00:13tpefreedom: What GPUs have the best nouveau support?
00:17HdkR: Kepler family
00:17HdkR: Which is Geforce 700
00:30imirkin: and 600
09:07HannaM: does a plan b exist, to react to the current state of freenode ? ( https://fuchsnet.ch/freenode-resign-letter.txt )
09:54karolherbst: HannaM: probably better discussed at a fdo level as this affects all of al
09:54karolherbst: *us
09:54RSpliet: HannaM: thanks for bringing that to our attention
09:56RSpliet: There are plenty of alternatives (libera, OFTC, Gnome IRC, Smurfnet, or fdo sets up their own infra), but right now I've only read one side of the story
10:13karolherbst: RSpliet: I am sure if you asked the fdo admins to host an IRC server, you get banned :D
10:14RSpliet: karolherbst: I'd highly recommend them to grow up if you get that sort of a knee-jerk reaction. I'm sure that's not true, but I also wouldn't insist that's necessarily the best solution.
10:20karolherbst: RSpliet: well, that's more work for them
10:20karolherbst: the only way to suggest that is if you are also willing to admin it yourself :p
13:49karolherbst: RSpliet: seems like the freenode staff stepped down: https://blog.bofh.it/debian/id_461
13:56karolherbst: RSpliet: btw, oftc seems to be the new hot IRC network
13:56karolherbst: and it seems like fdo generally prefers that one
13:56karolherbst: so I already claimed #nouveau there
13:58RSpliet: Cool. Apparently the former-freenode admins are forking to libera. I don't have strong preferences, but may be worth having a coordinated migration to one location if nouveau/fdo decides to side with the former admins on the matter.
13:59ajax: "new hot" is maybe a little misleading
13:59ajax: been around since '01, my goodness
14:01orbea: its hard to take "narcissistic Trumpian wannabe korean royalty bitcoins millionaire." seriously, what does that even mean? Would another network be even more trustworthy?
14:01karolherbst: ajax: :D
14:01karolherbst: RSpliet: well, llvm is also on oftc
14:02karolherbst: so that makes life easier for the AMD folks :D
14:02karolherbst: and CL
14:02karolherbst: orbea: it means what it means :D
14:02orbea: which is entirely unclear
14:02karolherbst: nope, it's entirely clear :D
14:03karolherbst: so somebody made some money with $hitcoins and bought the company managing the finances and now "bought" the infra and throws everybody out
14:03orbea: maybe you have some context that was not included in the link
14:03karolherbst: how is that not clear
14:03karolherbst: but yeah.. I think the details are a bit fuzzy atm
14:04orbea: then why not say that, what does trumpian and korean have to do with this?
14:04karolherbst: but what I've heard happening is, that somebod sold the domain+access rights who wasn't allowed to, to that person owning this freenode ltd. company managing only the finances originally
14:04orbea: but w/e, this is hardly important
14:04karolherbst: orbea: it's a comment describing what kind of person you deal with here
14:04karolherbst: but yeah..
14:05karolherbst: this "korean royalty" bit is a little strange though
14:05karolherbst: I think they could have left that out
14:05orbea: would make more sense without that
14:05karolherbst: yeah
14:05karolherbst: but maybe that person is korean royalty...
14:05orbea: heh
14:06RSpliet: https://gist.github.com/aaronmdjones/1a9a93ded5b7d162c3f58bdd66b8f491
14:06RSpliet: That's pretty clear
14:07karolherbst: anyway.. officially american, so I guess that was just a stupid remark
14:11karolherbst: RSpliet: yeah, that seems to explain a lot actually
14:14orbea: i feel like some pieces are missing still
14:15orbea: and Im not sure how much actually supports either side
14:15orbea: nor do I understand why the average user needs to be involved in their drama...
14:15RSpliet: Sure, we're hearing one side of the argument. I'm not involved enough to interpret and pick sides. I'll leave that to FDO ;-)
14:16karolherbst: orbea: yeah.. but that is a hostile take over, no matter how you look at it
14:16karolherbst: if you do a decision and your "staff" quits, you made a mistake
14:16karolherbst: it's really that simple
14:16RSpliet: orbea: different people have different opinions on the matter. I'm definitely a fan of the Free in "Free and Open Source Stuff". A hostile take-over is a red-flag for attempts to monetise.
14:17karolherbst: also, they advertised on the freenode hp for a different company also belonging to the owner of freenode ltd :)
14:17karolherbst: which has... nothing to do with freenode
14:17karolherbst: so that's a proof of corruption
14:18karolherbst: but yeah
14:18karolherbst: in the end I leave it to fdo to make a decision
14:19orbea: yea, i'd go with what fdo says
14:19karolherbst: just... preparing steps so random people won't claim channels we want to use :D
14:19orbea: just switching networks is a disruptive process, I hope if its done the reasons are actually fully thought through
14:20karolherbst: yeah.. I mean OFTC is actually fine, they have a track record of people open source friendly and some FOSS projects have their community there as well
14:20ericonr: (debian stuff is all on OFTC)
14:20orbea: good to hear oftc has a good track record
14:22RSpliet: And I guess it's more neutral than gnome's infra, and more proven than those ex-freenode's new network
14:22karolherbst: yeah
14:23karolherbst: I would even say that llvm being on OFTC makes a point of us going there as well, as I don't think we have related communities somewhere else
14:26ajax: gnome irc is the only other pre-existing irc network i might consider instead of oftc
14:29ericonr: karolherbst: oh didn't know about llvm
14:29karolherbst: ajax: right, but it's not really neutral, so I think generally people would prefer OFTC over gnome
14:40ajax:nods
16:22karolherbst: pmoreau: maybe at the end of the week I am done with the nouveau CI stuff :D
16:33Lyude: RSpliet: I actually saw this but I thought some other people on twitter said this had gotten released before it was ready
16:34Lyude: maybe the hostile takeover went through though
16:35Lyude: either way - not really the policy of fdo to dictate where projects have their IRC, but if that is the case at the very least it's probably worth us announcing it to x.org members in case any folks decide they want to move their projects to a different irc network
16:36karolherbst: Lyude: I kind of prefer some kind of vote and we all just use the same unless some project has strong opinions and want to be somewhere else
16:38Lyude: karolherbst: I mean, I don't really think we have any policy on this sort of thing
16:38karolherbst: that's why it's non binding
16:39karolherbst: I think it is best if all are on the same network, and we might want to discuss which to coose, but ultimately ebery project can decide on their won
16:39karolherbst: *own
16:39karolherbst: I still think it's good to coordinate
16:46RSpliet: yeah, I do think it's good to coordinate. Doesn't have to be top-down... OTOH it'd be good to get a recommendation from someone who actually wants to spend the time researching the situation at large, and judge things like how it would affect the long term reliability of freenode as a technical platform, how it affect privacy etc.
16:47RSpliet: FOSS-spirit is ultimately only one aspect of this, and although not negligible, the others matter too for an informed decision
16:52Lyude: hm - looks like the takeover did go through, found some tweets from freenode staff that seem to confirm it
17:18karolherbst: :/
17:22karolherbst: pmoreau, Lyude: :3 https://gitlab.freedesktop.org/drm/nouveau/-/merge_requests/2#note_924985
17:24RSpliet: karolherbst: merge requests are private?
17:25karolherbst: for now
17:25karolherbst: I don't want people to open MRs as long as we don't accept them :D
17:25karolherbst: but you should be able to see it, no?
17:25RSpliet: I was logged out :-D
17:25karolherbst: ahh
17:26karolherbst: yeah.. I just enabled it to work on the CI stuff
17:26karolherbst: RSpliet: the biggest blocker is, that forks need to be on the same gitlab instance, otherwise you have to wait like 20 minutes
17:27RSpliet: Oh, lol, this is infra testing. Cool cool
17:27karolherbst: yeah
17:28karolherbst: I mean.. long term we also want to be able to accept MRs besides emails
17:28karolherbst: and stuff
17:28karolherbst: well.. hw testing would also be nice, but.. that's a bit too much :D
21:32ccr: re: freenode stuff, https://www.scmp.com/magazines/style/tech-design/article/3124523/real-life-k-drama-how-did-andrew-lee-go-tech
22:28Lyude: jfyi - I'm already seeing a lot of people migrate to libera
22:28Lyude: also - probably worth mentioning in 10+ years I've never seen so many people migrate between IRC networks this quickly
22:30ccr: I personally just hope that this does not trigger a exodus of projects from IRC itself to some proprietary platform like Discord or .. urgh .. Slack :S
22:31RSpliet: Lyude: Yeah, the community don't take kindly to power grabs. I personally don't mind Libera or OFTC or whatever, it's all the same to my bouncer... but I think it'd be good to have a coordinated migration if we do migrate
22:32Lyude: ccr: yeah that's a big worry I've seen from a bunch of folks
22:32Lyude: RSpliet: yeah I added this to the agenda for tomorrow's board meeting
22:32RSpliet: Ah excellent, that's all I wanted to hear :-D
22:33Lyude: RSpliet: I honestly didn't expect to happen this fast lol
22:33ccr: would be good to know where this and other fdo channels move to :)
22:34RSpliet: Really? From my experience there was either going to be a really fast response, or it dies out with a lot of shrugging, "oh well"s and life goes on as before
22:35ccr: probably both of those responses around. it is unclear/unknowable what exactly this means for freenode's future, but I am kind of agreeing that it's probably nothing good.
22:36mwk: yeah, this is something unprecedented
22:36mwk: like half of the channels I was in are already on libera
22:36RSpliet: "Politics" aside, they just lost a metric shite-ton of maintenance- and development talent. That could threaten the continuity of the network...
22:36Lyude: RSpliet: yeah I guess you're right
22:37Lyude: (btw - jfyi I was going to bring it up during tomorrow's board meeting regardless :)
22:37mwk: they're also apparently bribing random people with giving them oper status so that they can "take revenge on their enemies", which is not exactly a sign of good management
22:37ccr: @__@
22:37RSpliet: Best case freenode stays operating as it is... which judging by the frequent netspliets isn't that great in the first place :-P
22:37Lyude: RSpliet: yeah they won't survive this, that's for sure. most networks need staff for a lot more then just moderation, but also just maintaining the ircd and other stuff
22:38Lyude: back when I ran ponychat we had our own fork of shadowircd (fork of atheme)
22:38Lyude: erm-charybdis
22:38Lyude: not atheme, atheme was the project
22:38RSpliet: charybdis is the thing that the freenode devs were working on to supersede current infra, and had to fork... right?
22:39Lyude: RSpliet: yep - they forked it and came up with ircd-seven or something like that
22:39ccr:remembers atheme.org and many of the projects under that flag
22:40RSpliet: "Solanum" is the latest fork
22:40Lyude: yeah, although I'll admit my memories of atheme and some of it's folks are definitely not positive :)
22:40RSpliet: or
22:40RSpliet: no
22:40RSpliet: wow
22:40RSpliet: so many forks
22:40RSpliet: it's like a restaurant
22:40Lyude: hah
22:42ccr: some open source projects seem to have a much larger chance of being forked. like DOOM engine forks.
22:42ccr: and on larger scale, Linux distros, I suppose
22:43RSpliet: in a dark past someone tried forking nouveau once
22:43imirkin: heh. pathscale.
22:43imirkin: are they still around?
22:44RSpliet: Wikipedia: "PathScale Inc. was a company"...
22:45RSpliet: Looks like they're a CDN now
22:46imirkin: from pathscale.com: "From 2009 until mid-2016 the new PathScale focused primarily on advanced compiler and toolchain solutions for scientific computing. ... Unfortunately, due to market conditions the viability of an independent compiler company did not last forever."
22:46RSpliet: Codeplay disagrees
22:47imirkin: just quoting what they have on their page.
22:47RSpliet: I know :-D
22:47RSpliet: Just trolling
22:47airlied: competence matters :-P
22:48imirkin: they definitely had some competent people working for them (some #nouveau regulars)
22:48imirkin: but i think selling a compiler for nvidia chips is a losing proposition
22:50Plagman: i remember an interesting irc chat circa 2010 (but don't remember with who) where the patchscale person was explaining to me that their stack would have leapfrogged ours in a year's time
22:50Plagman: (ours as in, nvidia's)
22:50RSpliet: I'm guessing the nickname was codestr0m
22:50Plagman: it was a fine goal
22:50Plagman: oh yeah possibly
22:53RSpliet: yeah, setting a goal like that is... brave
22:53imirkin: well, at the time, nvidia's gpgpu capability was fairly new
22:53airlied: codestr0m did come across as arrogant over competent :-P
22:54RSpliet: I mean, don't get me wrong, I've never had a look at NVIDIA's compiler. I wouldn't be surprised if it had, eh... 10 years of legacy making some optimisations hard to do.
22:54imirkin: and if you went with the assumption of "they will never touch it again", it might have been an achievable goal
22:54Plagman: yeah that assumption seems like the most flawed part
22:55imirkin: i've only looked at the code generated by nvidia, and in all the places it matters, it's _very_ good
22:55imirkin: it also has some weird stuff where it emits obviously suboptimal code
22:55imirkin: like some opt passes are misordered
22:55imirkin: but i don't think that stuff matters in practice
22:56imirkin: it's just a slight eyesore when you're looking at it
22:56RSpliet: Today I totally believe that. Not sure how well it was in 2010?
22:56RSpliet: Or when was that pscnv stint again. I think around that time
22:57RSpliet: Probably not too bad btw, there's quite a lot you can do with a small army of compiler engineers :-P
22:58Plagman: well, even if the gpgpu part was nascent, you could probably look at the graphics side and assume it was going to get to the same point soon enough?
22:58imirkin: Plagman: no, i mean the whole arch was nascent
22:58imirkin: DX10 vs DX9
22:58Plagman: oh you mean tesla was new-ish, etc?
22:58Plagman: yeah
22:59imirkin: anything you did for the DX9 gpu's would not apply
22:59imirkin: G80 came out in like late 2006 i think
22:59Plagman: yeah
23:01RSpliet: Plagman: well ultimately it depends on customer take-up. But you're right, either it worked out as a massive new market and NVIDIA would vastly outspend any other player, or it turns out there wasn't a killer application and would die a silent death - with no market for pathscale to conquer
23:51Plagman: yeah seems like a hard proposition