Slashdot Mirror
The Scalability of Linus
Hugh Pickens writes "Katherine Noyes writes at LinuxInsider that it may be time for Linus Torvalds to share more of the responsibility for Linux that he's been shouldering. 'If Linux wants to keep up with the competition there is much work to do, more than even a man of Linus's skill [can] accomplish,' argues one user. The 'scalability of Linus' is the subject of a post by Jonathan Corbet wondering if there might there be a Linus scalability crunch point coming. 'The Linux kernel development process stands out in a number of ways; one of those is the fact that there is exactly one person who can commit code to the "official" repository,' Corbet writes. A problem with that scenario is the potential for repeats of what Corbet calls 'the famous "Linus burnout" episode of 1998' when everything stopped for a while until Linus rested a bit, came back, and started merging patches again. 'If Linus is to retain his central position in Linux kernel development, the community as a whole needs to ensure that the process scales and does not overwhelm him,' Corbet adds. But many don't agree. 'Don't be fooled that Linus has to scale — he has to work hard, but he is the team captain and doorman. He has thousands doing most of the work for him. He just has to open the door at the appropriate moment,' writes Robert Pogson, adding that Linus 'has had lots of practice and still has fire in his belly.'"
"What If Linus Torvalds Gets Hit By A Bus?" - An Empirical Study
by Leonard Richardson
Published on segfault.org 02/23/2000
http://www.crummy.com/writing/segfault.org/Bus.html
It even coined the "Bus factor" phrase:
http://en.wikipedia.org/wiki/Bus_factor
1% APY, No fees, Online Bank https://captl1.co/2uIErYq Don't let your $$$ sit in a no-interest acct.
A good new fork will only be needed if and when Linus stops scaling. Until then, the reason that a fork has never really come together is because Linus' Linux is the best all round solution.
Linux is Linus's creation, he should have ultimate commit decision power
Any fork would either immediately or very quickly suffer from the same fate. If there's anything the open source community has a surplus of, it's egos.
--- Do you believe in the day?
...and still has fire in his belly
Perhaps he should eat less Mexican food.
If Pandora's box is destined to be opened, *I* want to be the one to open it.
Given that, in practice, virtually every distro maintains its own kernel, or set of patches, to suit their needs, I don't really see the big deal.
As long as Linus is performing his role of keeping the "official" repository basically the easiest and most standard starting point, all the peripheral kernel tweaks maintained by other entities will cluster more or less closely around it for cost reasons.
If he starts to slip, the center of gravity will shift toward one of the distro kernel repositories, or whatever other third party is doing the best job of filling the role, and the "official" repository will fade in prominence a bit.
Because of how kernel code is licensed, the "official" repository could either come back quickly(if Linus or his chosen successor get back on the ball, they could update from the prior leader, and start taking the comit lead again), or it could just fade away, mostly, and development could center around the RedHat tweak of the kernel, or the Debian one, or whatever...
More dangerous are situations(like the X11/X.org one) where there is a major licensing split that actually requires a decisive move one way or the other. Linux graphics are certainly not its strongest suit; but, had the defection to X.org not been so complete, things there could have been a lot uglier today.
The Linux kernel is not a company. Free software projects are a new kind of entity.
The debate is still open about whether it is correct to level "They should..." instructions at this kind of entity.
Possibly "I should..." statements are more appropriate.
-paul
It's called Andrew Morton
how long until
That will solve this problem once and for all.
No. The kernel is(at this point, whether anybody likes it or not) basically GPL2 permanently. Without any "copyright assignment" requirement to some organization, there are just too many interlocking owners for any re-licensing.
Already, most distros maintain slightly forked versions of the kernel, to suit their needs(ie. enterprise-ish ones like RedHat might do more driver backports, MontaVista introduces BSPs for a variety of oddball boards, etc.) Because novelty costs money, people don't generally go further from mainline than they have a good justification for; but there are already dozens of quiet, not-very-adversarial, slight forks floating around, mostly in the hands of the various distros, and some of the embedded engineering houses.
PREPARE THE ELECTRODES!
We've been on 2.6.X since 2003. Somebody needs to pull the cork out ...
politicians are like babies' nappies: they should both be changed regularly and for the same reasons
Isn't that what blogging is for?
A good new fork will only be needed if and when Linus stops scaling.
If? You say this as though it isn't inevitable. Linus could be hit by a bus tomorrow, or (more likely) die of cancer in 10 years. He could even retire from the project! Either way, there will eventually be an end to his influence.
... and then they built the supercollider.
Arrogant people who achieve power never give it up voluntarily.
Er, didn't Bill gates step down at Microsoft?
Free Martian Whores!
Some of the people he tells to fuck off are stupid, some are not. Some of the ideas he shits on are stupid, some are not. I seen plenty of times on LKML where he is dismissive and insulting only to later actually look at the ideas in detail and then accept them. The acceptance is sometimes in the form of repackaging the idea by a different, more favored developer so that there is never a need to acknowledge the original contributor may have been right.
He seems to function well enough, but do not pretend he is perfect.
He also said "and when" to indicate it was inevitable. This admittedly made the "if" part completely stupid.
which is totally what she said
Anything from Linuxinsider I would take with a healty dose of skepticism - it's better known for their anti linux shills.
Arrogant people who achieve power never give it up voluntarily. They hold onto every little bit of it for dear life.
Why must a pet project, the owner from inception, held in high regard around the entire world, which has provided both personal and professional benefit, as well as providing much prestige, be attributed to arrogance? Unless your Buddhist or some such thing, most reasonable people would argue you're foolish to give up such a thing until your damn well ready.
Really, in exchange for giving up all that, what does it get in exchange? Nothing aside from more free time, as far as I can tell.
What you descibe was the way kernels were being handled in the early 2000's. The 2.5.X kernels were getting all the dev time, and the distros were back porting fixes. You are right though, if the Linus branch starts slipping, a distro or some other maintainer branch will become the "base" branch. The way it is right now there are several branches I could follow/use if i wanted to.
[N] sys-freebsd/freebsd-sources (--): FreeBSD kernel sources
[N] sys-kernel/cell-sources (--): Full sources including the cell/ps3 patchset for the 2.6 kernel tree
[N] sys-kernel/ck-sources ((~)2.6.34-r1(2.6.34-r1)): Con Kolivas' high performance patchset + Gentoo patchset sources
[N] sys-kernel/gentoo-sources (2.6.32-r11(2.6.32-r11)): Full sources including the Gentoo patchset for the 2.6 kernel tree
[N] sys-kernel/git-sources ((~)2.6.35_rc6(2.6.35_rc6)): The very latest -git version of the Linux kernel
[N] sys-kernel/hardened-sources ((~)2.6.34(2.6.34)): Hardened kernel sources (kernel series 2.6)
[N] sys-kernel/mips-sources (--): Linux-Mips GIT sources for MIPS-based machines, dated 20100605
[N] sys-kernel/mm-sources ((~)2.6.28_rc2-r1(2.6.28_rc2-r1)): Andrew Morton's kernel, mostly fixes for 2.6 vanilla, some vm stuff too
[N] sys-kernel/openvz-sources (2.6.27.6.1(2.6.27-kuindzhi.1)): Kernel sources with OpenVZ patchset
[N] sys-kernel/sparc-sources (--): Full sources for the Gentoo Sparc Linux kernel
[N] sys-kernel/tuxonice-sources ((~)2.6.34-r1(2.6.34-r1)): TuxOnIce + Gentoo patchset sources
[N] sys-kernel/usermode-sources ((~)2.6.18-r2(2.6.18-r2)): Full sources for the User Mode Linux kernel
[N] sys-kernel/vanilla-sources ((~)2.6.35_rc5(2.6.35_rc5)): Full sources for the Linux kernel
[N] sys-kernel/vserver-sources ((~)2.3.0.36.30.4-r1(2.3.0.36.30.4-r1)): Full sources including Gentoo and Linux-VServer patchsets for the 2.6 kernel tree.
[N] sys-kernel/xbox-sources (--): Full sources for the Xbox Linux kernel
[N] sys-kernel/xen-sources ((~)2.6.34(2.6.34)): Full sources for a dom0/domU Linux kernel to run under Xen
[N] sys-kernel/zen-sources ((~)2.6.34_p1-r1(2.6.34_p1-r1)): The Zen Kernel Live Sources
Those are my 17 options in gentoo's main tree, I'm sure i can find others in some overlays. Thats not counting things like following one of the dev branches of kernel. In fact just about anyone can have a branch these days, `git branch myawesomebranch`.
All of the above was encrypted with a Quad ROT-13 method. Unauthorized decryption is in violation of the DMCA.
> Linux is his baby and he's a jealous parent.
I'll take your assertions with a (big) pinch of salt: remember Linus *created* a distributed version management tool (git) when he couldn't use anymore BitKeeper..
And the nice thing about DVCS is that anybody can have his own tree..
So yes, Linus is the ultimate authority about what goes in his own tree, but this is quite normal..
Linus is monolithic.
Do you even know who Jonathan Corbet is? Among other things, he created LWN.net, has been a Linux kernel contributor for longer than that, and has written books on Linux kernel development (for example, the O'Reilly "Linux Device Drivers" book).
He's been on the inside for a long time. This is an opinion you should at least respect, even if in the end you disagree.
Most of the bitching and moaning about Linus seems to be along the lines of "he didn't accept my patch" or "he tore me a new one for suggesting something", not "Linux sucks, I'm going to use HURD or FreeBSD instead". And that's an important distinction, because Linus' primary goal is to make Linux and its codebase as awesome as it can be, not stroke developer's egos.
So yeah, if you write up something that you think is a great memory management scheme that Linus decides isn't the best approach, you're going to be pissed at him, because you thought very carefully about it and worked very hard to create a patch. But that doesn't mean Linus is necessarily wrong, and also doesn't mean he's arrogant - it means he thinks there's better choices available. He's picking not from the best that you can come up with, but from the best that the much larger set of people who've ever considered this problem can come up with.
I am officially gone from
A couple of quick points:
The Linux kernel is open source. Anyone who thinks they can do better can just clone it in git and start their own fork. You don't have to replace Linus, you can just be your own kernel maintainer. There's no part of the mainline Linux kernel development that takes place in private, so you can even "play Linus" and just merge only the patches that you like from the kernel mailing list into your own personal tree.
The kernel that Linus releases is not meant to be used directly by end users. Distributions are responsible for integrating the kernel into their operating systems as they see fit. They can choose to track Linus' tree closely or not at all. Red Hat, for example, rolls their own kernel that bears little resemblance to any of Linus'.
Linus' tree is widely regarded as the official Linux kernel mainly because he invented it and has stuck to his vision of how the kernel should be developed over the past 18 years or so. Most of the top developers and open source companies trust Linus and his management over the mainline kernel. Many have been around from the very beginning. Suggesting that they would "dump" Linus as the core maintainer is outright laughable.
\
Perhaps on a related note, I was thinking that while Linus has obviously scaled vertically to his full extent, there is still plenty of room for him to scale horizontally.
Con? Is that you?
Slashdot: Where people pretend to be twice as smart as they really are by behaving like children.
Arrogant people who achieve power never give it up voluntarily. They hold onto every little bit of it for dear life. Torvalds would no more voluntarily give up his ultimate authority than he would jump off a cliff. You can make all the reasonable arguments in the world, it's not going to change who he is. Linux is his baby and he's a jealous parent.
Interestingly, humble but smart people would end up in the same situation : they know that arrogant and power-hungry people are there and want power for the sake of their ego. I don't know if Linus is humble or arrogant, but he gave up power a long time ago when he put his OS under an open-source licence. He has never hidden the fact that he was a "benevolent dictator" (some even say the expression comes from his second surname : Benedict). If Linus is a bottleneck and slows down kernel development, there will be a fork. Right now, as much as people say he is a problem, he is still the only solution available.
What is good about open-source is that you can say to power-hungry people "Want to be the boss of a team ? Well go find a team that will respect your work !".
The Wise adapts himself to the world. The Fool adapts the world to himself. Therefore, all progress depends on the Fool.
Actually, thinking about it logically I suppose the actual meaning of the structure does in fact simply equate to "if". When when is true, if will be true, and vice versa.
if ($if_) { $when = true; }
if ($when) { $if_ = true; }
if ($if_ && $when) { return true; }
# better written as
if ($if_ | when ) { return true; }
So the "and when" logically equates to "or when", both of which are completely pointless additions to the if, but the "and when" still somehow manages to make it sound like the speaker or writer believes that the situation is actually inevitable, something they've only just realised after saying "if"..
which is totally what she said
I was originally going to write "UK English" but settled on English English as a joke. It's just as correct as "American English".
Fair questions.. let's ask a guy who is Scottish and lives in Scotland, shall we?
Me: Hi, Somersault!
Me: Hello there, you handsome devil!
Me: Yeah, whatever. What do you think of all this "English English" stuff you were saying earlier?
Me: I find it amusing.
Me: Oh, I see!
Me: Yep.
Me: Well, bye!
Me: Bye.
which is totally what she said
Who cares? If Linus stops updating his repository tomorrow, we'll all just switch to whatever repository meets our needs.
It's only consensus that says that Linus' repository is the "official" one.
There are already plenty of people who track Andrew Morton's repository instead of Linus', so if Linus went away, it's not like we don't already have a tested mechanism to allow us to track "unofficial" repositories.
*sigh* back to work...
I'm a long-shot kind of guy. So I took "Eaten by a bear in 2021" in the Torvalds death pool. Sure it's unlikely, but the odds make the payout HUGE if it happens.
SJW: Someone who has run out of real oppression, and has to fake it.
This is true of programmers ... or humanity ... in general.
Mainframe/UNIX Bit Twiddler and long time Windows/Linux Hobbyist.
The Theorem Theorem: If If, Then Then.
Then 'when' part tells you that you mean to wait for the event to happen, rather than act beforehand in anticipation of the event.
Not like any of this matters much.
<cough>schedulers</cough>
Wow! Some people just can't win, can they? Get involved in big charity work, you're just feeding your ego. Anyway, those were some pretty strong words about Linus that you kicked off this discussion with. I take it you either know him personally or have worked with him fairly closely?
Aide-toi, le Ciel t'aidera - Jeanne D'Arc.
And yet some of this behavior is required to keep things going as well as it has. Just imagine a day in his life, a mix of technical issues, coordination, flames and people wanting attention because they think their idea is the best. That's assuming you ignore personal issues and demands for time.
I don't have anywhere near his skills / reputation / influence / pressure, and the days when I have to fend off multiple groups of people demanding time, technical issues, and balance that with my schedule are draining for me. Some days I get to relax with hands-on physical work instead of managing systems, even if it's crimping cables, moving equipment or rearranging my office. Having his responsibility and maintaining his pace long-term would probably (understandably?) increase my self-worth (ego) and frustration to the point that I'd be seen in a negative way.
When you work in a high-performance way you can become very intolerant of others' shortcomings because you just assume that if you can do something that they can do it too, and if you "get it", then they of course should be able to comprehend it too. The limitations of language and slow communication speed are bothersome, you want to do a brain dump but teaching one thing requires a host of background on other issues so you have to start from the beginning, which takes time and effort you probably don't want to spend.
So perhaps with his imperfections he is the best person for the job at this time, and perhaps that will be delegated more to others in the future. All we know is that at some point passing the baton will be required and hope for a successful switchover.
The same can be said of influential people in various sectors of industry. What happens with Apple when Steve Jobs is no longer there? Microsoft after Bill Gates (I know he's out of the picture somewhat, but really he's not totally out of the loop yet). Samba team after Jeremy Allison. There are so many more influential people in huge roles that I see this as part of the necessary burden of being so good at what you do that others pale in comparison. Future generations will have to sort it out when these people leave, and hopefully they're grooming the apprentices so they can be ready when needed.
Because it doesn't happen. It's an extremely rare person that ends up the head of an organization without being a psychopath. Ever notice the behavior of CEOs? Justifying the ridiculously huge pay packages because other CEOs get it? And worse are the ones that demand it even when they're not anywhere near average. Most well adjusted people just don't want to be the head, the higher up you go past the low levels of management the worse it gets.
Any fork would either immediately or very quickly suffer from the same fate.
People like Alan Cox and Marcelo Tosatti have maintained well-established patchsets of their own, comprised of the features they happen to think are important. This is part of the natural order of things ("managing senior programmers is like herding cats" - Dave Platt) and not to be discouraged.
Linus' Tree is *still* the official one not because of Linus, he could be replaced overnight and it would continue, it's because all the main contributors submit their patches to it, and the official kernel group analyse them before they are integrated, the all Linus actually needs to do is be the one person who actually commits patches (so there are no conflicts) and act as a final arbiter in disputes
As the final arbiter it does not matter if he is arbitary, egotistical etc ... as long as he only acts as the final arbiter, the majority rules, he just need to decide when opinion is split ...
Puteulanus fenestra mortis
Linus is at the top now because he does a very good job and people trust him. The actual development is done by thousands of developers (around 3000 contributors / release currently), number increasing. It sales just fine.
The way he is accomplishing this, is by using a network of trust (he talk about it in his talk about git).
This is very scalable, as he is not actually checking out every peace of code, he just merges them.
What would happen if he would suddenly go crazy or hit by a bus? The answer is simple: one of the core maintainers, like for instance Andrew Morton would take over the position. General development would continue as it is now, as Linus talked often about how and why he runs things the way he does, and many people agree with him there.
Which sounds exactly like a scalability problem.
The Tao of math: The numbers you can count are not the real numbers.
He may not be perfect, but he is by far the best "leader" (whether project or corporate) that I've *ever* seen. He may be insulting, but he does it to everyone to scare off the timid and to make people reconsider their ideas. He is extremely technically competent and humble, two traits which are invaluable in a technical manager. Something you may want to read: Linux Kernel Management Style.
As for taking ideas and giving them to "favored" developers: 1) at least he admits when he is wrong and 2) why do you think those developers are favored? Ideas are a dime a dozen; implementing them well (and in a way that fits in with the Linux kernel) is the hard part.
Nathan's blog
I firmly believe that Linus is in the "humble but smart" camp. In the time I spent doing kernel work and on LKML, he was nothing if gregarious and insightful. Sure, he would call people idiots, but he did that to everyone, with a smirk, and generally only when they deserved it. These people that claim Linus is some Machiavellian Bill Gates or Steve Jobs clone don't know WTF they are talking about. They probably had some hare-brained idea that they posted to LKML and someone (not necessarily Linus) told them it was stupid and why it was stupid, but all they heard was the part that bruised their fragile little feelings.
This is the second part most of the haters don't get, and why I firmly believe Linus is still doing a good job: people still listen to Linus and track his repo. If he was truly the asshole people make him out to be, people would stop working with him
Nathan's blog
Commit it to git, surely.
http://michaelsmith.id.au