Aalto Talk with Linus Torvalds (youtu.be) You have to love Linus for not being afraid to speak his mind. |
Aalto Talk with Linus Torvalds (youtu.be) You have to love Linus for not being afraid to speak his mind. |
GNU/Linux has no drivers: "Fuck NVIDIA!"
What is the problem?
The last point is somewhat ironic: if Linus had been reasonable and talked about Nvidia at length, Nvidia would probably not have noticed the talk; since Linus decided to be curt, the talk got onto HN and people from Nvidia have no doubt noticed it.
Besides, whatever you think about it, Linus's approach was indubitably effective. We are, after all, talking about the issue right now!
(ok, this is slightly unfair to intel's video drivers, they work fine as long as you don't need 3D acceleration)
But Linus' message was a bit disheartening. Most of our crew would really like to open source everything we write. But then there's the fact that our code would reveal the secrets of our hardware. And the fact that Nvidia is a hardware company with a strong Windows background, so open source software is quite a strange thing in comparison. It kinda feels like being between a rock and a hard place.
When it comes to mobile software, Nvidia has been doing more open source work and upstreaming quite a lot of our changes. A lot of work has been done in the kernel internals and arm stuff. Not a lot of customer facing stuff but very valuable work if you're trying to build ARM-based system on chips.
To whom? Certainly not your competitors.
Also what's wrong with releasing specs under NDA?
And of course, device driver would tell people all the secrets that NDA forbids you to disclose.
Nevermind, we were both just doing it wrong. This should work:
You can make the argument that nVidia Linux drivers are much better, but I don't see where you're getting the idea that it's hard to find AMD drivers.
The closed AMD driver is pathetically bad. Neither multi-monitor mode works for me in Gnome3, due to from what I gather is AMD not keeping up well with architecture changes. AMD is also dropping support for their oldest architecture in their next driver release, leaving a ton of cards - some still being sold today - with nothing but the open source driver anyway.
My next video card is going to be an Nvidia for sure. Whatever Nvidia's faults, their proprietary linux support cannot possibly be as poor as AMD's.
Many Linux kernel developers have offered to sign NDAs with Nvidia and many other companies.
And of course, device driver would tell people all the secrets that NDA forbids you to disclose.
How? It's easy to obfuscate certain details of driver code. Think of magic constants, for example.
And yes, I did. I actually spent half a dozen hours fiddling with xorg.conf etc trying to get multiple monitors to work with fglrx [AMD proprietary driver]. I should have known better. Especially since my not-that-old card is being dropped in the next fglrx anyway.
So, when people are reading the comments of this submission in the future, please keep this in mind as a historical note. (This, humorously, was actually the kind of situation that caused the complaint[1] that itself turned into a massive hullabaloo recently regarding what can be discussed on HN and what the policies regarding hell-banning are; to view the reference you will need showdead.)
At first glance it appears to be a flamebait title, and hence a mod who sees it may feel it's a no-brainer - correct it and move on.
However, there's a nuance to it - that is exactly how Linus Torvalds expresses himself, and the original title ("Linus to Nvidia: Fuck You!", or something close to that) captured his sentiment accurately, so maybe it's not flamebait after all.
Or maybe it is flamebait even despite that, since Linus's SOP is to sometimes start flamewars to make a point, break through the red tape, or otherwise just make a command decision and move on.
Clearly plenty of room for moderation error, a nd that's just one submission. What's a mod to do?
So on the one hand, there has been a spate of godawfully-titled submissions in recent months:
1. "X things you should ... whatever" type titles (clearly banned in the HN posting guidelines)
2. Too short and uninformative (like a word or three).
3. Sensationalism, flamebait, miscategorized comparison results, etc.
4. more I'm sure...
But on the other hand, the mod system has problems as well:
1. Nobody even knows what the mod system is
2. Nobody knows who the mods are.
3. There's no way to give feedback on moderations, for the ones that were incorrectly modded.
4. Too many false positives (posts that shouldn't be modded but are, resulting comments like saurik's above, and entire threads complaining this problem).
5. Too many false negatives that slip through anyway.
And of course, not part of the mod system, but too many submitters just don't know how to descriptively, accurately, concretely title submissions anyway, increasing the volume a seemingly too-small group of mods has to deal with.
HN isn't the first social media site to have problems like this, but most others have a full-time dev team working on solving them, and they evolve certain solutions like Slashdot's meta-moderation or Reddit's user-run/modded subreddits.
So I don't think the mod system in its current form can scale with those problems, but on a more meta level I'm not sure that PG can scale as the developer of the mod system, given that YC takes 110% of his time.
Just trying to identify the problem before attempting to solve it, any thoughts?
Personally I'd love to see another smaller, more refocused community spin off where better, more open and reasoned tech/startup discussions can take place. Keep it invite-only perhaps.
> 1. Nobody even knows what the mod system is
That is not a problem. This is excellent feature of a moderation system. Meta is death - really. I'm never going to make another meta post after this one. (Unless it's t help a new user.)
> 2. Nobody knows who the mods are.
Again, that's not a problem. It avoids turning moderating into character battles. A mod who makes a mistake anonymously has no investment of face-saving; they can easily undo the error.
> 3. There's no way to give feedback on moderations, for the ones that were incorrectly modded.
There is an email address clearly listed in the guidelines.
I really don't think the problems are as big as people are making out. Sure, some things are frustrating. Taking this video and post as an example: Don't link to the small part of the video where Linus tells nvidia to go fuck themselves, link to the entire video (and to the start of that video) and then give it a better title. Trying to attach blame to mods because someone made a weird sub-optimal choice when submitting a link is un-good.
1. If a post is renamed then the original title(s) should be available to view somehow. Maybe via a moderation summary link.
2. The downvoting mechanism should require the downvoter to post their reasons.
Edit: Seriously though, is the new rule that we aren't allowed to submit anything other than the page's title?
"People who get offended should be offended!" - Linus
"I like offending people because I think people who get offended should be offended."
Of course, in this case, it is a descriptive title for the talk but completely takes away the specific thing the submission is about. This is slightly less bad but still bad.
Her question actually starts a minute earlier than the link:
http://www.youtube.com/watch?v=MShbP3OpASA&t=48m14s
PS - Bravo Linus. This is issue is a real PITA, and a bit incongruent considering the historically awesome driver support Nvidia has provided for Linux.
I've found it while watching the video after the "fuck you nVidia" discussion: http://news.ycombinator.com/item?id=4121698
Edit: Actual quote: "I like offending people, because I think people who get offended should be offended" (without the "the" before "people").
Quote (~min 11:30): "I have never in my love done any web programming because I'm not interested, I think that kind of stuff... there's MIS people to do that for you, right? I'm interested in programming"
[edited some spelling bugs]
Reasons include: Keeping competitors away from what you think are valuable secrets and maintaining an advantage. Keeping people away from features that, if misused, could result in chip damage. Keeping security holes secret (e.g., badly designed DMA hardware that could be exploited, if the flaws were known). Limiting access to known buggy features, or unfinished features that either don't work or that could leak damaging hints about strategic direction. You have purchased or licensed 3rd party technology that you contractually cannot divulge details of. For interoperability with other products you have embedded knowledge of them in the product, under NDA.
More: It's expensive to document chips to the point that outside development can be done. Perhaps the documentation doesn't exist, at all, and would have to be reverse-engineered out of the chip design (yes, this happens). It's expensive to write drivers for multiple platforms, or even to get software into a state where it can be consumed by an outside party (just dumping a tree onto GitHub is /not/ a release). You feel that "forking" would result in a loss of control of your own product (and would dramatically increase the cost of future releases, lest you break things). You regularly rev chips and cover the changes transparently in the software layer, and this would /not/ be transparent if you released product details (thus increasing the cost of revisions).
More (the slimey side): You have misappropriated technology and divulging it would be harmful to you. There are design errors or bugs verging on malfeasance that could expose you to litigation. You have lied about the product's capabilities and a release would reveal this (whereupon, litigation).
Or, it's a pain in the ass, the market is significantly less than 1 percent of your total, and you have a horizontal skyscraper of engineers already behind schedule. "Good faith and being nice" doesn't pay the bills.
[I have also heard, from other parts of the industry, that the company in question is hard to deal with].
Some companies have a financial incentive to effective tell the open source movement to f-- off. Sure, when they take that, they may not be being evil on a grand scale. But hey, if they are telling linux to f-- off, it seems appropriate Linus return the favor. He's just making things clear.
Is there anything wrong that?
"You can disagree with me as much as you want, but during this talk, by definition, anybody who disagrees is stupid and ugly, so keep that in mind."
It's still in its infancy, and faces a lot of challenges that open source software doesn't, but it's come a long way in a few years.
1. One of the highly distinctive characteristics of being a Free Software project leader is having the freedom to speak your mind. What Linus does (hacking the Linux kernel) and who pays him to do it (presently the Linux Foundation) are pretty loosely linked. The primary objective of LF is to fund Linux development, and Linus is pretty much the guy to get that done. If LF didn't pay him for it, someone else would. He can state his opionions on relevant technical matters with few if any fears of repercussions. I'm looking forward to next week's press releases from Nvidia.
2. Linus addresses what Nvidia are doing wrong at a few points, both directly and indirectly.
Around 15 minutes in he talks about what Free Software provides in the way of developer freedoms: you can focus on what you are interested in and what you are good at. In Linus's case, issues such as maintaining Linux-related websites, init, QA, and Linux distributions is stuff he fundamentally doesn't care about (while other bits such as, eventually, creating a useful revision control system he does). Free Software lets you focus on your own core competencies.
He also makes the point, around 35 minutes, that it's very important that people need to know how he feels about things. Including how he feels about support received from hardware vendors.
More specifically, for hardware manufacturers, playing nice and closely with the kernel development community leads to both better product performance and customer relations. The woman asking the Nvidia question clearly wasn't happy with her Nvidia experience. I've learned in assessing hardware compatibility to treat any Nvidia componentry as at best a red flag if not a show-stopper. I'll actively go out of my way to avoid their products (Intel have gone out of their way to ensure compatibility and open specs, my most recent purchases centered on Intel chipsets, in particular for graphics). Playing well with devs also means that issues are addressed in a timely manner, compromises can be reached, and in general communications are open and positive. I don't know the full backstory on the Nvidia front (though searching the LKML mailing list should turn up some bits).
3. ... and yes, the HN moderators fubared this one.
You should have titled it "Linus Torvalds Angry at Nvidia, Flips the Finger at Speach" or something.
That would have been more descriptive, just a heads up for next time! ;)
Apparently, a clueless moderator decided to change it.
Edit: I expected downvotes and I'll gladly take them if someone can explain to me how Nvidia is being unreasonable.
It's not really incongruent. Nvidia has provided driver support for Linux, but that does no good for, say, FreeBSD. This driver support is in lieu of providing specs. They're keeping a lid on things and maintaining control. There are very good hackers who would make open source drivers if they had specs. Arguably less buggy drivers. Since they make sure they have control it's not surprising that they either 1) think this isn't worth their trouble, or 2) have actively decided they don't want this to happen, for whatever business reason.
It's worth noting that just last night they released a stable driver that (finally) supports RandR 1.2. From what I understand this was also a significant change and it may bode well for future standardization improvements like KMS (moves multi-monitor management away from the proprietary TwinView driver components into standard xrandr, for instance, so that is complexity the KMS version will not have to handle).
Fingers crossed that they release this imaginary driver soon. :)
The XRandR support in the driver is not yet stable [1]. The 302.xx line is currently in beta, but will support both XRandR 1.2 & 1.3 at the same time. That part also works flawlessly, however, your mileage may vary in regards to suspend/hibernate, since both I and multiple others on the nvidia forums experience serious problems with resume from either, i.e. the computer hard locks (and doesn't even respond to ssh etc.), which requires you to powercycle it. So while the XRandR is a huge step forward, it is not without problems yet.
[1]: ftp://download.nvidia.com/XFree86/Linux-x86_64/latest.txt shows you the latest stable driver, which is 295.59
I guess he's talking about pushing out some new driver support.
Using a fragment identifier to demarc the time offset works better for me:
There was a stink a while back where Adaptec had been providing drivers for FreeBSD, but then stopped. People were stuck either not upgrading their boxes (bad!) or moving off their RAID setup (painful, expensive!). It's ugly. With specs, there is not this issue.
The (pervasive) idea that vendors providing drivers for your system of choice, for the time being is being open source friendly is something the vendor promotes, but it's not actually friendly.
support
ArchLinux has already sent out packages for this release in its stable update channel: https://projects.archlinux.org/svntogit/packages.git/commit/...
(As an aside, it still hardlocks my computer...but according to their forums, they "are working on a fix" for the soon-to-come 304.*)
Metafilter solves this with http://metatalk.metafilter.com. Perhaps HN needs a dedicated place for meta discussion as well.
www.shirky.com/writings/group_enemy.html
Well I probably count as the masses. But one model might be a sort of web of trust. You start off with a few people who are ultimately trusted. And then as your web grows you can track who let who into the site. People who consistently let in poor members get their invite privileges revoked or are kicked. Have a reputation threshold before you can invite people.
Second, if "poor members" get invited in large numbers, then the system you propose will fail, because they get to dictate what is good or not (by upvoting it), which will in term make them appear good (because they like/submit the 'good stuff').
Moderation has been a feature of complaint across every web forum (and many non-web forums) I've ever been a part of.
See, for example, the comments here (and even there) about StackExchange; 4chan; suicidegirls (NSFW) "spring cleaning"; the HUGE amount of meta / drama on Wikipedia (ANI alone is gigabytes of guff stretching over years about moderating that community. The holy books of millennia old religions are smaller than ANI. International trade argrements are smaller than ANI.)
But this meta bike-shedding has also been a feature of older systems. It's frequently created flame wars on Usenet - leading to various trolling groups sporging Usenet feeds. It's a feature of mailing lists.
I don't know what the average[1] age of HN is, but here's a result from a Usenet search for results before 1990 - before a lot of HN would have been born.
(https://groups.google.com/groups/search?safe=off&q=moder...)
(Also, Google, please give me a shorter URL option to cut n paste. Don't make me have to learn what your URLs are doing; don't expect me to use a nasty URL shortener (which are blocked on many boards)).
Here's a mildly interesting message discussing some of the problems of moderating Usenet:
(https://groups.google.com/group/mod.comp-soc/msg/ee189feb225...)
Case in point: my parent comment is currently downmodded. Thanks to that downmodder for marking against my personal experience. But I have no way to find out why I've been affected, only that some random person somewhere in the world doesn't like what I said for some unknown reason, and my words are literally diminished in the eyes of others because I have (at least) one single dissenter.
Since I started here, I've noticed that people on HN have been complaining about the moderation more consistently than on any other web community I've been involved with, with the exception of Wikipedia (as you point out, but it's not the same flavour of moderation I mean)
I wonder about that. Is it b/c PG is the only the HN source committer, but without enough time to evolve the site fast enough to keep up?
Or is it something else?
Nvidia is being unreasonable in the same general way as Apple is unreasonable with its walled garden and other companies are unreasonable with pollution and poor working conditions. Just because you can get away with something and it makes economic sense does not justify it and does not make it reasonable!
I really don't think we should accept every company's actions just because they make business sense. We, as consumers (and as developers) should promote altruism and good behavior. Being a corporation is no more a license to be a jerk than being a normal person--it's legal in both cases, and we should not condone it in either.
Good communities have their own standards above and beyond the law. In the academic community, plagiarism is not tolerated even if it does not infringe on copyright. The open source community should similarly not tolerate companies and individuals who use the open source software and then refuse to cooperate. And this is all that's needed: cooperation. I don't think anybody even expects Nvidia to write open source drivers; all they want is enough information to not waste time reverse engineering each chip.
In short, Nvidia is unreasonable because it is acting like a jerk. Being legal and making business sense should not justify being a jerk, and we should disparage and avoid companies that act like jerks. So yeah, fuck Nvidia.
If you depend on free software to sell your hardware, it is probably in your interest to see to it that the people that write the software that is required to sell your hardware don't hate you.
Not to disagree with your point (which is right), but It becomes in NVidia's financial interests when they have negative press coverage because someone like Linus says bad things about them.
I hope you understand what I am saying. Linus isn't just talking in an abstract-intellectual environment, his voice is strongly publicised and is here he is using it as an activist, to force NVidia into action. Its all part of the game...
It's not like more than 0.01% potential buyers will read this Linus quote and choose a competing product. Especially if NVidia's product is the best for their use case in the first place.
Why should Nvidia devote resources, potentially a significant amount, to make Optimus work on Linux? I do not buy the argument that they should do so out of gratefulness that the Linux based Android OS let's them sell a large number of unrelated chips.
I'm open to being convinced otherwise, but citing a lack of "gratefulness" is not compelling.
Similarly, you say I disregard a list of sites when I mention three of the four in my response about my personal experience being part of web communities (and seriously, 'as good as 4chan' is hardly an argument for content quality, particularly on a site where the users pride themselves on intellect and quality).
Your issues with not knowing who downvoted you for what reasons seem nearly identical to Slashdot, then (modulo I guess being marked "troll" when you were really downvoted for something more readily corrected). You had previously mad it clear that you cared deeply about the moderators being anonymous, and boy if Slashdot didn't maintain that.
Regardless: you have still, to this point, not stated even a single other web community you were a part of, only that they existed. The conversation becomes interesting an useful when you can point at actual examples (as it has now for Slashdot).
To look at this from another analogy, imagine it you claimed that you had never seen a bank that charged fees. Someone else responds saying they had never seen a bank that didn't charge fees and then lists ten banks that charge fees. You then respond with simply "I don't agree those are banks" but don't provide a list of banks that don't charge fees. I fail to see how you can claim you are then being helpful in that conversation.
Don't worry about the down votes. They are frustrating, but anyone on HN knows that you occasionally get a random down vote. By themselves they're usually meaningless. Yes, people should really either engage in the debate and not down vote, or not engage, and down vote, and explain if needed. (Personally, reasons for down voting should be restricted to "this comment does not belong here; it adds nothing to the discussion".)
I think there's general agreement about voting on HN - up vote things that are well written, even if you disagree. Don't down vote just because you disagree. Down vote things that add nothing to the discussion. Flag spam or other blatantly harmful stuff.
But all of that is separate from the issue of trusted people within HN locking threads, deleting threads, deleting posts, changing titles, banning users, etc.
> No insight and no accountability = bad moderation practices.
I honestly truly disagree with this. Strongly.
It's really simple to say "Here are a list of rules. Don't break them or we mod your posts". But when Bob starts skirting the rules there can be huge flamewars. Bob says he clearly wasn't breaking any rules, mods say he was, other people jump in saying that he wasn't and the mod is dumb and has made a bunch of similar stupid decisions, then other people jump in and say that maybe Bob was a bit close but he does so much good that we can make an exception for him, then other people leap in and say it's a stupid rule anyway. Meanwhile, this thread is getting so much attention that the real stuff of the forum is being ignored. New users are not interested in that stuff. The thread creates an unpleasant atmosphere which is noticed by other people. Now whenever Bob replies to someone who's a mod there's suspicion that Bob is just flame-baiting.
Anyway, I'm doing what I hate, so I apologies to everyone on HN. I'm setting my noprocrast settings for 48 hours. This is my lost post on this thread!
Now try to find a laptop with a discrete Nvidia GPU that doesn't have this Optimus junk and you'll see the problem.
That may be true for specific features like Optimus--I can't say. But if you're talking about writing a full 3D driver stack for a modern GPU, you're almost certainly underestimating the effort involved by several orders of magnitude. NVIDIA has hundreds of full-time software engineers working on the various parts of the GPU driver stack. Now imagine doing it without the immense institutional knowledge.
Less fun edition: I'm sure it is. As it is with Linux based 3D render farms. Do both of those represent a large portion of Nvidia's income or an insignificant one?
I presume it's the latter: there are billions of desktops/laptops but only several thousands of scientific computing / 3D rendering installations. The fact that Nvidia doesn't seem to go out of its way to help people using Linux seems to corroborate that.
And clearly if I had to have worked to some niche fields to know about that use, it's not that widespread.
I seriously doubt Nvidia's bottom line is affected by such professional uses of GPUs.
>And how about a large portion of android phones?
What about them? You think companies like Samsung or Motorola are gonna switch just because Linus said something negative? Or that Android customers are gonna demand a different GPU?
Not to mention that Nvidia is, if not "the only game in town", one of the "counted on the fingers of one hand with some digits cut-off in a tragic saw accident" games in town.
