Slashdot Mirror
Linux 'Awfully Cathedral-Like' - Java's a Bazaar
jg21 writes "LinuxWorld draws attention to a curious use of ESR's The Cathedral and the Bazaar by the Sun Microsystems exec who currently talks about Linux more than he does even about Java. Apparently Sun's President and COO Jonathan Schwartz said at a press briefing last week that Java with its JCP is more like ESR's Bazaar than Linux, which he dismissed as being "awfully cathedral-like" since Linus is the final arbiter (or Great Dictator), and not a committee." But be sure you don't mis-use the word Java in this Bazaar or the Mall Police will totally get you.
Sun is completely lacking in clue here, as always.
Sun, of course, feels heavily threatened by Linux and is merely spreading FUD in order to cement Sun's (TINY) market share and bolster Sun's (TINY) share price.
I have been an active member of the Linux community since its inception and we have been exorbitantly friendly to new users and developers. Sun, by contrast, makes you sign restrictive participatory agreements and agree to non-Free licences for community-owned code.
Sun is dead. Long live Linux.
I'm not Seth Finkelstein. I still speak the truth.
Anybody can fork the kernel. Most distributions do. Multiple threads of development happening independently versus everything having to go through a single party is what characterises the bazaar as separate from the cathedral, and this means that Linux is the epitome of the bazaar development process.
. . .anyone else could fork it off (from any point) and maintain it however they wanted.
More to the point many do, including major distros.
The whole idea that Linus dictates what goes in the kernel is utter bollocks, whereas Sun is infamous for maintaining the "true vision" and "purity" of Java.
Isn't their very argument against open sourcing Java that what happens to Linux would happen to it?
KFG
Cathedral vs Bazaar (redux)
Corp-SW is a cathedral; big, complex, rigid, takes AGES to make, intrinsically linked to a strict hierarchical power structure. Nominally direction and decision making comes from the top.
OSS is a bazaar; big, complex, but a collection of LOADS of different things, each able to do its thing its way. Nominally the whole system supports redundant/competing sections, anyone can stake themselves out a chunk and hawk their wares.
The original article, as I recall, goes off on one about some hippy-comune "noo-space" bollocks, and how projects garner their development effort via "coolness" and "reputation" (positing that you gain reputation in the bazaar by having either a really cool idea, or having successfully done something cool before (cool includes hard to do in the "noo-space")
Essentially the worthwile aspect of the article is describing the difference between "monolithic" system architecture and "collective" architecture in terms of project management (as in rather than monolithic computer systems)
I recall it actually bemoans that Linux as such has NO overall dictator of strategy. (Linus IIRC controls what goes in the Kernal, not whether a rival to the GIMP gets developed).
I haven't written much either, but then I don't describe myself as "one of the senior technical cadre that makes the Internet work"
Athletic Scholarships to universities make as much sense as academic scholarships to sports teams.
What I took away from the article (C vs B) was that in the Bazaar everyone WAS free to do what the hell they wanted, but that a process of pseudo natural-selection would starve out weak/crap work; either through not ever being used, or through not getting developent time.
This would not particularly require a single decision making point (individual or comittee), just time and community consensus.
That aside, I don't think the Bazaar was ever meant to apply to the _kernel_ but to the Linux / other OSS system as a whole.
Filesystems would, to my mind be the ideal example:
Cathedral(Windows); which version of OUR proprietary FS would you like?
Bazaar (Linux); which of these basically unrelated systems would you like?
There's actually two separate issues being discussed here. One is control over the reference. One is control over the implementation. Sun maintains control over the Java language reference. That's why they sued MS - because they extended the language in such a way that it broke things. Sun doesn't maintain control over all of the implementations. That's why you can get a Sun run time and an IBM run time and a GNU run time and an MS run time ...
Sun could turn the standard over to an independent committee. They don't want to do that. You can argue the merits (or lack thereof) of their position but that's a different conversation and isn't comparable to Linus' control of the kernel (which is arguably an implementation of the POSIX standard.)
"The legitimate powers of government extend only to such acts as are injurious to others." Thomas Jefferson.
Eric S. Raymond has just responded to Sun's Jonathan Schwartz and he says, among other things "any time [Sun] try to use my work to justify retaining proprietary control or argue that Linux is somehow less open, that's either culpable stupidity or dishonesty and they should expect to get kicked in the teeth for it by the entire open-source community, starting with me."
It's the Spanish Inquisition Sketch from Monty Python.
What happens if he gets hit by a bus?
Then one of the other major kernel developers will take it over. This has in fact already happened (control being turned over, not Linus getting hit by a bus).
Heck what happens today when large factions of kernel developers disagree with him? ( Kernel debugger )
If a large fraction of the kernel developers have a fundamental disagreement with Linus, then they'll fork the kernel and maintain their own. This in fact happens fairly regularly, as with the -mm kernels in 2.4 and the -ac kernels in 2.2. If the majority of the Linux community finds this new branch of the kernel more useful than Linus' branch, Linus will find his branch becoming marginalized. We've already seen this in GCC when EGCS forked off in response to frustration over lack of support for newer CPUs in GCC itself.
I suspect that this hasn't happened to the kernel because, to take the kernel debugger as an example, while lots of people may disagree with Linus none of them have an argument good enough to conclusively trump his arguments. The kernel debugger is a good one for me because while I tend to agree a kernel debugger would be a really useful thing, I also agree with Linus that it'd make it awful tempting for newer developers or those not intimately familiar with the subsystem they're bug-hunting in to find and fix only the immediate symptom and not do the rooting around in the code which eventually leads to an understanding of the underlying cause of the problem.
Microsoft removed RMI from their JVM (available as a separate download) which is what Sun used against them specifically. Extensions are allowed. For example Apple's JVM has extensions, but they don't get sued b/c they also implement the entire J2SE specification. Sun's interest is in having 100% Pure Java apps work everywhere, and Microsoft broke that by implementing a subset of the Java platform.
Also, the GNU Java runtime is doomed because of Sun patents on technologies in the J2SE specification. Read about the GNU Classpath project and Kaffe and you'll find that although they have made great progress, keeping up in the future is hindered by patent encumbrances. J2SE is not free and cannot be free for this reason.
Sun used to espouse "open" meaning proprietary implementations of an open standard, competing on quality (and presumably, extensions beyond the standard). That's a decent approach, but not viable if the "open" standard really has patents attached that cause clean-room implementations to be subject to patent infringement lawsuits.
Linus is very authoritative, and has yet to form an official public community/legal entity that develops and protects the kernel in the 10 year+ that he has been doing it.
Like OSDL? It doesn't get much more official than that.
What happens if he gets hit by a bus?
The succession has been arranged, but it's based on people, not organizations, since the Linux community is based on personal respect, not respect for an organization.
They actually pass through Redhat and SuSE's compilers. Not Linus'. Most of the mainstream distros have custom kernel patches and the like that aren't in Linus' vanilla kernel. It is still very much so a bazaar. Linus maintains a 'reference' version, and other people tweak with it as they see fit.
My blog. Good stuff (when I remember to update it). Read it.
This is the same guy who wants to redine words to suit his meaning -- even the term 'Open Source' itself:8 #rewriting_history_and_vocabulary/
http://blogs.sun.com/roller/page/jonathan/2004080
I still don't think it's appropriate to talk about any cathedral forming around Linus - he is a central figure as a matter of respect and skills (of which diplomacy is getting more important than technical skills), not by virtue of any other authority. Unlike the priest in a cathedral, which has authority from the church and can ignore the suggestions from the people in the bazaar outside entirely, if Linus does something unpopular he'll be thrown aside and ignored.
The recent issues with the XFree86 license has shown that WHEN someone running an open source project truly try to operate a cathedral inside the bazaar, their lack of externally imposed authority WILL cause people to turn away from them when they misstep.
There's always authority structures, some more formal than others, but what matters is where that authority derives from. If the authority is a sign of respect for the work you do, then that is not necessarily a sign of a cathedral in itself. Particularly not when others are blatantly ignoring your authority all around you (maintaining alternative trees).
In fact, Linus has distanced himself more and more from authority in some ways by letting others handle maintenance of older versions, by more or less encouraging distribution vendors from maintaining their own patchsets and not distributing Linus kernels unmodified at all, by encouraging many projects to keep developing their stuff outside mainline if he doesn't think it's suitable for his tree.
Linus ISN'T a strong authority for Linux in many ways - there's lots of stuff he refuses to put in HIS tree that ends up in lots of the alternative trees and in the distributions anyways. Linus is important because he's pragmatic enough that most sane stuff sooner or later does make it in, and most silly stuff stays out, and hence he's a good middle neutral ground from the people that are driving major components forwards.
Linus doesn't direct or control the direction of Linux development, he asserts some degree of authority over when major changes makes it into a kernel that acts as a common baseline and clearinghouse.
Look at the variations in the kernels distributed with major distributions today. The differences aren't just minor changes, but major components like file systems, realtime support etc.