On kernel mailing list behavior [LWN.net]

On kernel mailing list behavior

Posted Jul 17, 2013 17:54 UTC (Wed) by b7j0c (guest, #27559)
Parent article: On kernel mailing list behavior

this should be an amusing thread

should linus try to be more diplomatic? i suppose. telling people their volunteer contribution sucks in a diplomatic way will always be challenging, and the single response from linus doesn't seem to indicate that he is interested in behavior modification.

everyone has the freedom to unsubscribe. there are no employment contracts, w2s, employer-employee relationships between linus and those who volunteer patch submissions. everyone is free to exit the process at any time, or just take the code and start anew. the captain of the boat won't get far if all the rowers abandon ship.

it would be nice if everyone on the list would consent to some basic ground rules, and in short order, as i presume there will be a major backlash against any trend that seeks to take the lkml too far off-topic.

to post comments

On kernel mailing list behavior

Posted Jul 17, 2013 18:14 UTC (Wed) by smurf (subscriber, #17840) [Link] (2 responses)

I have no major problem whatsoever with people telling me that my code sucks *censored* through heavily-clogged straws. I can deal.

I do have a major problem with people telling me that *I* did that when I wrote my code, much less insinuating that I do it habitually.

The line can be blurry at times. It certainly depends on the recipient's cultural background. Professional courtesy requires erring on the side of caution.

I stopped participating in Debian discussions umpteen years ago, not because I couldn't take it, but because I didn't want to subject myself to the kind of abrasive language that was then common there. (Things have since improved, though I refuse to quantify.)

The bottom line is that the kernel community (a) does get older and (b) is not especially diverse. IMHO, if we don't want to have major problems in ten years because maintainers are leaving for age and/or burn-out reasons, we need to increase the appeal of participating *now*.

Of course, a Code of Conduct by itself will not by itself increase the appeal of kernel programming to people who are not western, white, and male. But it cannot exactly hurt either.

On kernel mailing list behavior

Posted Jul 17, 2013 19:00 UTC (Wed) by raven667 (subscriber, #5198) [Link] (1 responses)

> increase the appeal of kernel programming to people who are not western, white, and male. But it cannot exactly hurt either.

I actually think that this is a corollary to the main point, the abusive behavior is what is in question and it is not something that is discriminatorily directed at non-western, non-white, non-male developers on LKML. A scaling back of the abuse will make LKML more welcoming to all professional developers, including those who are not western, white and male, but especially those who don't feel the need to tolerate others abusive bullshit.

On skin thickness

Posted Jul 18, 2013 3:44 UTC (Thu) by zblaxell (subscriber, #26385) [Link]

Indeed.

Some western, white, male, professional, etc. people have to subscribe to LKML and interact with developers there in order to be verbally abused. Other people get their quota of abuse from non-LKML sources without making any special effort to receive it.

Abuse is not discriminatorily directed at these various groups, but is discriminatorily received by them.

There also seems to be some warped social status exchange going on. In a group it's usually a privilege to have an elite's attention at all, and if that attention leads to a response in public so much the better. Getting an emotional and memorable response (i.e. an abusive one) from an elite elevates the recipient's status among members of the group--but only members of the group that understand the structure of this interaction will see it this way. Outsiders see only one stranger hurling public abuse at another for no apparent reason.

On kernel mailing list behavior

Posted Jul 17, 2013 18:42 UTC (Wed) by dlang (guest, #313) [Link]

There are actually quite a few replies from Linus in this thread.

It's actually well worth taking the time to go ane read. (although since it's still going on, you may want to wait a bit or you will end up tracking new posts to several different subthreads)

On kernel mailing list behavior

Posted Jul 18, 2013 8:34 UTC (Thu) by daniel (guest, #3181) [Link] (21 responses)

should linus try to be more diplomatic?

The real problem is not Linus's well known antics. I can take the bad with the good in Linus's case, because Linus is Linus. You know, the Linus. Broke the mould after they made him and all that.

The real problem is people who are not Linus, but feel entitled or even obligated to act just as antisocial as Linus. These are the people who are primarily responsible for turning LKML toxic. If it was just Linus acting up from time to time and everybody else carrying on in a civilized adult manner, LKML would never have earned the reputation it now has.

Let me add my own data point. A major reason I dropped out of the kernel community for three years and did other things was the toxic environment. I could name names, but I won't. (You know who you are.)

To be sure, this was not the major reason I dropped out... the actual major reason was, when I joined a hedge fund one of the conditions was that I stop contributing to open source. Of course that was not forever, but it is telling that after I came up for air from the even more caustic financial world, I did not choose to go back to LKML. I went and did something fun for a year instead, where contributors are actually respectful to each other. (Computational geometry if you want to know.) The only reason I came back to the kernel community is because somebody I respect (Hirofumi) asked me to. Now I am glad to be back, but I sure was not coming back on my own. And I still feel terrible about the toxic environment on LKML, and the people who make it that way. I do believe that Hirofumi would agree with me.

Just my personal data point. Carry on.

Regards,

Daniel

On kernel mailing list behavior

Posted Jul 18, 2013 14:06 UTC (Thu) by viro (subscriber, #7872) [Link] (20 responses)

Would that, by any chance, be about requests to <gasp> turn htree code into something more or less readable? They did *not* start toxic; it only got to the point of really unpleasant conversations about your taste after your directly expressed opinion that taste was optional, because, you know, htree appears to work and WTF do all those unreasonable people want, anyway? Functions somewhat smaller than 400 lines of spaghetti? If somebody can't follow those, they are wimps and that's that; you won't be wasting time on that...

On kernel mailing list behavior

Posted Jul 18, 2013 20:24 UTC (Thu) by daniel (guest, #3181) [Link] (19 responses)

It wasn't you Al, though it easily could have been. However you are doing an excellent job right now of demonstrating classic toxic style. Sit back and look at your post. Did you attack a) the HTree code or b) me? Right. You attacked me. Please do not do that again. Ever. I am sure you are smart enough to catch yourself before you hit send next time.

Let's get something else clear. Cleaning up HTree is not my job, nobody pays me for that. My job was inventing it, thus handing Ext2/3/4 a new lease on life, and making it work well (many thanks to Ted for supplying the nasty telldir bit). Don't get me wrong, I find the HTree code style as cringe worthy as anyone else. And for the record, I often cringe when I read your code, think about that one please.

Cleaning up HTree is something I might have become interested in, if the environment were less toxic. Without monetary reward, what is my reward? The abuse stops? Sorry, not enough in itself. Ted is the Ext4 maintainer, he gets paid full time for that. If you want HTree cleaned up, talk to him. You might start by asking why Andreas's cleanup/generalization patches from Lustre still aren't merged.

Or instruct a kernel newbie on how to clean up HTree stylistically, that accomplishes the dual purpose of instructing the newbie (and perhaps bringing in some new blood, goodness knows we need it) and reshaping it to your taste.

Finally, you should read about Shardmap, which will obsolete HTree except for legacy volumes. See, you actually *want* me spending my time on stuff like Shardmap as opposed to rolling around in the mud and the blood and the beer, as you try to bite off a piece of my ear.

On kernel mailing list behavior

Posted Jul 19, 2013 2:19 UTC (Fri) by jimparis (guest, #38647) [Link] (1 responses)

> > Would that, by any chance, be about requests to <gasp> turn htree code into something more or less readable? They did *not* start toxic; it only got to the point of really unpleasant conversations about your taste after your directly expressed opinion that taste was optional, because, you know, htree appears to work and WTF do all those unreasonable people want, anyway? Functions somewhat smaller than 400 lines of spaghetti? If somebody can't follow those, they are wimps and that's that; you won't be wasting time on that...

> However you are doing an excellent job right now of demonstrating classic toxic style. Sit back and look at your post. Did you attack a) the HTree code or b) me? Right. You attacked me.

I find this really interesting because it shows that there can be a huge difference in how a message is interpreted.

I know nothing about you or HTree or previous disagreements. When I read Al's message, what I got out of it was:

1) htree code is unreadable
2) you thought it was fine
3) htree code seems to work
4) htree code is huge spaghetti code

My thought after reading it was "wow, sounds like htree code was really crappy", _not_ "wow, sounds like this daniel guy is terrible".

On kernel mailing list behavior

Posted Jul 19, 2013 6:14 UTC (Fri) by daniel (guest, #3181) [Link]

He was justifying being rude, pure and simple. The comments on the code are merely thin decoration of a personal attack. This: "it only got to the point of really unpleasant conversations about your taste after your directly expressed opinion that taste was optional". Count the fallacies. Then note your selective interpretation.

Also: "did not start toxic", which clearly implies that the discourse did turn toxic. To tell the truth, I don't remember anything about it, except for not feeling in any way compelled to "work for the kernel" when not being paid to do so, after I had already made a major contribution. In the event, I think Ted went in and made some cosmetic changes to it. Good. He was getting paid to do that, I wasn't. See, everything was working fine except for one thing: the toxic behaviour on the sidelines.

Also, how do you know HTree code is actually crappy? Because Al said so? Did you read it? Did you see the part where I think Al's code is ugly? I mean it. Cosmetic beauty is in the eye of the beholder. And honestly, beauty is secondary to the code being good.

And the code is good in spite of Al smearing it (and by implication smearing me). It is in fact the fastest directory indexing code in the known universe, with something like one bug turning up over the last ten years, that being a year or so after going into service. For the record, I did "come out of retirement" and help fix that. It mattered, you see.

Anyway, I told Al it wasn't him and I meant it. I never expected him to act visibly defensive about whatever he was thinking of, which I have long since forgotten.

But here we have another point: certain people feel justified in attacking contributors who are strictly volunteers. My lord. Where on earth does that come from? In the Tux3 project I have gotten some great contributions from really talented people, and of course I want more. I don't browbeat them to get more. See? Basic principle of decency.

Now here you are, jumping on the side of toxic entitlement. Try this: say thank you for that crucial piece of the Linux tapestry I provided. And think hard about whether the kernel would have gotten more like that from me and others if the culture was a little less toxic.

On kernel mailing list behavior

Posted Jul 19, 2013 12:14 UTC (Fri) by tytso (✭ supporter ✭, #9993) [Link] (16 responses)

The htree code was significantly cleaned up before it got merged -- by me. That was a special case, though. because we very much wanted the feature, and it was clear you wouldn't do it. If you think what is there now isn't great, it was ten times worse before I was done getting it to the point where I was comfortable putting the changes into mainline.

It is *not* the maintainer's responsibility to clean up, rewrite, and merge in every single code submission. That just doesn't scale. There have been features that have never made it into ext4 simply because the submitter couldn't get the code cleaned up enough.

Also, for the record, I am not paid to the ext4 maintainer. Google is kind enough to let me spend a certain amount of my time on it --- but it's not what pays my salary. And my hats are off to many of the other ext4 contributors who go above and beyond what is strictly needed "for their company", but who then make their code contributions clean enough and general enough that it is fit for the upstream kernel.

On kernel mailing list behavior

Posted Jul 19, 2013 18:57 UTC (Fri) by daniel (guest, #3181) [Link] (15 responses)

My advice to you, Ted, is that if your Google manager ever states that working on Ext4 is not part of your job description, you should leave for a more enlightened employer. Obviously, there are many who would be happy to have you.

Thanks for "cleaning up" HTree to your taste. Bear in mind that it might not have been to my taste. And indeed, I have issues with the taste in that code myself, but curiously, my issues seem to be completely different ones than your issues. Moral of the story: taste is a subjective thing.

One thing is abundantly clear, and we should get this straight: maintainers should not be assigning make-work projects to active contributors. It is a most excellent way to turn an active contributor into an ex-contributor. Instead, try to use other means at your disposal. For example, put an intern on it. I know you can.

Ted, you may not realize it but you are perpetuating the toxic culture in several ways right now:

1) You should not have been harassing me about cosmetic changes at any time, and you should not be justifying it now. If a polite suggestion results in me going in and making a series of cosmetic changes to bring the code up to "kernel standard" (whatever that is, the kernel is full of ugly code written by the same people who complain about ugliness in the code of others) then fine. Everybody happy. But harassment - and that is exactly what it was - is way over the top.

2) You should politely let this particular matter drop instead of dragging it on forever and confirming to the world that the rumoured toxic culture does in fact exist.

3) You should avoid fomenting tension in public amongst community members who are supposed to be setting an example. Please email me privately if you want to continue this thread, otherwise, beer is on me at the next meet up.

4) Mobbing. That is the technical term for it. Recognize and actively defuse, whether it happens in your own thread or somebody else's.

By the way, what about those patches from Andreas? As I understand it, they do clean up some substantive issues, such as the arbitrary limitation to two level trees.

On kernel mailing list behavior

Posted Jul 19, 2013 20:12 UTC (Fri) by dlang (guest, #313) [Link] (12 responses)

> maintainers should not be assigning make-work projects to active contributors.

one person's "make-work" project is another persons "critical for long term maintainability" project.

It's _very_ common for people to some to an open-source project and hand over a bunch of code to be implemented.

It's almost unheard of for the same person to be involved a decade later to fix or clarify that same code.

As a result, the maintainers (many of who _have_ been around for a long time) are FAR more concerned about maintainability issues than new contributers.

Sometimes the maintainer will do the work to convert a contribution to be more maintainable, but as Ted notes, that doesn't scale. What the maintainer needs to do is to push back on the contributer to get them to improve the submission.

In most cases, the contributer does so, and their next contribution is better (requiring less change).

Sometimes, the maintainer will ask the contributer to fix something else that the contributer sees as related to the contribution. sometimes the contributer does so and the result is much better, sometimes the contributer doesn't do so and their change goes in anyway, and other times the contributer refuses and goes away, and their change doesn't go in. All of these are acceptable results

On kernel mailing list behavior

Posted Jul 20, 2013 7:52 UTC (Sat) by daniel (guest, #3181) [Link] (11 responses)

You are being rather presumptuous, assuming that Ted's cleanups were critical for long term maintainability.

I will give you this: Ted's legendary hack to make telldir/seekdir work was if fact critical for long term usage of HTree, or even short term usage. But I have always said that, haven't I. Now maybe you should go read Ted's telldir code, then come here and tell me more of your theories about long term maintainability. Dare you.

By the way, did you see the rule 4 above, no mobbing? Please think about how you might have worded your comment so that it does not come across as a defence of the right to abuse/harass/whatever toxic behavior LKML has earned a reputation for.

On kernel mailing list behavior

Posted Jul 20, 2013 8:20 UTC (Sat) by dlang (guest, #313) [Link] (9 responses)

you made up those rules, I never agreed to them.

By your definition, anyone who argues with you is now guilty of 'mobbing' and violating the rules. This is exactly the type of rules gamesmanship that makes people very reluctant to allow any formal rules to be defined.

As for maintainability, where there is disagreement between contributors and existing long-term maintainers (especially when there are multiple maintainers talking), the maintainers get the benefit of the doubt. They have earned it over time.

In programming and system administration, people come into existing situations and always blame the prior person for the problems they see, and for that person being short-sighted.

When a person stays in a job long enough, they look at something and have the same reaction, but then realize that they were the one responsible, and they can remember why they did things that way. After a few times of this, they get MUCH more careful about what they are willing to accept, because they realize that they will have to maintain it over the long term.

Until you've been on something long enough to be in the position of replacing your original work with the next generation (and seeing the next generation of co-workers and their reaction to your prior work), youreally don't think about maintainability in the same way.

In my case, it's system and network administration and design, for Linus and his Lieutenants, it's maintainability of the kernel code. There are a LOT of things that they did 15 years ago that they would not accept from anyone today

On kernel mailing list behavior

Posted Jul 20, 2013 8:32 UTC (Sat) by daniel (guest, #3181) [Link] (8 responses)

Feh. You are just sticking up for the current toxic culture, including mobbing. Let Ted argue his own points please. Oh wait, it was Al. Oh wait, it's a mob, including you.

Your post is full of fallacies but we already wasted enough bandwidth on this, you find them please.

On kernel mailing list behavior

Posted Jul 20, 2013 8:39 UTC (Sat) by dlang (guest, #313) [Link] (7 responses)

I actually expected you to dismiss anything that I said.

i also expect that you will dismiss anything that Ted says.

After all, anyone who doesn't agree with you must be blindly in favour of abusing people.

On kernel mailing list behavior

Posted Jul 20, 2013 8:53 UTC (Sat) by daniel (guest, #3181) [Link] (5 responses)

You posted a patronizing lecture, proceeding from an assumption that is dubious at best. Your behaviour is toxic, and you react defensively to that being pointed out, rather than doing the right thing and considering how you might become part of the solution instead of part of the problem. By the way, you never said thank you. Good night.

On kernel mailing list behavior

Posted Jul 20, 2013 13:47 UTC (Sat) by shmget (guest, #58347) [Link] (4 responses)

" Your behaviour is toxic, and you react defensively to that being pointed out, rather than doing the right thing and considering how you might become part of the solution instead of part of the problem"

Have you ever hear of that wonderful technical invention called : 'The mirror'.
you should try it one day, dear kettle.

On kernel mailing list behavior

Posted Jul 20, 2013 20:35 UTC (Sat) by daniel (guest, #3181) [Link] (3 responses)

Your behaviour is called "blaming the victim". Part of the toxic pattern.

On kernel mailing list behavior

Posted Jul 20, 2013 20:39 UTC (Sat) by daniel (guest, #3181) [Link] (2 responses)

And further, you joined in the mobbing. And also forgot to say thank you. Time to go get that mirror out.

By the way, what is your real name? Are you a kernel developer?

On kernel mailing list behavior

Posted Jul 21, 2013 0:33 UTC (Sun) by bronson (subscriber, #4806) [Link] (1 responses)

Please just stop.

On kernel mailing list behavior

Posted Jul 21, 2013 2:35 UTC (Sun) by daniel (guest, #3181) [Link]

Et tu Bronson?

On kernel mailing list behavior

Posted Jul 22, 2013 20:05 UTC (Mon) by Baylink (guest, #755) [Link]

I've seen the behaviour Daniel is exhibiting here, before, here on LWN.

It was aimed at me, a couple years ago, in a thread also prominent for pointing me to Derailing For Dummies to acquire an education in how not to do ... something I wasn't doing in the first place.

Old timers will remember the thread; *EVERYONE* is invited not to, um, derail *this* thread into that one.

On kernel mailing list behavior

Posted Jul 20, 2013 13:44 UTC (Sat) by shmget (guest, #58347) [Link]

"By the way, did you see the rule 4 above, no mobbing?"
Now who's being presumptuous ?

"your comment so that it does not come across as a defence of the right to abuse/harass/whatever toxic behavior LKML "
Nice begging the question fallacy... but the point has not being conceded, no matter how much creative you choose to be with your vocabulary to try to pain a favorable picture of your position.

On kernel mailing list behavior

Posted Jul 20, 2013 18:25 UTC (Sat) by tytso (✭ supporter ✭, #9993) [Link] (1 responses)

The cleanup I was referring to wasn't the telldir/seekdir hack, but the fact that the code was deeply nested with huge numbers of goto commands. I fixed it by refactoring the code and adding some functions, so the flow of control was more understandable. I don't consider such changes to be "cosmetic" by the way; that would be changing the names of variables, cleaning up english/spelling mistakes, fixing whitespace. And I do a certain amount of that too. The changes I was referring to, and the problems which Al alluded to, were not cosmetic problems, but things which are fundamental to the code's long-term maintainability.

People who are curious what the difference might be are invited to look at the your original code submission, and compare and contrast that to with is in ext3 and ext4 today.

On kernel mailing list behavior

Posted Jul 21, 2013 20:26 UTC (Sun) by daniel (guest, #3181) [Link]

I know that Ted, and that can be plainly seen from my post. You should also realize that I consider your "cleanups" purely cosmetic, however I would never dream of claiming that you should not have done it. It made you happy, you see. It was pointless to me, since I regarded the original as well within the bounds of what any professional ought to be able to read without effort.

Maybe I am wrong about that. However if you want to see a lot of long functions, go read the FreeBSD code. Same type of subsystem. Multi-hundred like functions everywhere you look. They like it that way. I like it either way: I am generally more concerned about the substance.

Taste is very much subjective. If you want to exercise yours in the subsystem you maintain, then by all means do so. That is your business, not mine. If you do exercise your taste, you can do so in a polite way.

Anyway, it is clear that we differ. I am saying, clearly cosmetic, you are saying fundamental. Well, the great thing is, you were able to change it so you were happy. Why you were berating me then, and still are now, is beyond me. Now we as a community have put a word to such behaviour: toxic. And you are not by any means a prominent offender. This is a great time to sit back, take a deep breath, and ask yourself why you felt compelled to act that way, and why you feel compelled to defend it now. I am telling you it is harmful and hurtful, you respond by doing it more. It is not even the real you, this is something more akin to road rage, the main difference being "behind a keyboard" as opposed to "behind a steering wheel".

On kernel mailing list behavior

Posted Jul 18, 2013 22:30 UTC (Thu) by nyc (guest, #91222) [Link]

This was never a big issue for me while I was still an active contributor. I don't know if I'm so much thick-skinned as just oblivious. But there is a large issue with the comment to which I'm replying:

> everyone has the freedom to unsubscribe. there are no employment
> contracts, w2s, employer-employee relationships between linus and
> those who volunteer patch submissions. everyone is free to exit the
> process at any time, or just take the code and start anew. the
> captain of the boat won't get far if all the rowers abandon ship.

This putative "freedom" is preposterous. A number of the industrial contributors' job performance is judged at least partially on community acceptance of their work. And it's not that easy to jump ship, especially not in this economy.