Conflict over a code
Conflict over a code
Posted Mar 19, 2015 15:12 UTC (Thu) by corbet (editor, #1)In reply to: Conflict over a code by dunlapg
Parent article: Conflict over a code
I'm sort of curious as to where your numbers come from...if we were driving away 60% of our potential contributors, it would be hard, to say the least, to continue to grow the community. But the community does grow, so I suspect the situation is not anywhere near that bad.
There is no doubt we can do better; I acked the CoC for a reason. But I think we should be careful in how we characterize the problem and not turn it into something rather bigger than it really is.
Posted Mar 19, 2015 16:37 UTC (Thu)
by dunlapg (guest, #57764)
[Link] (14 responses)
My impression comes from my own interaction with submitting patches (which is relatively minimal), and what I hear from a number of my colleagues who work in the community on a regular basis.
It may be heavily skewed by the particular maintainers that I've had to work with. Every mail I've received from the maintainers I've worked with -- even mail approving a patch to be accepted -- has been technically sound but laced with emotional poison and made me angry. I like arguing and discussion, but there's no way that I could work with those guys on a regular basis. I know a number of developers who minimize their interaction with the LKML for similar reasons. I would be really surprised if more than 40% of all the developers in the world could put up with it. And I know a number of people who aren't happy with the status quo, but don't want to talk about it for fear of alienating the maintainers on whose cooperation they rely to do their work.
Just to be clear -- I think there are three independent axes on which any communication can be:
* Direct vs indirect
Whenever Linus is asked about the communication style on the list, he only talks about the direct/indirect axis, and says that he thinks direct is the best way. But the reality is that you can be any combination of these (direct / polite / technical; direct / polite / personal; indirect / polite / personal; &c)
From my observation, although Linus himself is occasionally rude, he is almost never personal: he always describes what someone *did* in their code, not what they *are*. Rude but technical doesn't bother me, but a lot of people I know find even that unpleasant and a drain to work with. But a lot of other people on the list are personal, which is the real poison.
40% of the developers in the world is an awful lot of space to grow in; and it's a testament to Linus' leadership in most areas of the project that it's succeeded as well as it has. But I can't help but think that it would have grown *even more* if the development community weren't alienating and making life difficult for so many people who want to participate.
Some discussion of this, including some quantification of this effect in other open-source projects, can be found here:
Posted Mar 19, 2015 17:36 UTC (Thu)
by bfields (subscriber, #19510)
[Link] (1 responses)
Very much agreed on the conflation of directness, politeness, and criticism. It's not that hard to manage all three at once.
Posted Mar 19, 2015 19:00 UTC (Thu)
by bfields (subscriber, #19510)
[Link]
Posted Mar 19, 2015 17:39 UTC (Thu)
by viro (subscriber, #7872)
[Link] (8 responses)
<googles a bit> The message(s) approving the patch to be accepted were "Looks good to me - Peter, any objections?" from mingo and "Nope, looks good" from peterz... Where's the emotional poison?
Or am I completely misidentifying you (or patch in question)?
Details, please. I assume that your "paint broad strokes" (rather than being exactly accurate) referred to percentages of developers with different reactions to l-k and not to the actual events you were mentioning right after that...
Posted Mar 19, 2015 19:48 UTC (Thu)
by dunlapg (guest, #57764)
[Link] (7 responses)
Yes, this is George Dunlap. Unfortunately I thought my nick was "gwd" and so would be a bit more anonymous. "dunlapg" doesn't even look like it's pretending to be. :-)
You may have noticed that I was purposely vague about the who and the what. I don't want to get into a public accusation about particular people, for a number of reasons. If you want an analysis of what was said and why I had the reaction I did because you genuinely want to understand what I'm talking about, then I'll try to write one up and mail you privately. If you just want to prove to yourself that everything's OK, I won't bother.
In any case, as the video above says, it's not about individual interactions, but about a pattern. After all, maybe the handful of interactions I had were on bad days -- for them (they were uncharacteristically harsh) or for me (I was having a bad day and misinterpreted what was said). But if a significant number of people have a significant number of negative interactions, then that's evidence of a real problem.
The tone of the LKML is infamous; it's brought up on a regular basis in public forums (e.g., https://www.youtube.com/watch?v=1Mg5_gxNXTo around 14:40 and 27:50), and mentioned in private even more. If you don't hear complaints about it, I'm not really sure what to tell you.
Posted Mar 19, 2015 20:23 UTC (Thu)
by corbet (editor, #1)
[Link] (6 responses)
Posted Mar 19, 2015 20:54 UTC (Thu)
by cesarb (subscriber, #6266)
[Link]
Posted Mar 19, 2015 21:29 UTC (Thu)
by bronson (subscriber, #4806)
[Link] (4 responses)
So true. Hacker News mounts up every few years to whine and moan about how poisonous the LKML is and why doesn't someone else do something about it, and then goes away.
Here's a somewhat recent example: https://news.ycombinator.com/item?id=8415667
Without exception (that I can find), those screaming the loudest are also those who have never subscribed.
To flip out about someone's public behavior, especially when your entire knowledge is based on cherry-picked tabloid quotes and not any first-hand experience... That feels somewhat hypocritical TBH.
Posted Mar 19, 2015 21:32 UTC (Thu)
by bronson (subscriber, #4806)
[Link] (1 responses)
But, taken as a whole, it's a pretty decent place to work. It's nowhere near as bad as the loudest commenters claim it is.
Posted Mar 19, 2015 22:31 UTC (Thu)
by viro (subscriber, #7872)
[Link]
The same goes for a lot of active developers; personally, I still hadn't unsubscribed, but I end up deleting unread considerably more than 90%.
These days l-k is more of a public archival mechanism; way back then it used to function as a primary development list, but that role had already been on decline back when I first subscribed to it ('98 or so). A lot of development threads get Cc'ed there, as a courtesy and for archival purposes, but that's it. More specialized lists are different (fsdevel, linux-arch, netdev, etc.), but l-k proper is too high-volume for that.
So complaints about "LKML toxic culture" sound somewhat quaint, to be honest...
BTW, speaking of the thread where the patch from dunlapg had been discussed and accepted - he had some reasons for being annoyed in that thread, but not by emotional abuse (check yourself, it's on https://groups.google.com/forum/#!topic/linux.kernel/XIEi...; everyone had been nice and polite by any possible standards). But having the patches sit uncommented for 4 weeks total *is* annoying. The bottleneck is neither on the producing side (there's a lot of contributors and they send a lot of stuff) nor in merge bandwidth (bk had helped and git had pretty much resolved that one); it's in the bandwidth of _reviewers_. Had been that way for more than a decade now.
And getting more contributions would do nothing to alleviate that bottleneck, obviously. FWIW, last year I had been on the receiving end of rather indignant comments about the inferior workflow of the kernel developers. With python, of all things, being offered as an example of How To Do It(tm). Question that had really changed the tone of the discussion was simple - "how will your superior workflow scale to <that many> commits per month and <that many> contributors?" Both being a _lot_ more than in python git repository...
It's not that we don't need the contributions or would rather have fewer folks contributing, but that's _not_ where the worst shortage is. If the submission rate would double, we would almost certainly hit a serious crisis - most of us often have long queues of things to review and integrate as it is and doubling the inflow would flat-out exceed the available bandwidth in a lot of places.
What we need is more reviewers familiar with the kernel and it takes a while to become such ;-/ I really wish more newbies would pick an area and start RTFS looking for bugs and learning the things in there - would be a lot more useful than ten thousand first patch tweaking the amount of whitespace, both for them and for us. Oh, well...
Posted Mar 21, 2015 4:46 UTC (Sat)
by marcH (subscriber, #57642)
[Link] (1 responses)
Unfortunately just the main way for "Journalism" to make enough money to survive. Nothing specific to hi-tech.
Posted Mar 21, 2015 6:13 UTC (Sat)
by alison (subscriber, #63752)
[Link]
I do think that there are particular maintainers who tend to be abusive. If one were to criticize Linus' leadership, it might be to wish that he should rein in those individuals. (Don't go scrutinizing my patches, as maintainers have been unfailingly polite and encouraging to me.)
Posted Mar 20, 2015 1:09 UTC (Fri)
by jschrod (subscriber, #1646)
[Link] (2 responses)
> 2. In the past (in my observation) he hasn't distinguished, in other
This was your main accusation, about the main problem that has to change at LKML.
Then you replied, after being asked to substantiate your accusation:
> From my observation, although Linus himself is occasionally rude, he is
This doesn't match. Does Linus attack code, or does he attack people?
Especially interesting in your GGP posting about Linus' bad influence on LKML style is the reported fact that Linus isn't subscribed to LKML and doesn't really participate in discussions there. See Al Viros comment about that, in this thread.
Posted Mar 23, 2015 17:19 UTC (Mon)
by dunlapg (guest, #57764)
[Link] (1 responses)
What I was saying was that although Linus himself (in my limited experience) tends to attack code and not people, it doesn't seem to me from his responses to questions to understand the difference between being harsh to code and being harsh to people. It's no contradiction to say that although he doesn't do a certain behavior, he doesn't seem to recognize that behavior when someone else does it.
(I'll try to address the other issues in another comment.)
Posted Mar 29, 2015 3:02 UTC (Sun)
by jschrod (subscriber, #1646)
[Link]
I didn't miss it; I didn't understand it. Maybe because I'm not a native English speaker.
> What I was saying was that although Linus himself (in my limited
OK; that i can understand and subscribe to, to a certain extend.
What I observe is: Linus only cares about his own conversation style. He reacts and contemplates only attacks that are about his personal email style; the rest get ignored by him. He doesn't accept criticism of his frank communication style readily, since it's usually targeting people he knows well and who know (and often publicly acknowledge) the reason why he sends vitriol to them.
You wrote, he attacks code, not people. Well, that's not quite true. I saw him "attack" people as well, people he trusted and who planted changes on Linux he didn't expect them to do. But this kind of "attack" is more akin to the scolding a sub-project leader gets from his guidance committee on their weekly/monthly meeting, in my experience words are much harsher there, not that I think that's appropriate and goal-oriented...
The bigger issue, IMHO, is that people seem to want Linus more to do than he's willing to provide. Linus is expected to set an example of "how to communicate on LKML". He's implicitely expected to reign in those how overstep communication style there. His own communication style to his "lieutnants" (btw, a communication style that would be private in a propriate development culture) is often attacked, sometimes not; but most often it is attributed as perpetuating communication style on LKML. That he's not even subscribed to LKML is not acknowledged and not taken into account into these postings. He get's accused to the fact that non-reigning on a mailing-list he's not subscribed to does not happen -- even though it's acknowledged at the same time that he himself is not guilty of "attacking people" (which he does, as I mentioned above).
Well, the problem seems to be: Linus doesn't seem to be feel himself to be a member of the "LKML community" however they constitute themselves. He doesn't seem to care about the communication style there, isn't even subscribed himself -- witness his commit remark, paraphrased "let's see how his pans out". He obviously thinks himself as an obeserver here, not as an actor -- and you want him to be an actor. He declines.
In my opinion, it's an question open for discussion if that is a problem that we can attribute to Linus, or if it's your problem. Who can decide what the role of Linus in Linux development shall be? Currently, only Linus himself.
You seem to want to establish more responsibilities on him that he is not willing to take. In a proprietary environment, that is easy: convince his managers and tell him to take the task. But in Linux development, you don't have an executive with that power and thus other means are sought. IMNSHO, it's a pity that the intentions behind those other means -- convincing Linus to take on an additional non-technical task that's advantageous to the overall project but that he doesn't want to do -- are not clearly spelled out.
PS: My own thoughts about that topic are not finalized. I recognize the bias in Linux' development culture towards male dominant personalities. After all, I'm married to a CS Geek myself and share her dread experience in IT culture. (Myself, actually, I would paint Ingo Molnar more as an actor in the push-contributors-away camp than Linus Torvalds, btw. Maybe you want to talk with Con Kolivas about that.) But I'm not at the point to pin that mis-balance on Linus, personally.
Even more to say, I don't think it's his reponsibility to change our communities' behavior. It's our common responsibility, for each of us. We shouldn't duck away saying "Linus does it as well". He doesn't do it, not in the sense that he's attacked of. All the other guys, with their scolding remarks targeting newbies, not following Linux path, need to touch their own noses. *We* need to tell *them* to shut off; we don't need to tell Linus how he communicate with his lieutnants.
Posted Mar 20, 2015 0:45 UTC (Fri)
by rodgerd (guest, #58896)
[Link]
I recently spoke to someone about their experiences working for a company that contributes to upstreams regularly, and their comment was that the company was great, doing work that the community could do was great, but the experience of dealing with free software projects was a *constant, grinding misery* of cockish behavior. I will note that person's next couple of jobs have been in non-free software companies.
Sure, it's anecdata. But it's not exactly isolated.
Conflict over a code
* Polite vs rude
* Technically critical vs personally insulting
Conflict over a code
(But, also agreed with corbet and viro that the numbers and examples here are very hand-wavy.)
Conflict over a code
Conflict over a code
Conflict over a code
The tone of the LKML is infamous, but I suspect it's most infamous among those who don't actually hang out there. I probably read more of LKML than just about anybody, and I often wonder where this comes from; it seems like a very 1990's view of the community. If the tone is so bad there should be plenty of examples, from within the last month, say, of unpleasant behavior. Care to point some out to me? What am I missing?
Conflict over a code
Conflict over a code
Conflict over a code
Conflict over a code
Conflict over a code
Conflict over a code
Conflict over a code
Conflict over a code
> people, the difference between attacking code and attacking people.
> almost never personal: he always describes what someone *did* in their
> code, not what they *are*.
What's your opinion? Please back it up by citations (facts).
Conflict over a code
Conflict over a code
> experience) tends to attack code and not people, it doesn't seem to me
> from his responses to questions to understand the difference between
> being harsh to code and being harsh to people. It's no contradiction to
> say that although he doesn't do a certain behavior, he doesn't seem to
> recognize that behavior when someone else does it.
Conflict over a code