| |||||||||||||
ELC: Morton and Saxena on working with the kernel communityELC: Morton and Saxena on working with the kernel communityPosted Apr 21, 2008 18:06 UTC (Mon) by Frej (subscriber, #4165)Parent article: ELC: Morton and Saxena on working with the kernel community
Sorry for this highly offtopic/political and nontechnical reply. Please don't hate me ;). >The decision about when to merge a new feature is hard for some to understand. Many consider Linux a dictatorship, which is incorrect, it is instead "a democracy that doesn't vote". The merge decision is made on the model of the "rule of law" with kernel hackers playing the role of judges. Unfortunately, there are few written rules. I'm not a native english speaker, so i might be reading it wrong. But the above paragraph does seem a bit "loaded" (In lack of a better word). Don't claim that some thing is hard to understand "for some". That's arrogant? Especially since the author seems to understand, from the following sentence(s): >Many consider linux a dictatorship, which is incorrect, it is a "democracy that doesn't vote". That's a phrase coming from a dictator defending himself ;). At the least the argument "that's incorrect" certainly doesn't make it easy to understand. >The merge decision is made on the model of the "rule of law" with kernel hackers playing the role of judges. Unfortunately, there are few written rules. A democracy with no voting and judges that have no rules. Don't claim the linux model is a democracy then, it obviously isn't. Democracy should more or less be equivalent to "Majority rules" - when you don't vote how can majority rule? Explain why the model works for linux instead ;). Being a dictatorship does not have to be bad. As long as people are free to go their own way and leave(fork) it counters the arbitrary power of maintainers for the specific project(with name). Even though the consensus is that linus is a benevolent dictator. Which is my claim, google is my reference ;). Maybe it's more of an an oligarchy today with many more maintainers.
(Log in to post comments)
ELC: Morton and Saxena on working with the kernel community Posted Apr 21, 2008 18:34 UTC (Mon) by darwish07 (subscriber, #49520) [Link] >A democracy with no voting and judges that have no rules. Don't claim the linux model is a democracy then, it obviously isn't. Democracy should more or less be equivalent to "Majority rules" - when you don't vote how can majority rule? Explain why the model works for linux instead ;). No it's a democracy to the right limit. i.e. you can post a feature, even if Linus himself said OK, and another developer found something bad in your code, Andrew or subsystem maintainer won't merge the code till it has zero objections. It's just hidden democracy. i.e., there's no formal process but there's informal way which is more pragmatic and solve the 'merge/not-merge' problem quite well (in most cases).
Flawed democracy Posted Apr 30, 2008 18:07 UTC (Wed) by man_ls (subscriber, #15091) [Link] That doesn't look like a true democracy to me. For me the ultimate test to check if something is a democracy is: can a subverter (e.g. someone like Fidel Castro) do it and claim that it is still a democracy? In this case the kernel doesn't pass muster; Castro might subvert a system without written rules.Not that it is necessarily a bad thing; as others have said, as long as anyone can elect to be out of the process it has to be a fair process or languish. That is better than a democracy IMHO.
Linux does past your test Posted May 2, 2008 5:17 UTC (Fri) by anandsr21 (guest, #28562) [Link] Even Linus cannot subvert the process. His request can be turned down by the maintainer, and if he tried to force the kernel will fork. Even if Linus, Morton and the subsystem maintainer try together to hoodwink the whole community, it will not be easy to prevent a fork.
Depends on how you define it Posted May 2, 2008 10:22 UTC (Fri) by man_ls (subscriber, #15091) [Link] Think about it as "gaming the system", OK? Maintainers are after all elected by Linus himself, without any external oversight, so you can assume they are of one mind. And the same goes for Morton. They could stage all a revolt against Linus but the outcome would not be sure either way.Forking as you say is hard to prevent, but IMHO it is out of the question; in this case people are getting out of the system. It is like people fleeing a tyrannic country, which may or may not be prevented by the authorities (remember when Castro encouraged balseros to go and "invade" Miami). Of course in this case it is the crucial point, so maybe we are splitting hairs here.
ELC: Morton and Saxena on working with the kernel community Posted Apr 21, 2008 18:38 UTC (Mon) by darwish07 (subscriber, #49520) [Link] > Even though the consensus is that linus is a benevolent dictator. Which is my claim, google is my reference ;). Maybe it's more of an an oligarchy today with many more maintainers. In any project with large number of humans, there *must* be a final authority. Without that authority, people will just go in endless debates and *stop* doing the useful work. There's a fine line between dictatorship and 'final authority' but I think Linus is walking on that line very well.
ELC: Morton and Saxena on working with the kernel community Posted Apr 21, 2008 21:19 UTC (Mon) by mattmelton (subscriber, #34842) [Link] Lord Hailsham would have called it an "elective dictatorship" (the form of the British constitution) -- but lets not get too constitutional because it requires some study or hell will break loose on convention vs written.
Yup. The only working model Posted Apr 22, 2008 0:15 UTC (Tue) by khim (subscriber, #9252) [Link] A democracy with no voting and judges that have no rules. Don't claim the linux model is a democracy then, it obviously isn't. The linux model is obviously democracy. You don't even need 50% of votes to fork the kernel: 10-15% of developers will be enough. See: FreeBSD, NetBSD, OpenBSD, etc. So far Linux does not have significant forks - that means that system works. The fact that there are no known rules is advantage, not weakness. If you have written rules you have a way to game the system, but if the only way to get anything done is rough consensus it's much harder. Take a look on Microsoft's attack on two systems: IETF (no strict rules) and ISO (rigid structure with a lot of bureaucracy). First attack failed: after four years SenderID is still "an Experimental Protocol for the Internet community" and not "an Internet standard of any kind". Second attack succeded: now we have this huge mess called ISO/IEC 29500 standard. Rules are means to speed up decision process, nothing more - when they are allowed to work as substitute for common sense bad things happens.
Forking Linux? Fugettaboutit! Posted Apr 22, 2008 1:15 UTC (Tue) by pr1268 (subscriber, #24648) [Link] You don't even need 50% of votes to fork the kernel: 10-15% of developers will be enough. I just can't see the Linux kernel forking anytime soon. As long as Linus still maintains a presence, it would be impossible from a marketing perspective (not to mention a complete loss of credibility) to fork the kernel. After all, Linus wouldn't go "both ways" and actively maintain divergent forks, and having the "Kernel supported by Linus himself" as a sales/market pitch would certainly do little for the competing fork. Not that it couldn't happen legally (I'm thinking GPLv2 here), nor that it hasn't happened before (the aforementioned BSDs). As for the parent comment by Frej, please understand that regardless of any "loaded" comments in the article with respect to how Linux is managed, the management style works because (1) Linus plays a big part of kernel development, even after almost 17 years of hacking Linux; (2) he is unusually skilled and talented at many aspects of operating system theory and kernel hacking, software engineering, and, most importantly, people skills (arguably the most difficult of these), and (3) he commands a great deal of respect because of (1) and (2). Whatever you (or the article) call it, the management style is what it is due to 17 years of continuous development. And, besides, ELC keynote speaker Andrew Morton works with Linus on a daily basis, so I suppose they know each other quite well. Forgive me for all the Linus "gushing"; I've done it here on LWN.net before. But I still firmly believe that Linux owes much of its success to Linus' positive attributes mentioned above.
GPL ensues democracy, Linus make it work Posted Apr 22, 2008 6:27 UTC (Tue) by khim (subscriber, #9252) [Link] It's totally different question: why the democracy work. GPL ensues democracy - with or without Linus. Since you can not create proprietary fork the only reason to fork is if Linus makes bad decisions (BSD license, in comparison, allows democracy, not ensues it). And if Linus will make bad decisions on regular basis you'll see that this "Kernel supported by Linus himself" will quickly become moot point (see gcc/ecgs story, for example: forks like pgcc were frowned upon and rarely used till a lot of developers "voted" against FSF's rule). Now, it's true that Linus is enforcing some decisions - but so do "real" democracies (have you seen any where all decisions are made by votes of everyone?). They have government for everyday's work and even fundamental things like taxes are rarely voted by the "population at large". Free software is a democratic institution because everyone have exactly one "vote" - it's just that this vote must be cast with your feet, not with your ballot. Unlike real life (where you'll be forced to change lifestyle and your spouse and kids will suffer) "vote with your feet" in virtual world only affects your work, nothing else - so it's proper vote. Proprietary software is, of course, planned economy: what the management says - goes. And as usual when resources are scarce and goal is clear planned economy work best (recall 20th years of last century in USSR and US) but when you have enough resources to try different approaches and don't have a clear goal... take a look on Windows Vista and Linux today...
Yup. The only working model Posted Apr 22, 2008 13:13 UTC (Tue) by Frej (subscriber, #4165) [Link] >The linux model is obviously democracy. You don't even need 50% of votes to fork the kernel: ... So far Linux does not have significant forks - that means that system works. Forking is individual freedom and not a form of government. Actually a key part of democracy is that "majority rules", which is obviously >50%. I agree it's important that the possibility of forking exists, as i wrote "...As long as people are free to go their own way and leave(fork), it counters the arbitrary power of maintainers for the specific project(with name)." >The fact that there are no known rules is advantage, not weakness. If you have written rules you have a way to game the system, but if the only way to get anything done is rough consensus it's much harder. Take a look on Microsoft's attack on two systems: IETF (no strict rules) and ISO (rigid structure with a lot of bureaucracy). First attack failed: after four years SenderID is still "an Experimental Protocol for the Internet community" and not "an Internet standard of any kind". Second attack succeded: now we have this huge mess called ISO/IEC 29500 standard. Yes rules is always an complication. I don't disagree with this. The linux model has obviously worked. But that doesn't make it a democracy. My gripe was about the use of democracy as the true solution for all problems. If that holds, then for linux to be a solution, we must argue that it is some kind of democracy.
Forks are out there Posted Apr 30, 2008 18:27 UTC (Wed) by man_ls (subscriber, #15091) [Link] There are a lot of forks that have existed for a long time (-ac, -aa), and some are still out there (-rt, -mm). Apart from explicit forks (called "trees", but diverging in substantial ways) there are distribution kernels, which are often maintained in a parallel state. And we can assume that there are large numbers of private forks all over, either experimental, outdated or temporary, but these are not "significant" as in your message.If Linus went crazy and stopped merging new drivers while including OS/2 Warp code, people would probably switch to an externally maintained kernel. There must be excellent maintainers out there: every distro has one or more. This is not really a democracy except etymologically: power by the people, for the people.
ELC: Morton and Saxena on working with the kernel community Posted Apr 22, 2008 2:54 UTC (Tue) by fjf33 (subscriber, #5768) [Link] I think the problem is that we have created this dichotomy between democracy (good) and everything else (bad) so we cannot apply some of the other terms that are probably closer to the reality of the kernel development. To me it looks more like an aristocracy where the best and the brightest have absolute control based on internally developed conventions and pressures. Their power comes from 'knowing' what is best and right, which is the main reason why they don't abuse their powers. The threat of a peasant revolt (fork?) is what keeps them inline plus the ego massage from belonging to the elite group. A pure democratic method would probably produce a much lower quality product in the sense that it would tend to converge on a boring consensus kernel that tries to be something to everybody without pleasing anyone.
ELC: Morton and Saxena on working with the kernel community Posted Apr 22, 2008 13:33 UTC (Tue) by Frej (subscriber, #4165) [Link] >I think the problem is that we have created this dichotomy between democracy (good) and everything else (bad) so we cannot apply some of the other terms that are probably closer to the reality of the kernel development. Exactly what i wanted to get across - but much clearer than what wrote. Thank you! :)
ELC: Morton and Saxena on working with the kernel community Posted Apr 22, 2008 13:35 UTC (Tue) by Frej (subscriber, #4165) [Link] Exactly what i wanted to get across - but much clearer than what I wrote. Thank you! :)
ELC: Morton and Saxena on working with the kernel community Posted Apr 22, 2008 15:44 UTC (Tue) by madscientist (subscriber, #16861) [Link] > To me it looks more like an aristocracy where the best and the brightest > have absolute control based on internally developed conventions and > pressures. Their power comes from 'knowing' what is best and right, which > is the main reason why they don't abuse their powers. This is called a meritocracy. An aristocracy is where rule is hereditary.
ELC: Morton and Saxena on working with the kernel community Posted Apr 24, 2008 3:11 UTC (Thu) by lysse (guest, #3190) [Link] In fact, the word "aristocracy" derives from the Greek for "rule by the best". (It's also worth noting that the coiner of the term "meritocracy" - presumably "rule by the most deserving" - saw it as a *dystopian* ideal; and it doesn't take too much imagination to see that what starts out as meritocracy soon ossifies into aristocracy.)
Linux kernel: aristocracy? Posted Apr 25, 2008 3:40 UTC (Fri) by giraffedata (subscriber, #1954) [Link] In fact, the word "aristocracy" derives from the Greek for "rule by the best". Sure, but that doesn't mean the best individuals to rule. The original users could have meant rule by the class of people that is defined to be superior rulers or even superior people. In 1561 (which is apparently the earliest use of it), it was common to divide people into such classes, and it was usually done the same way species are: by bloodlines. I just looked to see what two American English dictionaries had to say about it, and one says solidly that it's a "best class," based on heredity. And the other says its either that or the best individuals.
ELC: Morton and Saxena on working with the kernel community Posted May 2, 2008 11:16 UTC (Fri) by dale77 (guest, #1490) [Link] Nice post. Lets try and avoid calling things by labels we like, as opposed to labels which actually apply.
ELC: Morton and Saxena on working with the kernel community Posted Apr 24, 2008 23:31 UTC (Thu) by aegl (subscriber, #37581) [Link] I think that Andrew's comments about a democracy that doesn't vote relates to the high rate of abstentions when a patch is posted.Everyone CAN vote by posting a reply to any patch submission with either a YES ("Acked-by:") or a NO (reason for not merging). But hardly anybody does. There are currently 4816 subscribers to the linux-kernel mailing list. Most "votes" are carried with 2 YES (proposer of the patch and 1 Ack) and 4814 abstentions. This is a good thing. A well thought out and obviously correct patch should get merged with minimal discussion. If everybody did start voting, the mailing list would grind to a halt (If every one of the 12,000+ commits that got us from 2.6.24 to 2.6.25 was discussed by all 4816 readers of LKML, then vger.kernel.org would melt under the load).
Linux kernel: democracy that doesn't vote Posted Apr 25, 2008 3:25 UTC (Fri) by giraffedata (subscriber, #1954) [Link] Are you claiming that every LKML subscriber has equal weight? I've never seen any rules for how the decision to merge works, but I'm pretty sure there's nobody counting up votes on the mailing list and accepting a patch if there are more yesses than noes.
Linux kernel: democracy that doesn't vote Posted Apr 25, 2008 4:14 UTC (Fri) by aegl (subscriber, #37581) [Link] No ... every vote doesn't have equal weight. People who have demonstrated expertise in a particular area will generally have far more influence on a merge decision than people without a track record, who in turn may have more weight than people who have publically demonstrated their lack of knowledge.But even Linus gets "outvoted" from time to time (he just merged kgdb even though he really doesn't like the idea of kernel debuggers). So the process is rather vague, and sometimes surprising ... but there is a general philosophy behind it. Maybe a merge decision can be summed up by answering the question: "Will this patch make the kernel better?" ... but this is still poorly specified unless you have the same opinion about what "better" means in the context of Linus' (and many of the other maintainers and key developers) goals. Some of the people who find it "hard to understand" the process are under the mistaken impression that the goal is commercial success. It isn't. The goal is to be the best kernel for its users. Commercial success might follow as a result of being good. But if you ask Linus how much he cares about the "market segment share" that Linux has, I bet his answer will be something like "not at all".
Linux kernel: democracy that doesn't vote Posted Apr 25, 2008 15:46 UTC (Fri) by giraffedata (subscriber, #1954) [Link] OK, that's not voting. Voting is a mathematical process; there are lots of other words for that: consensus, giving opinion, giving advice, etc.If merely stating your position on LKML could be called voting, then citizens that go before a city council meeting to argue against a new zoning ordinance would be said to be voting on the ordinance. Voting is the simplest way to implement democracy, and I think many people assume it's the only way, which is why Andrew feels the need to call it "democracy that doesn't vote."
|
|||||||||||||