When I read your version, I learn that spin locks are not a drop-in replacement for uncontested mutexes. That's great, except that I don't really know for sure what either of those terms mean, and also I didn't gain any knowledge that doesn't relate to this specific situation. It also wasn't super convincing -- I could just as easily see someone post a reply saying "well actually spin locks ARE a drop-in something something mutex" and I'd have no way to tell who was right.
Linus, on the other hand, told me about the mechanism by which the problem happens. I learned about how the scheduler sees spin-locks' loops, and that the main advantage of "real" mutexes is that the scheduler knows who's waiting on what. I also wouldn't necessarily have realized the "take clock measurement, then get descheduled before you release the lock" issue if he hadn't called it out specifically.
Sure, I guess "garbage" comes off a little mean, but "the author has a common misconception" almost sounds more passive-aggressive to me.
>Sure, I guess "garbage" comes off a little mean, but "the author has a common misconception" almost sounds more passive-aggressive to me.
The latter might be slightly passive-aggressive (although it doesn't have to be), the former is simply aggressive-aggressive.
I think Linus got upset that a badly designed benchmark got popular and therefor is spreading misconceptions. His strongly worded answer is getting popular too and as a result everybody learns and he has stopped an urban legend from spreading. A success on every points.
I think your comment is just wrong and generally pure garbage. I can't believe some poor server will have to host this nonsense for years to come. It's the most inane thing I've read all year.
I'm just talking about your comment, not you though, so clearly no hard feelings here and it's definitely the right way to frame my argument to have a nice, peaceful and constructive technical discussion.
It is certainly possible to describe a code with bad words without hurting the feelings of person who made it, and I think, in order for it to work everyone has to understand that everybody writes “garbage” or “crap” code sometimes, so when it happens, it’s not a big deal; it shouldn’t hurt your professional dignity.
On my job, in team of 4, we do it from time to time and it’s just much mentally easier, streamlined - and yes, constructive to conduct code reviews this way. I understand this may not work for other, bigger teams, but it’s certainly a valid way.
And it was more than just the "garbage" line. It was also things like "If I haven't convinced you of that by now, I don't know what I can say.", and "See how simplistic - and self-centered - your expectation is?".
Writing this way doesn't "streamline" anything. I interpreted it as a negative interaction, and clearly others have as well including the original author (since he said as much in his response).
Exactly. Blameless and non-violent language can streamline communication. It does this by removing any trigger that would make your communication partner feel defensive. Needlessly emotional and charged language is sure to distract from the subject at hand. When you use calm language, there is only one thing to focus on: the content. When you call someone's code garbage, there are now two things that that person could focus on -- the content and the tone.
To me, it is completely reasonable that the original author might have made a mistake like this. After all, when he shared his findings it seemed like he had found something interesting to me! Many people on this very site commented positively on his efforts. There is nothing "garbage" about the code that he wrote, it just doesn't quite do exactly what he thought it did.
It was blogged about, so carries some authority on the topic, and was convincing to readers, but the code did nothing close -- not just "not quite exactly" -- to what it was portrayed to be doing in the benchmarks or article. It is garbage; useless; should be binned. I tried to come up with better adjectives -- "dangerous", "useless", etc -- but they came up short, in subtle ways. Who cares; Linus explains what he means better than the one, concise word.
It always contains some information payload. If you felt triggered by it and failed to parse the content then maybe you're overly emotional.
For some people who aren't it's going to be useful assessment of how bad this code was.
I see this ongoing trend (Twitter) that by limiting the ways in which people can express themselves, you can in theory broaden the recipient circle (to include people who otherwise would be offended and wouldn't read).
In practice you not only limit the ways in which information can be distributed but often also barter away some people who weren't offended and would get useful information this way.
Looking at speech this way is also a bad idea because today's acceptable is tomorrow's unacceptable word and the only way to allow all information is to allow all speech.
"maybe you're overly emotional." Feel free to label me however you please. It does not change the fact that if your goal is to communicate ideas efficiently, my way will help you avoid alienating some folks, and has no practical downside.
> and has no practical downside That's the place when we differ in opinion. It has one direct downside of preventing some people from expressing their opinions and conveying the point. And another in the long run – setting up a precedents of what can and cannot be said.
> Feel free to label me however you please It was general note not something directed at you.
Certainly, ad hominem attacks are worse than making substantive attacks, but is that really the standard we should judge it by? Instead of a sprawling thread of visceral, wouldn't it be better if he just explained the reasons why the approach is bad in a professional manner?
Writing like this just seems sloppy and non-professional. It's rambling and unclear. It could have been far more effective if written point-by-point with far less emotion.
I would answer your rhetorical questions with an emphatic "no." People's emotions and mental state are important. It's also important that we communicate, understand, and think about people's interior worlds.
Question for you: What else is Linus communicating with his writing style? Does he show that he cares? About what? What can we tell about his values? Are these values we want in a leader?
Sometimes it's good to take off the prescriptivist hat and try on Honest Curiosity.
Explaining why and educating about the concepts, is the better approach, and was there as well. Short and concise would be best, while limiting the negative to exactly what need be pointed out: A spade is a spade after all.
If it stops people from developing "their own userspace spinlocks", it may even be kind of necessary to be quite direct.
I don't see how this is the case. One can critique a style of communication professionally just like any endeavor. I'm not "bashing" Linus, or even suggesting that he should be less honest. Just that there are far more effective ways to make the same point.
Has that ever happened? My impression was that he only ever takes a rough tone with high profile coders and is nice and polite to "poor coders". Also, he never insults them personally, but says that their code/argument is bad and that they are smart enough to know that already. Hence, they should stop making excuses and fix it already.
This is especially inexcusable coming from Linus because it's not like he needs to resort to these tactics to be heard or make an impression, he's one of the most respected software devs in existence (and for good reasons). You could remove all the abuse from his posts and you'd lose exactly nothing IMO.
But this discussion is hackneyed, I know many people love Linus' style and think it's perfectly acceptable and even a good thing, I'm just definitely not one of them.
It’s sad that there are people who think calling garbage code garbage is inappropriate. What goes in their mind? Don’t they smell or call a smell a smell?
If this was a truly idiotic / trivial mistake I doubt Linus would put much thought or time into such a long post. I also think that if you are going to write a blog post, you are implicitly supposing a place of authority and therefore opening yourself to criticisms. In other words, writing a blog post that is incorrect (in this case, very incorrect) is actively harmful as it misleads others, and it is important to call out wrong posts when necessary. I've posted dumb advice with a know-it-all voice before too (see [1] for an embarrassing HN comment) and I am not at all angry/annoyed that I was similarly called out.
[1] https://news.ycombinator.com/item?id=20017832 (edit: wrong link)
In my opinion calling the code garbage is much nicer than "The author of that post has a common misconception", which sounds like a passive-aggressive way to call someone dumb. Everyone writes bad/'garbage' code once in awhile.
Not at all. It's a polite way to say they're ignorant of a particular thing. It would be rather insecure of them to read into it and think they're being called dumb across the board.
When you say "the person who wrote this code has a common misconception", that's a statement about the person. You've jumped to a conclusion about the reason the person wrote the bad code. You're telling the person to accept that they don't know as much about mutexes as you, that they're ignorant, that they're a worse programmer (albeit in a fairly narrow category).
This is why I think GP's wording comes off worse even though it's superficially more polite.
Giving someone an out is a great thing to do, especially for people for whom saving face is important. But I don't think "this code is garbage" does that. Something like, "that code is measuring timing incorrectly because X", conveys the same information without being inflammatory. "Garbage" is a value judgment, while just pointing out that it's doing something incorrectly is merely stating a fact.
Saying that stuff at the beginning only clutters the reasoning, as one should not "just believe" that that code was "indeed garbage", just because he said so.
Additionally, _sometimes_ offering someone an out to save their ego actually does them a disservice.
If you were forced to summarize the entire first third of the email into ten words, that might be valid. As-is, it's way too short to be a proper explanation. It's more of a reminder note. If someone doesn't know exactly what you mean at a single glance, it's useless. Linus has an actual explanation in his email.
Imo, he can call something garbage if the mentioned garbage makes false and bold claims about his and contributors decades of work.
By the way, Linus' answer is much more informative than yours. I got a lot of useful insights from it and had fun just reading his lively language.
It's a bit disingenuous to suggest that Torvalds meant it in that sense, though. He very obviously meant it in the sense of a piece of trash that should be thrown into a trash bin.
(Also I've never heard of dead code referred to as "garbage code".)
People should be shamed for being ignorant. It stops ignorance from spreading.
reacting in a haughty manner doesn't stop you from being wrong, it just makes you a dick.
"The author of that post has a common misconception, which is that spin locks are a fast drop in replacement for uncontended mutexes. They aren't. You can only use them well when you know you won't be descheduled, like we do in the kernel. In a generic userspace program, you don't have that knowledge. The possibility of your time slice ending skews all the time measurements, for example."
Boom. Roughly the same message. Most people will get the same amount of edification from this as Linus' original letter. You could go in to more detail if you want but it's not really necessary. Smart people will figure it out if you just point out the general shape of the error. No need for calling code "garbage."