On user program breakage from Linux kernel changes | Hacker News Reader

On user program breakage from Linux kernel changes

article.gmane.org

103 pointsintel_44 comments

Puhh... Linus is known to be quite reckless and he may be right in essence, but the language... he might calm down a little and learn one or two things about social interaction.. there are a ot of ways to say what you want without plainly insulting someone over and over again.

It seems he calmed down a little afterwards and answered without all that rage and curse words, clearly he's upset but i don't imagine him going postal anymore:

http://article.gmane.org/gmane.linux.kernel/1414130

  >  But since applications *do* care, and since we *do* have 
  >  multiple error values, we stick to the old ones, unless 
  >  there are some *very* good reasons not to.

Perhaps the original email could have been about 10 lines long with the quote above being the main point? My point is: is it worth attempting to convey anger in a text only medium?

I can give as good as I get in a good old shouting match f2f. My thing is waving my arms about. Other colleagues find that funny and start laughing. Then I realise, and we all calm down, then we find a way out of the problem. I find it slightly strange trying to do that in a text only medium.

Well, i like to argue about things i am passionate about too. With friends, with colleagues, with my boss. But i try to reason with them. I don't just tell them to "shut the fuck up".

Yes, good point, but I think Torvalds was reminding the unfortunate developer of the need not to break user space applications. I can see why that is very important. But still...

If you think of Linux Kernel Development as a company then i see nothing wrong with that. In the eyes of CEO, one of his SVP did something totally unacceptable by SVP standard.

So, you seem to be ok with your boss shouting at you and insulting you for like half an hour because you make a one-time mistake. I wouldn't be ok with that in a company i am working. I'd quit. Either the SVP in your example is completely incompetent and absolutely not fit to do his job (then demote him, fire hime, whatever) OR it's a misstep (then tell him his error, make certain it doesn't happen again, reason with him but don't just insult him, what good is supposed to come out of that?!).

I certainly agree that a company/team/project needs strong leaders but being strong/focused/passionate/ambitious and being a dick to people has nothing to do with each other. At all.

Agreed. Also: praise in public, chastise in private.

If I'd done something truly stupid and irresponsible at a job, I would expect to get yelled at by my boss, but in private, not in front of the whole company.

Clearly this sort of behavior seems to work for Linus, given that there's no shortage of talented kernel developers flocking to the project, but it still rubs me the wrong way.

Good point.

> ... but it still rubs me the wrong way

The thing is you have to take it context. People that work with Linus know how he acts, and know how to interpret what he does. Linus's tendency to (temporarily) go ballistic very often rubs those unfamiliar with it the wrong way, but you learn after a while that he really isn't some kind of mega-aggressive jerk.

I don't like (and won't stand for) "abusive" control-freak type-A bosses, but Linus isn't one of those.

Points one knows after being on the LKML for a while and seeing a few Linus rants:

(1) He doesn't do this every day, but he does it periodically so it's Not Just You.

(2) He almost always does it For A Good Reason; it's not just because he's in a bad mood, it's really because you did something stupid. If you didn't do something stupid, and he's just confused, see points (4) and (5).

(3) Even if you think he's got the volume up too high, it's generally fairly proportionate to the stupidity of the original act (subtle confusion is less stupid than knowingly breaking huge numbers of apps), and proportionate to the "you should have known better" factor (i.e., his direct lieutenants are held to a much higher standard that Little-Joey-the-first-time-driver-author).

(4) He usually Calms Down Quickly, and gets more technical; really it's not personal. He wants to solve the problem, not argue. However ranty his initial message, it's not just an insult; it says what you did wrong and why.

(5) If you were really right, and he was wrong, he'll quickly admit his error after being presented with a good argument, and will start calling himself an idiot. He doesn't hold grudges. [Judging from the exchanges I've seen, though, he's usually not wrong; he has very, very, good instincts, and I suppose that's one reason why he so confidently goes off ranting...if he's not sure about something he'll ask a question instead.]

My gripe here isn't with Torvalds in particular. I just don't think anyone should behave that way in public toward other people. The context isn't particularly relevant. I don't care if Torvalds is a good or bad person, or is or isn't abusive. You just don't act that way in public toward other people. You just... don't.

Clearly opinions on this vary; I'm just stating mine.

This is not a minor mistake, to go with the company analogy, this is a kind of mistake that would fire you instantly if you did it. There are few worse things to do in the kernel than breaking user space. Would you prefer being fired for incompetence rather than being insulted?

If you fired every programmer that ever introduced a regression into bleeding edge pre-release code, you would not have very many programmers left. I don't think that is a realistic mode of operation.

Complaining about analogies is like complaining that the morphism between (Q, +, ✕) and (R, +, ✕) is not a good example because there's no √: Q -> Q. That's simply wrong, it's a very good example. They are both vector spaces and the morphism, the analogy, captures the essence of the similar behaviour in the context required.

The analogy was not between introducing regressions in open source vs. commercial software, it was between introducing this particular regression, in the way that it happened, and the abstract idea of doing something so severe in an arbitrary abstract company -- not software company -- any company; something so serious that it warrants immediate dismissal. Surely anyone can think of something that satisfies the above criteria in whatever field they are working without having to bound the free variables and twist the analogy in a way that it fails just to complain that it does.

If you have to use abstract algebra to explain your business process analogy, you probably don't have a very good analogy.

I'm not complaining about your analogy; I'm pointing out that the conclusion is simply incorrect.

Bugs happen, that's a fact of life. How and why they happen is what drives the reaction. For example, a bug caused by a simple typo is nothing to sweat over, whereas a bug caused by lack of attention to quality or by ignoring the established procedures will normally result in a 'talk'. LKML 'talks' just happen to be public, and everyone knows that going in.

That language would be less acceptable in a company, especially in public.

The employee would have grounds for disciplinary action against the person making the rant. And if it's a pattern of behaviour the employee could have a case for constructive dismissal.

> there are a ot of ways to say what you want without plainly insulting someone over and over again

Sure, and in the above context they are less effective. So why would he care?

> he might calm down a little and learn one or two things about social interaction

Why? What purpose would that serve? The LKML is one of the last few places where politically correct bullshit does not prevail. Communication is efficient and effective. There is no evidence that changing the tone would make it more efficient and more effective, but there's evidence galore that the opposite would happen. Experience has thought me that the only people hurt in the process are technical morons, and such people have no place in a technical discussion.

If someone cares more about the language or social niceties than about the correctitude of technical solutions they have no place in a technical community.

Writing a wall of text with insults and a low amount of real communiction is not effective. It's also nothing to do with political correct. Linus' second post in the mail thread is much more technical and much less insulting.. more effective.

On the contrary, reading Mauro's e-mails before and after Linus' post proves that it is very effective.

I would hardly say that it proves it. It doesn't prove that being a little bit more polite and to-the-point wouldn't be as effective.

The only thing it proves is that people working at Red Hat have to deal with linus rage outbursts by the very nature of their jobs.

You are saying that a man that brought to reality one of the biggest and successful software projects in the history of humanity is wrong calling mediocrity for what it is.

I don't know if you have already accomplished something like that, if you don't go ahead and prove that you can with your political correct methods.

I don't understand how linus being a good programmer makes him less wrong.

There are a lot of people who are nice and successful software project leaders. Being a dick is not inherent to being successful in the software field.

I think that what matters the most about Linus is not being a good programmer (in the beginning of linux maybe it was important) but his capacities for project management, maintaining a vision for so long, and calling shit "shit" when is necesary for the goodness of the project.

Well, posting an angry rant is beneficial in at least one way: It's guaranteed to be read by all the kernel devs. Hell, we're reading it, and I doubt everyone on HN is a kernel dev or follows their mailing list. If Linus had been more polite, perhaps his comment would only have been noticed by those involved with the patch in question. Instead, he has simultaneously corrected Mauro's personal misconception and reinforced an important design constraint in the minds of all budding kernel devs.

I do agree that the second post was more effective in explaining Mauro's mistake, and would prefer myself to err on the side of politeness, but I suppose there can be benefits to acting like an asshole from time to time. Some changes are not ok to make, and perhaps it's justifiable to "make an example out of someone" from time to time to ensure that everyone understands why.

"There is no evidence that changing the tone would make it more efficient and more effective, but there's evidence galore that the opposite would happen."

What is this evidence? Do you have an example of a project that failed because the tone of its community changed from hostility to politeness?

Everything you've said here is simply dismissive.

"Kernel dev. policy is not to break user space. ENOENT is not a valid return from ioctl. Kernel maintainers are expected to know and adhere to this" is the only point that Linus had to make in this discussion.

People just don't understand why he added a bunch of swearing and questioning of the dev's basic intelligence. That's all.

Obviously, to make it hurt and set an example.

"Sure, and in the above context they are less effective"

This is absolutely, 100% false. Do you have any sort of data to back this up?

"There is no evidence that changing the tone would make it more efficient and more effective, but there's evidence galore that the opposite would happen."

Again, completely, 100% could not be more incorrect.

Do you really think that just saying "X is true" makes it so?

"If someone cares more about the language or social niceties than about the correctitude of technical solutions they have no place in a technical community."

I'm not really sure its an issue of "more" one versus the other, but why can't you communicate with someone in a correct manner without screaming and swearing at them in all caps? Its not mutually exclusive that one can either talk like my grandmother or be a technical wiz.

This is probably the worst post I've ever seen on HN, and that's saying something. Every single point relies on evidence you seemed to make up in your head?

I find it easy to have empathy for Linus. He's human, just like any of us, but unlike any of us, he's responsible for the development and organization of one of the core pieces of software that drives a huge amount of stuff in the world.

He's allowed to have a cranky day sometimes, especially when somebody screws up in a way that he has been very specific about in the past.

I'm sure Linus loses a lot of sleep worrying about what buster on HN thinks regarding the tone of his language...

As a Linux user, I'm happy that Linus is passionate about not breaking USERSPACE. Had I done something of the scale that Linus is pissed about, I would have expected a much worse reaction.

There is only 2 rules in kernel development which you can't break:

1. Kernel APIs to userland are always backwards compatible

2. If you break something, you will take responsibility

If maintainer breaks both of those rules, how can you accept patches from him/her ever again? Remember that amount of patches handled by kernel core team is enormous, you have to able to trust subsystem maintainers at least some degree. In corporate world this would have caused immediate "clear your desk" reaction.

Good project must have passionate leader. Linus or Theo or Richard or Guido or Igor or Salvatore you name him..

Otherwise it will become a "kolkhoz". )

The unbalanced paren at the end of your comment really irks my OCD.

I read it as a "soft-smile".

this is russian smiley. usually used like that )))))

Passionate, sure. Insulting, no. To be honest, I think Linux is successful despite Linus' temperament, not because of it. There are many counterexamples of successful open source projects that maintain a respectful tone: WebKit, LLVM, GCC, GNOME, KDE, Perl, etc.

This is a side-effect.

I can appreciate his enthusiasm, but that man is gonna have a heart attack one of these days.

He's right of course... Breaking semantics is bad.

I wonder what he says when he's really angry ;)

I wonder if they've fixed the various breakages of dup3 and the equivalent fcntl that I reported about 4 months ago?

Why are you not linking to your report if you want to know? Your post is almost useless as it is.

Reply on news.ycombinator.com