If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Some Guy)   Apparently Linus Torvalds can channel Steve Jobs when he wants to. (warning: serious nerdspeak mixed with Not safe for work language)   (article.gmane.org) divider line 55
    More: Amusing, regression, kde  
•       •       •

2653 clicks; posted to Geek » on 29 Dec 2012 at 10:40 AM (1 year ago)   |  Favorite    |   share:  Share on Twitter share via Email Share on Facebook   more»



55 Comments   (+0 »)
   
View Voting Results: Smartest and Funniest

Archived thread

First | « | 1 | 2 | » | Last | Show all
 
2012-12-29 03:13:53 AM  
Seriously....
 
2012-12-29 08:39:22 AM  
Well, "If a change results in user programs breaking, it's a bug in the kernel. We never EVER blame the user programs" seems like pretty much the opposite of Jobs to me.

But I can get a little tunnel vision when talking about Torvalds.
 
wee
2012-12-29 09:19:56 AM  

Gonz: seems like pretty much the opposite of Jobs to me.


Yeah, I don't see it either.  Also, what Torvalds is saying is 100% valid from a technical standpoint.
 
2012-12-29 10:25:07 AM  
Isn't this like the 3rd or 4th time that Linus has warned Mauro that he doesn't "_ever_ want to hear that kind of obviousgarbage and idiocy from a kernel maintainer again?"
 
2012-12-29 10:53:44 AM  
i'd be yelling too.

kernels don't need quick fix mentality.
shiatty changes can always be merged in after they are cleaned up and kernel ready.
 
2012-12-29 10:55:05 AM  
burn.gif

And.... Daaaaamn
 
2012-12-29 11:30:46 AM  
Uh, this is almost exactly the opposite of how I feel Apple was run under Jobs.
 
2012-12-29 11:33:44 AM  
i236.photobucket.com
 
2012-12-29 11:36:32 AM  
Omg Linus *Jizzes pants*

/sad pathetic nerd
//and damn proud of it
 
2012-12-29 11:36:57 AM  
The Olympic shot-putter?
 
2012-12-29 11:44:17 AM  
Torvalds is correct in every point in this. If a program works before a kernel commit and doesn't afterwards it is obviously the kernel fix that broke things. What the maintainer is doing is like saying a car was running just fine until you put water in the gas tank and start blaming the radio because the car wont run.
 
2012-12-29 11:50:39 AM  
Where's the farking hero tag? If only Microsoft (or EMC for that matter for anybody who's suffered through EMC Networker) was run this same way!
 
2012-12-29 11:55:37 AM  
Linus Torvalds? I hear he sleeps with nunchucks.
 
2012-12-29 11:59:56 AM  

Trollin4Colon: The Olympic shot-putter?


You're thinking of Torvald Utne.

Do you not even skim the briefings in your inbox?
 
2012-12-29 12:12:42 PM  
I wish we had a Linus Torvalds where I work.

We've got some developers that could use a proper talking-to.
 
2012-12-29 12:22:59 PM  

wee: Gonz: seems like pretty much the opposite of Jobs to me.

Yeah, I don't see it either.  Also, what Torvalds is saying is 100% valid from a technical standpoint.


Talk to someone who worked at Apple on a pet project of Jobs. Sure, he had firm conviction that his design decisions were correct and would not hesitate to insult customers who disagreed. However, this pales in comparison to the insults, scorn and vitriol he would heap on his own engineers if they made a mistake, or what he perceived to be a mistake.
 
2012-12-29 12:26:23 PM  
Some days I think Linus would love that his benevolent dictatorial powers included summary executions - For The Encouragement of Others...

Seriously though, He is EXACTLY what the manager of a very complex software project should be like.  If only the developers in the company I work for could be exposed to that...

/ No Chuck, there is no C:\ in Linux....
 
2012-12-29 12:37:35 PM  
agree with pretty much everyone here so far.

Also, the impression I get of him is that he's made his point now and he can probably have a jolly good time with the guy at lunch tomorrow. Blunt, honest, and forgiving is a good way to be and a good kind of manager to have.
 
2012-12-29 12:55:04 PM  
ENOENT from ioctl does deserve a few curse words
 
2012-12-29 01:04:48 PM  
This is the difference between the Bazaar and the Cathedral. The Bazaar is not afraid of telling you to SHUT YOUR WHORE MOUTH WHEN REAL CODERS ARE TALKING! And they are not afraid to do it publicly.

Good on Linus. Some dumbasses need a biatchslap now and then.
 
2012-12-29 01:16:11 PM  
Also, coders will cuss liberally. (NSFW language, obviously.)
 
2012-12-29 01:17:26 PM  
You don't screw around in the kernel. All the kernel driver programmers I know are serious coders writing scary code.
 
2012-12-29 01:46:14 PM  
We should fly Linus and Gaben together for semi-weekly meetings.  If for no other reason than to get them into the same room so I can hug them mercilessly.

Two of my favorite people in the world.
 
2012-12-29 01:55:13 PM  
Linus Torvalds
 
2012-12-29 02:06:01 PM  
What's interesting is that reading further down the thread, turns out the change in kernel behavior Linus got so pissed about was actually a bug (the changed error code was supposed to stay inside the kernel, not sent to userspace processes). So, like so many managers, he burst in guns blazing way too soon.

Linus actually does this quite often, and 90% of the time for the right reasons. Unfortunately this is not one of those times.
 
2012-12-29 02:14:20 PM  

wee: what Torvalds is saying is 100% valid from a technical standpoint.


If you say so...

*shrugs*


Kuroshin: We should fly Linus and Gaben together for semi-weekly meetings.  If for no other reason than to get them into the same room so I can hug them mercilessly.

Two of my favorite people in the world.


Are they making a Linux Steam box?
 
2012-12-29 02:14:24 PM  

rekoil: What's interesting is that reading further down the thread, turns out the change in kernel behavior Linus got so pissed about was actually a bug (the changed error code was supposed to stay inside the kernel, not sent to userspace processes). So, like so many managers, he burst in guns blazing way too soon.

Linus actually does this quite often, and 90% of the time for the right reasons. Unfortunately this is not one of those times.


Three points:

1. Why change error codes? Period. This is a definite backwards-compatibility issue. Even if it is "wrong", you don't change it and risk breaking things.
2. As somebody pointed out, "ENOENT from ioctl does deserve a few curse words". I/O has nothing to do with pathing.
3. Mauro whole approach to the issue was farked from the beginning and he deserved a few more cuss words for that.
 
2012-12-29 02:29:14 PM  
The sad thing is the monkey will still most likely put the kernel maintainer position on his résumé. Even worse, most interviewers will never know this screw-up ever happened.
 
2012-12-29 02:53:27 PM  
This used to be the same attitude that Dave Cutler applied when he ran Windows NT development. If a program can bring down the OS, it's entirely the OS's fault.

No excuses.
 
2012-12-29 03:20:23 PM  

moviemarketing: wee: what Torvalds is saying is 100% valid from a technical standpoint.

If you say so...

*shrugs*


Kuroshin: We should fly Linus and Gaben together for semi-weekly meetings.  If for no other reason than to get them into the same room so I can hug them mercilessly.

Two of my favorite people in the world.

Are they making a Linux Steam box?


That's one rumor.
 
2012-12-29 03:20:35 PM  

DicksWii: Linus Torvalds? I hear he sleeps with nunchucks.


Ninja gaijin? I love that game!
 
2012-12-29 03:52:28 PM  
The Slashdot thread on this is fascinating in its own right as an examination of social mores.
 
2012-12-29 04:08:23 PM  

blue_2501: rekoil: What's interesting is that reading further down the thread, turns out the change in kernel behavior Linus got so pissed about was actually a bug (the changed error code was supposed to stay inside the kernel, not sent to userspace processes). So, like so many managers, he burst in guns blazing way too soon.

Linus actually does this quite often, and 90% of the time for the right reasons. Unfortunately this is not one of those times.

Three points:

1. Why change error codes? Period. This is a definite backwards-compatibility issue. Even if it is "wrong", you don't change it and risk breaking things.

The change in the error code wasn't intended to be noticed by userspace processes, it was intended to stay within the driver code and then "normalized" once passed to the kernel (and to userspace from there). The fact that it was seeing externally was a bug that was caught by QA folks running the release candidate, which is exactly the job that QA is supposed to do.

2. As somebody pointed out, "ENOENT from ioctl does deserve a few curse words". I/O has nothing to do with pathing.

Agreed, not the best choice of error codes, but no one would have noticed or cared if not for the bug noted above.


3. Mauro whole approach to the issue was farked from the beginning and he deserved a few more cuss words for that.

Agreed on that point, kernel behavior changes that aren't documented should never NOT be considered a regression. But Linus's tirade IMO was based on an assumption that Mauro had knowingly allowed the error code change to go in, which simply wasn't the case.
 
2012-12-29 05:22:50 PM  
i.imgur.com
 
2012-12-29 05:55:16 PM  

Phil Payne: Torvalds is correct in every point in this. If a program works before a kernel commit and doesn't afterwards it is obviously the kernel fix that broke things. What the maintainer is doing is like saying a car was running just fine until you put water in the gas tank and start blaming the radio because the car wont run.


Just out of curiosity...

if a program has worked for a very long time by doing something that it should never had been able to do, say by exploiting a known bug in the OS. And then along comes a patch that fixes the bug in the OS and the program stops working, that is considered something to be avoided because it "breaks a userspace application" and therefore is bad?
 
2012-12-29 06:43:19 PM  

Benjimin_Dover: Phil Payne: Torvalds is correct in every point in this. If a program works before a kernel commit and doesn't afterwards it is obviously the kernel fix that broke things. What the maintainer is doing is like saying a car was running just fine until you put water in the gas tank and start blaming the radio because the car wont run.

Just out of curiosity...

if a program has worked for a very long time by doing something that it should never had been able to do, say by exploiting a known bug in the OS. And then along comes a patch that fixes the bug in the OS and the program stops working, that is considered something to be avoided because it "breaks a userspace application" and therefore is bad?


Yeah, that's dumb. Linus is being pretty over the top here. There are no absolutes in this arena--there are userspace apps that exploit kernel bugs (intentionally OR inadvertantly). If the bug is harmless, it may make sense to grandfather in this behavior and create a new API. Or special case what the app is doing.

Not all bugs are harmless, though, even if the apps that exploit them are.
 
2012-12-29 06:51:36 PM  

wee: Gonz: seems like pretty much the opposite of Jobs to me.

Yeah, I don't see it either.  Also, what Torvalds is saying is 100% valid from a technical standpoint.


And wrong from a political standpoint. Trovalds is a one trick pony and he would serve his pony better by not being a fark'n asshole in public.
 
2012-12-29 07:16:03 PM  

Benjimin_Dover: Just out of curiosity...

if a program has worked for a very long time by doing something that it should never had been able to do, say by exploiting a known bug in the OS. And then along comes a patch that fixes the bug in the OS and the program stops working, that is considered something to be avoided because it "breaks a userspace application" and therefore is bad?


No. Code should work to a specification, and the users should base their code on that specification.

The interface as specified is like a contract, what the code is supposed to do. It's the developers responsibility to make it do what it is supposed to do, and nothing more. They shouldn't have to be thinking about all the people using undocumented features in the code. If you're a user taking advantage of those undocumented features, well, enjoy it while it lasts. If they plug the hole then you're program is going to break.
 
2012-12-29 08:24:21 PM  

farkeruk: undocumented features


There is no such thing as documentation in Linux and you damn well know it.

Everything every Linux program does is an undocumented feature.
 
2012-12-29 09:19:14 PM  

Benjimin_Dover: And then along comes a patch that fixes the bug in the OS and the program stops working, that is considered something to be avoided because it "breaks a userspace application" and therefore is bad?


Yep. You work with the maintainers of the app to make sure nothing is broken. Application developers are the users of the kernel. Your users are your life. Customer is always right.
 
wee
2012-12-29 09:27:15 PM  

ryarger: Sure, he had firm conviction that his design decisions were correct and would not hesitate to insult customers who disagreed.


Linus, in this case, is "Apple".  The "customers" are everyone who buys their crap. You don't make a change willy-nilly that breaks everything you've "sold".  "Why'd this last iTunes update prevent me from syncing my iPod?"  "Why can't I sync that iPod with the computer after I install a new browser?"  That's the sort of thing we're talking about, breaking user-space programs, not Jobs berating some geek who wants to cower in his cube long enough to cash in his stock and leave.
 
2012-12-29 09:56:14 PM  

Benjimin_Dover: Phil Payne: Torvalds is correct in every point in this. If a program works before a kernel commit and doesn't afterwards it is obviously the kernel fix that broke things. What the maintainer is doing is like saying a car was running just fine until you put water in the gas tank and start blaming the radio because the car wont run.

Just out of curiosity...

if a program has worked for a very long time by doing something that it should never had been able to do, say by exploiting a known bug in the OS. And then along comes a patch that fixes the bug in the OS and the program stops working, that is considered something to be avoided because it "breaks a userspace application" and therefore is bad?


Yeah, that falls under "It's not your fault, but it IS your responsibility."  I always hated that, but always had to live by it.

What it comes down to is that the users won't care.  All they will know is that their software had run fine for a LONG time until YOUR crappy update screwed everything up.  They will be PISSED. And maybe for good reason if the software they're running is business/mission critical, was insanely expensive to buy and transition to, and rarely if ever updated -- IOW, most business and government software running today.

Don't even TRY to talk to them about "bugs in the legacy application." It was YOUR update that broke it, and you as sure as hell better fix it ASAP or there will be Hell to pay.

That reminds me of a story about early development of MS Windows:

Apparently SimCity for DOS had a bug where it would allocate memory, free it, then keep using it.  Since DOS (generally) wasn't multitasking, the bug never was an issue until Windows came along.

Under Windows, the SimCity would crash when it tried to access memory it had freed-- as it should.

Still, SimCity was a damn popular game.  If it didn't work under Windows, that could really hurt Microsoft's reputation-- people wouldn't care that it was a bug in the old program, just that a very popular bit of software wouldn't work under Windows. So instead of releasing a version of Windows that behaved "properly" and just let SimCity crash like it was "supposed to," Microsoft modified Windows to treat SimCity as a special case and not actually free memory that SimCity releases.

Apparently the list of "special case" software titles grew into the dozens if not hundreds, With Windows bending the "rules" in different ways for each one.

t2.gstatic.com
 
2012-12-29 10:11:55 PM  
Oh, yeah, and not only is Linus being a total asshole, he is being a terrible manager.  You are always supposed to praise in public and criticize in private.

He just publicly humiliated some really talented guy who was donating his time and skill.  Maybe he was screwing up (maybe not when you look at the details.)  It takes a REALLY mature person to be able to go through a drubbing like that and not feel deeply angry and resentful-- probably only a tiny percent of people could.

shiat like this only hurts Linux in the long run. There are precious few people who understand how the Linux kernel works even in general terms, and far fewer still who know it well enough to work on it.  Linus should be inspiring the not-so-great kernel programmers to do better, not drive them away.
 
2012-12-29 10:31:36 PM  

gingerjet: wee: Gonz: seems like pretty much the opposite of Jobs to me.

Yeah, I don't see it either.  Also, what Torvalds is saying is 100% valid from a technical standpoint.

And wrong from a political standpoint. Trovalds is a one trick pony and he would serve his pony better by not being a fark'n asshole in public.


One trick pony? The guy also created Git. You put your code on Github? You got Torvalds to thank for that, and not just for the OS that it's running on.

Benjimin_Dover: if a program has worked for a very long time by doing something that it should never had been able to do, say by exploiting a known bug in the OS. And then along comes a patch that fixes the bug in the OS and the program stops working, that is considered something to be avoided because it "breaks a userspace application" and therefore is bad?


Yes. It's called "backwards compatibility". Even if it's unexpected behavior, you don't break it. Now, how you do or don't break backwards compatibility varies per application/module/whatever. But, the kernel is the CORE, so BC is absolute here.
 
2012-12-29 10:32:23 PM  

wee: ryarger: Sure, he had firm conviction that his design decisions were correct and would not hesitate to insult customers who disagreed.

Linus, in this case, is "Apple".  The "customers" are everyone who buys their crap. You don't make a change willy-nilly that breaks everything you've "sold".  "Why'd this last iTunes update prevent me from syncing my iPod?"  "Why can't I sync that iPod with the computer after I install a new browser?"  That's the sort of thing we're talking about, breaking user-space programs, not Jobs berating some geek who wants to cower in his cube long enough to cash in his stock and leave.


If you're subby then you truly have no clue. Apple's OS is the most rock solid consumer grade OS (out of a field of two, admittedly). Most apps from Rhapsody days (I ran Yellow Box on a Pentium II) still run today.
 
2012-12-29 11:40:40 PM  
One thing i love about s/w development is that its so easy to recognize those with "kruger dunning". Just by the opinions of some in this thread, i can distinguish between those who "get it" and those who simply echo what they heard in a first year university course without really being able to put info into context. Linus is absolutely right and has enough power/position/credibility to handle the situation in the most efficient manner possible. Its my opinion that he has had it with this Mauro guy and wants him out, because there is no way to change this kind of developer. Mauro is tying his ego to his code and is rationalizing his choices. There is not much you can do to change this fundamental flaw; your best bet is to cut the guy loose and encourage him to go work for some mediocre company where he will fit right in.

Hasn't anyone noticed that excellent s/w products come from teams with ruthlessly honest leadership that doesn't tolerate (even slight) incompetence? The "politician type" leader that most companies foster do not fit this mold, and as a result just about all software from these companies suck balls.
 
2012-12-30 03:39:07 AM  
Linus Torvalds has been like that for years. This is nothing new. It's awesome that he's managed a project of this magnitude for 20+ years.
 
2012-12-30 04:45:22 AM  
It's one thing to do "oh it must be on your end" but to get caught at it and then still try to shift blame... damn. And Mauro didn't even write the code he was trying to stand up for at first. I've done programming. I've met people practically married to their code.

"No no it must be on your end! My precious!!!! Hisssssss" while practically hugging their computer.

Related mildly creepy story: When I was a first year comp sci student, one of the 3rd year students was obsessed with Heidi Wall (yes, daughter of the guy who invented perl). He had a Heidi Wall desktop wallpaper, Heidi Wall mug etc etc. Which wouldn't have been so creepy except he was in his mid to late 20s and she was, I think, 15.
 
2012-12-30 12:03:46 PM  

GiantPeon: Hasn't anyone noticed that excellent s/w products come from teams with ruthlessly honest leadership that doesn't tolerate (even slight) incompetence? The "politician type" leader that most companies foster do not fit this mold, and as a result just about all software from these companies suck balls.


I am furiously jacking it to this sentiment. I have worked in the worst of the worst software groups you could imagine, and have seen this pattern play out over and over. What happens to your average shiatty software group is so sad and painful to watch: slowly but surely, they lose all their customers to upstart development houses with 10 developers.

But first, they shift more and more resources from R&D to development, then from development to maintenance, then finally from maintenance to production support. shiatty software isn't cheap to build, contrary to popular belief. It's also hideously expensive to maintain. Angry customers require even better support than happy ones, which just magnifies the problem. Takes about two-four years for it all to play out.

Corrective action is usually carried out wrong: moving work to India to get it done cheaper, quality slips even more, more work required, more Indians hired, the last 20% of the company's value ends up in a ditch in Bangalore. And it all starts with "meh, good enough".

The Linux kernel is too important to sacrifice quality over someone's feelings. Millions of lives depend on it. No mistakes allowed.
 
2012-12-30 04:46:38 PM  
I don't agree that the kernel is necessarily at fault as there might have been a very old bug in the kernel and everybody using the API coded accordingly, so everything "worked", but you would have to be an idiot to push out a kernel change that would affect any legacy code.

/otherwise agree with Linus
 
Displayed 50 of 55 comments

First | « | 1 | 2 | » | Last | Show all

View Voting Results: Smartest and Funniest


This thread is archived, and closed to new comments.

Continue Farking
Submit a Link »
On Twitter





In Other Media


  1. Links are submitted by members of the Fark community.

  2. When community members submit a link, they also write a custom headline for the story.

  3. Other Farkers comment on the links. This is the number of comments. Click here to read them.

  4. Click here to submit a link.

Report