open source contribution that Nvidia is making. > Embracing open source outreach is the way to generate more. I think you're sacrificing large long-term wins to make small short-term wins. Maybe your userbase will double in size when you merge this patch. But what if the entire Linux desktop grows by 100x thanks to the proliferation of high-quality open source drivers because you chose to say "no" to a bully who wouldn't play ball with everyone else? I don't see what NVIDIA is doing as open-source outreach. I think they see contributing to KWin et al as a necessary evil they need in order to support their proprietary driver's goals.