5	1	0.007127	0.005337	0.000818	The Cathedral and the Bazaar
3	1	0.002659	0.002067	0.000642	Eric Steven Raymond
2	1	0.002450	0.001254	0.000377	Thyrsus Enterprises
4	0	0.007519	0.010269	0.000087	This is version 3.0
6	1	0.009085	0.043685	0.005934	Copyright © 2000 Eric S. Raymond
22	0	0.046469	1.579056	0.001172	Permission is granted to copy, distribute and/or modify this document under the terms of the Open Publication License, version 2.0.
1	1	0.001768	0.000436	0.000379	Abstract
28	0	0.154677	1.086724	0.000089	I anatomize a successful open-source project, fetchmail, that was run as a deliberate test of the surprising theories about software engineering suggested by the history of Linux.
29	1	0.039139	0.813800	0.144030	I discuss these theories in terms of two fundamentally different development styles, the ``cathedral'' model of most of the commercial world versus the ``bazaar'' model of the Linux world.
17	1	0.028103	0.737190	0.039621	I show that these models derive from opposing assumptions about the nature of the software-debugging task.
48	1	0.150897	50.067749	2.194525	I then make a sustained argument from the Linux experience for the proposition that ``Given enough eyeballs, all bugs are shallow'', suggest productive analogies with other self-correcting systems of selfish agents, and conclude with some exploration of the implications of this insight for the future of software.
5	1	0.005104	0.004490	0.000520	The Cathedral and the Bazaar
3	1	0.007613	0.005292	0.000346	Linux is subversive.
44	0	0.234736	294.434601	31.724010	Who would have thought even five years ago (1991) that a world-class operating system could coalesce as if by magic out of part-time hacking by several thousand developers scattered all over the planet, connected only by the tenuous strands of the Internet?
3	1	0.005206	0.002684	0.000361	Certainly not I.
26	1	0.047577	0.741004	0.013688	By the time Linux swam onto my radar screen in early 1993, I had already been involved in Unix and open-source development for ten years.
12	1	0.017389	0.182813	0.003257	I was one of the first GNU contributors in the mid-1980s.
35	1	0.225647	36.043491	9.958281	I had released a good deal of open-source software onto the net, developing or co-developing several programs (nethack, Emacs's VC and GUD modes, xlife, and others) that are still in wide use today.
8	1	0.035114	0.052787	0.002370	I thought I knew how it was done.
9	1	0.022077	0.035396	0.001908	Linux overturned much of what I thought I knew.
17	1	0.033045	0.129285	0.007046	I had been preaching the Unix gospel of small tools, rapid prototyping and evolutionary programming for years.
20	1	0.085694	0.450975	0.009904	But I also believed there was a certain critical complexity above which a more centralized, a priori approach was required.
47	0	0.420621	88.924141	0.000023	I believed that the most important software (operating systems and really large tools like the Emacs programming editor) needed to be built like cathedrals, carefully crafted by individual wizards or small bands of mages working in splendid isolation, with no beta to be released before its time.
27	0	0.098344	2.404486	0.000024	Linus Torvalds's style of development—release early and often, delegate everything you can, be open to the point of promiscuity—came as a surprise.
53	0	0.180014	46.461777	0.001906	No quiet, reverent cathedral-building here—rather, the Linux community seemed to resemble a great babbling bazaar of differing agendas and approaches (aptly symbolized by the Linux archive sites, who'd take submissions from anyone) out of which a coherent and stable system could seemingly emerge only by a succession of miracles.
17	0	0.042335	1.573397	0.000626	The fact that this bazaar style seemed to work, and work well, came as a distinct shock.
47	0	0.138113	139.984161	0.000023	As I learned my way around, I worked hard not just at individual projects, but also at trying to understand why the Linux world not only didn't fly apart in confusion but seemed to go from strength to strength at a speed barely imaginable to cathedral-builders.
10	1	0.029789	0.159596	0.004081	By mid-1996 I thought I was beginning to understand.
29	1	0.067459	3.871741	0.192140	Chance handed me a perfect way to test my theory, in the form of an open-source project that I could consciously try to run in the bazaar style.
10	1	0.022640	0.045939	0.002116	So I did—and it was a significant success.
7	1	0.010619	0.020981	0.000718	This is the story of that project.
13	0	0.018563	0.113434	0.000147	I'll use it to propose some aphorisms about effective open-source development.
25	0	0.048452	0.558493	0.000539	Not all of these are things I first learned in the Linux world, but we'll see how the Linux world gives them particular point.
35	0	0.081091	5.566368	0.002429	If I'm correct, they'll help you understand exactly what it is that makes the Linux community such a fountain of good software—and, perhaps, they will help you become more productive yourself.
5	0	0.013380	0.028803	0.000190	The Mail Must Get Through
26	0	0.069895	2.205941	0.002074	Since 1993 I'd been running the technical side of a small free-access Internet service provider called Chester County InterLink (CCIL) in West Chester, Pennsylvania.
21	1	0.047754	0.239788	0.008392	I co-founded CCIL and wrote our unique multiuser bulletin-board software—you can check it out by telnetting to locke.ccil.org.
10	0	0.012188	0.024621	0.000047	Today it supports almost three thousand users on thirty lines.
25	0	0.050081	0.333063	0.000023	The job allowed me 24-hour-a-day access to the net through CCIL's 56K line—in fact, the job practically demanded it!
9	1	0.033904	0.175878	0.004100	I had gotten quite used to instant Internet email.
14	1	0.041429	0.833437	0.017572	I found having to periodically telnet over to locke to check my mail annoying.
33	1	0.119218	22.921207	0.698325	What I wanted was for my mail to be delivered on snark (my home system) so that I would be notified when it arrived and could handle it using all my local tools.
40	1	0.270766	5.717487	0.196952	The Internet's native mail forwarding protocol, SMTP (Simple Mail Transfer Protocol), wouldn't suit, because it works best when machines are connected full-time, while my personal machine isn't always on the Internet, and doesn't have a static IP address.
24	1	0.062737	2.836792	0.347688	What I needed was a program that would reach out over my intermittent dialup connection and pull across my mail to be delivered locally.
20	1	0.057234	0.700884	0.050810	I knew such things existed, and that most of them used a simple application protocol called POP (Post Office Protocol).
25	1	0.094632	2.028427	0.052891	POP is now widely supported by most common mail clients, but at the time, it wasn't built in to the mail reader I was using.
5	1	0.008190	0.007303	0.000565	I needed a POP3 client.
10	1	0.024462	0.103680	0.005429	So I went out on the Internet and found one.
6	1	0.013834	0.036768	0.001289	Actually, I found three or four.
31	1	0.089087	2.895347	0.030360	I used one of them for a while, but it was missing what seemed an obvious feature, the ability to hack the addresses on fetched mail so replies would work properly.
14	1	0.038649	0.321760	0.004413	The problem was this: suppose someone named `joe' on locke sent me mail.
28	1	0.063753	8.099276	0.264280	If I fetched the mail to snark and then tried to reply to it, my mailer would cheerfully try to ship it to a nonexistent `joe' on snark.
15	1	0.041831	0.337544	0.010942	Hand-editing reply addresses to tack on <@ccil.org> quickly got to be a serious pain.
12	1	0.025882	0.085230	0.001605	This was clearly something the computer ought to be doing for me.
9	1	0.015751	0.047760	0.002536	But none of the existing POP clients knew how!
9	1	0.015795	0.047752	0.001540	And this brings us to the first lesson:
14	1	0.027696	0.165674	0.005301	1. Every good work of software starts by scratching a developer's personal itch.
37	0	0.089812	4.730495	0.001447	Perhaps this should have been obvious (it's long been proverbial that ``Necessity is the mother of invention'') but too often software developers spend their days grinding away for pay at programs they neither need nor love.
24	1	0.037249	1.074131	0.073229	But not in the Linux world—which may explain why the average quality of software originated in the Linux community is so high.
23	1	0.042152	0.271828	0.014883	So, did I immediately launch into a furious whirl of coding up a brand-new POP3 client to compete with the existing ones?
4	1	0.003872	0.003655	0.000709	Not on your life!
21	1	0.076104	1.098717	0.018183	I looked carefully at the POP utilities I had in hand, asking myself ``Which one is closest to what I want?''
2	1	0.002006	0.000670	0.000128	Because:
7	1	0.017144	0.040128	0.002848	2. Good programmers know what to write.
8	1	0.021749	0.061097	0.003759	Great ones know what to rewrite (and reuse).
14	1	0.024809	0.382705	0.011642	While I don't claim to be a great programmer, I try to imitate one.
10	1	0.015324	0.016387	0.000880	An important trait of the great ones is constructive laziness.
32	0	0.071840	13.530821	0.003568	They know that you get an A not for effort but for results, and that it's almost always easier to start from a good partial solution than from nothing at all.
12	1	0.027783	0.209597	0.009401	Linus Torvalds, for example, didn't actually try to write Linux from scratch.
19	1	0.054780	0.394961	0.016058	Instead, he started by reusing code and ideas from Minix, a tiny Unix-like operating system for PC clones.
28	0	0.137562	4.092783	0.000478	Eventually all the Minix code went away or was completely rewritten—but while it was there, it provided scaffolding for the infant that would eventually become Linux.
23	1	0.092283	3.554973	0.298366	In the same spirit, I went looking for an existing POP utility that was reasonably well coded, to use as a development base.
36	1	0.093145	1.072371	0.086126	The source-sharing tradition of the Unix world has always been friendly to code reuse (this is why the GNU project chose Unix as a base OS, in spite of serious reservations about the OS itself).
20	1	0.051909	0.078822	0.003335	The Linux world has taken this tradition nearly to its technological limit; it has terabytes of open sources generally available.
26	0	0.083345	2.140271	0.000741	So spending time looking for some else's almost-good-enough is more likely to give you good results in the Linux world than anywhere else.
5	1	0.008565	0.006588	0.000651	And it did for me.
29	0	0.145261	50.011833	0.003276	With those I'd found earlier, my second search made up a total of nine candidates—fetchpop, PopTart, get-mail, gwpop, pimp, pop-perl, popc, popmail and upop.
12	1	0.026674	0.087542	0.003533	The one I first settled on was `fetchpop' by Seung-Hong Oh.
21	1	0.068112	1.283791	0.013890	I put my header-rewrite feature in it, and made various other improvements which the author accepted into his 1.9 release.
21	1	0.050327	0.309204	0.005578	A few weeks later, though, I stumbled across the code for popclient by Carl Harris, and found I had a problem.
40	1	0.629215	2.338012	0.159124	Though fetchpop had some good original ideas in it (such as its background-daemon mode), it could only handle POP3 and was rather amateurishly coded (Seung-Hong was at that time a bright but inexperienced programmer, and both traits showed).
28	0	0.083704	0.286223	0.000023	Carl's code was better, quite professional and solid, but his program lacked several important and rather tricky-to-implement fetchpop features (including those I'd coded myself).
3	1	0.005006	0.004345	0.000604	Stay or switch?
21	0	0.053177	0.289487	0.000110	If I switched, I'd be throwing away the coding I'd already done in exchange for a better development base.
12	1	0.022185	0.033837	0.001306	A practical motive to switch was the presence of multiple-protocol support.
17	1	0.031480	0.243933	0.014447	POP3 is the most commonly used of the post-office server protocols, but not the only one.
39	1	0.320269	9.054539	2.095369	Fetchpop and the other competition didn't do POP2, RPOP, or APOP, and I was already having vague thoughts of perhaps adding IMAP (Internet Message Access Protocol, the most recently designed and most powerful post-office protocol) just for fun.
24	1	0.076950	23.790789	0.386378	But I had a more theoretical reason to think switching might be as good an idea as well, something I learned long before Linux.
9	1	0.020245	0.109567	0.008759	3. ``Plan to throw one away; you will, anyhow.''
8	0	0.016293	0.081227	0.000208	(Fred Brooks, The Mythical Man-Month, Chapter 11)
22	0	0.047768	0.251788	0.000606	Or, to put it another way, you often don't really understand the problem until after the first time you implement a solution.
11	1	0.025109	0.333267	0.011874	The second time, maybe you know enough to do it right.
16	1	0.045732	0.493517	0.007270	So if you want to get it right, be ready to start over at least once.
13	1	0.035817	0.173627	0.002324	Well (I told myself) the changes to fetchpop had been my first try.
3	1	0.010051	0.004549	0.000491	So I switched.
30	1	0.127445	11.117170	0.841408	After I sent my first set of popclient patches to Carl Harris on 25 June 1996, I found out that he had basically lost interest in popclient some time before.
11	1	0.026697	0.331062	0.007328	The code was a bit dusty, with minor bugs hanging out.
23	1	0.093514	0.224496	0.002749	I had many changes to make, and we quickly agreed that the logical thing for me to do was take over the program.
8	0	0.024640	0.033697	0.000085	Without my actually noticing, the project had escalated.
13	1	0.032014	0.158206	0.006191	No longer was I just contemplating minor patches to an existing POP client.
24	1	0.070843	1.814994	0.026895	I took on maintaining an entire one, and there were ideas bubbling in my head that I knew would probably lead to major changes.
18	1	0.038505	0.279812	0.020571	In a software culture that encourages code-sharing, this is a natural way for a project to evolve.
7	1	0.017253	0.070283	0.000965	I was acting out this principle:
12	1	0.024098	0.081132	0.002793	4. If you have the right attitude, interesting problems will find you.
9	1	0.021232	0.132114	0.002228	But Carl Harris's attitude was even more important.
3	0	0.003609	0.004435	0.000045	He understood that
22	1	0.040580	0.263045	0.016309	5. When you lose interest in a program, your last duty to it is to hand it off to a competent successor.
22	1	0.051862	0.207915	0.008657	Without ever having to discuss it, Carl and I knew we had a common goal of having the best solution out there.
20	1	0.047150	0.420356	0.006682	The only question for either of us was whether I could establish that I was a safe pair of hands.
10	1	0.019591	0.031974	0.001755	Once I did that, he acted with grace and dispatch.
12	1	0.037020	0.354464	0.016750	I hope I will do as well when it comes my turn.
5	1	0.011760	0.011750	0.001160	The Importance of Having Users
5	0	0.007542	0.006447	0.000053	And so I inherited popclient.
9	0	0.018442	0.010851	0.000019	Just as importantly, I inherited popclient's user base.
24	0	0.069830	1.478782	0.000459	Users are wonderful things to have, and not just because they demonstrate that you're serving a need, that you've done something right.
6	1	0.013666	0.016641	0.000866	Properly cultivated, they can become co-developers.
23	1	0.046555	0.427255	0.006812	Another strength of the Unix tradition, one that Linux pushes to a happy extreme, is that a lot of users are hackers too.
10	1	0.016244	0.038265	0.001733	Because source code is available, they can be effective hackers.
9	1	0.014910	0.080620	0.002317	This can be tremendously useful for shortening debugging time.
24	1	0.059223	1.428882	0.020917	Given a bit of encouragement, your users will diagnose problems, suggest fixes, and help improve the code far more quickly than you could unaided.
18	1	0.038177	0.262068	0.026717	6. Treating your users as co-developers is your least-hassle route to rapid code improvement and effective debugging.
9	1	0.016937	0.027568	0.001547	The power of this effect is easy to underestimate.
34	1	0.146535	6.020511	0.127494	In fact, pretty well all of us in the open-source world drastically underestimated how well it would scale up with number of users and against system complexity, until Linus Torvalds showed us differently.
29	1	0.100556	0.558240	0.006531	In fact, I think Linus's cleverest and most consequential hack was not the construction of the Linux kernel itself, but rather his invention of the Linux development model.
38	0	0.101109	13.881523	0.018289	When I expressed this opinion in his presence once, he smiled and quietly repeated something he has often said: ``I'm basically a very lazy person who likes to get credit for things other people actually do.''
4	1	0.005877	0.005795	0.000468	Lazy like a fox.
15	0	0.026100	0.100512	0.000471	Or, as Robert Heinlein famously wrote of one of his characters, too lazy to fail.
27	1	0.045722	0.972929	0.121569	In retrospect, one precedent for the methods and success of Linux can be seen in the development of the GNU Emacs Lisp library and Lisp code archives.
30	1	0.121849	0.658695	0.013917	In contrast to the cathedral-building style of the Emacs C core and most other GNU tools, the evolution of the Lisp code pool was fluid and very user-driven.
17	1	0.040536	3.616614	0.047720	Ideas and prototype modes were often rewritten three or four times before reaching a stable final form.
13	1	0.025271	0.084239	0.003371	And loosely-coupled collaborations enabled by the Internet, a la Linux, were frequent.
50	1	0.563616	3.042894	1.609654	Indeed, my own most successful single hack previous to fetchmail was probably Emacs VC (version control) mode, a Linux-like collaboration by email with three other people, only one of whom (Richard Stallman, the author of Emacs and founder of the Free Software Foundation) I have met to this day.
21	1	0.070379	1.423730	0.242710	It was a front-end for SCCS, RCS and later CVS from within Emacs that offered ``one-touch'' version control operations.
12	0	0.037172	0.054285	0.000091	It evolved from a tiny, crude sccs.el mode somebody else had written.
24	0	0.058996	1.707616	0.000665	And the development of VC succeeded because, unlike Emacs itself, Emacs Lisp code could go through release/test/improve generations very quickly.
6	1	0.012560	0.014920	0.000773	The Emacs story is not unique.
28	1	0.043570	0.269927	0.012025	There have been other software products with a two-level architecture and a two-tier user community that combined a cathedral-mode core and a bazaar-mode toolbox.
11	1	0.017088	0.066060	0.002393	One such is MATLAB, a commercial data-analysis and visualization tool.
39	1	0.115684	10.256047	0.049233	Users of MATLAB and other products with a similar structure invariably report that the action, the ferment, the innovation mostly takes place in the open part of the tool where a large and varied community can tinker with it.
4	1	0.010438	0.051874	0.002456	Release Early, Release Often
13	1	0.019286	0.038490	0.001492	Early and frequent releases are a critical part of the Linux development model.
37	1	0.105008	2.811294	0.173271	Most developers (including me) used to believe this was bad policy for larger than trivial projects, because early versions are almost by definition buggy versions and you don't want to wear out the patience of your users.
13	1	0.019830	0.083063	0.003684	This belief reinforced the general commitment to a cathedral-building style of development.
37	0	0.114315	51.687325	0.035662	If the overriding objective was for users to see as few bugs as possible, why then you'd only release a version every six months (or less often), and work like a dog on debugging between releases.
8	1	0.021084	0.018619	0.001043	The Emacs C core was developed this way.
36	1	0.329030	15.682634	0.174057	The Lisp library, in effect, was not—because there were active Lisp archives outside the FSF's control, where you could go to find new and development code versions independently of Emacs's release cycle.
25	1	0.057229	0.822180	0.087139	The most important of these, the Ohio State Emacs Lisp archive, anticipated the spirit and many of the features of today's big Linux archives.
33	1	0.113363	8.869771	0.327075	But few of us really thought very hard about what we were doing, or about what the very existence of that archive suggested about problems in the FSF's cathedral-building development model.
23	1	0.053703	8.085107	0.276384	I made one serious attempt around 1992 to get a lot of the Ohio code formally merged into the official Emacs Lisp library.
9	1	0.020602	0.042138	0.000983	I ran into political trouble and was largely unsuccessful.
23	1	0.071754	0.718000	0.044993	But by a year later, as Linux became widely visible, it was clear that something different and much healthier was going on there.
12	1	0.032101	0.084983	0.003023	Linus's open development policy was the very opposite of cathedral-building.
11	1	0.040110	0.136450	0.004259	Linux's Internet archives were burgeoning, multiple distributions were being floated.
15	1	0.041775	0.076218	0.001880	And all of this was driven by an unheard-of frequency of core system releases.
14	1	0.031153	0.315035	0.004713	Linus was treating his users as co-developers in the most effective possible way:
3	1	0.005886	0.011717	0.001719	7. Release early.
2	1	0.004652	0.002770	0.000530	Release often.
5	1	0.005684	0.007345	0.000889	And listen to your customers.
47	1	0.209917	63.145958	10.960053	Linus's innovation wasn't so much in doing quick-turnaround releases incorporating lots of user feedback (something like this had been Unix-world tradition for a long time), but in scaling it up to a level of intensity that matched the complexity of what he was developing.
21	1	0.041009	1.144919	0.011224	In those early times (around 1991) it wasn't unknown for him to release a new kernel more than once a day!
19	1	0.030406	0.135263	0.009555	Because he cultivated his base of co-developers and leveraged the Internet for collaboration harder than anyone else, this worked.
5	1	0.011390	0.006529	0.000699	But how did it work?
18	1	0.033794	0.065112	0.002599	And was it something I could duplicate, or did it rely on some unique genius of Linus Torvalds?
4	1	0.009999	0.007422	0.000526	I didn't think so.
7	1	0.027772	0.043771	0.001272	Granted, Linus is a damn fine hacker.
15	1	0.018757	0.075622	0.004736	How many of us could engineer an entire production-quality operating system kernel from scratch?
9	1	0.010383	0.022683	0.001154	But Linux didn't represent any awesome conceptual leap forward.
28	0	0.074241	0.203144	0.000056	Linus is not (or at least, not yet) an innovative genius of design in the way that, say, Richard Stallman or James Gosling (of NeWS and Java) are.
40	0	0.280789	2.344296	0.000023	Rather, Linus seems to me to be a genius of engineering and implementation, with a sixth sense for avoiding bugs and development dead-ends and a true knack for finding the minimum-effort path from point A to point B.
19	1	0.030045	0.424885	0.024787	Indeed, the whole design of Linux breathes this quality and mirrors Linus's essentially conservative and simplifying design approach.
33	0	0.149179	0.995340	0.000063	So, if rapid releases and leveraging the Internet medium to the hilt were not accidents but integral parts of Linus's engineering-genius insight into the minimum-effort path, what was he maximizing?
8	1	0.016024	0.023190	0.001263	What was he cranking out of the machinery?
7	1	0.024920	0.036450	0.001792	Put that way, the question answers itself.
37	1	0.102482	5.391131	0.395507	Linus was keeping his hacker/users constantly stimulated and rewarded—stimulated by the prospect of having an ego-satisfying piece of the action, rewarded by the sight of constant (even daily) improvement in their work.
36	1	0.095101	18.080654	3.182134	Linus was directly aiming to maximize the number of person-hours thrown at debugging and development, even at the possible cost of instability in the code and user-base burnout if any serious bug proved intractable.
11	1	0.031057	0.148257	0.002351	Linus was behaving as though he believed something like this:
23	1	0.054751	0.301521	0.009588	8. Given a large enough beta-tester and co-developer base, almost every problem will be characterized quickly and the fix obvious to someone.
10	0	0.029326	0.121103	0.000372	Or, less formally, ``Given enough eyeballs, all bugs are shallow.''
7	1	0.013332	0.026705	0.001720	I dub this: ``Linus's Law''.
12	1	0.022456	0.129622	0.004661	My original formulation was that every problem ``will be transparent to somebody''.
23	0	0.030266	0.272714	0.000022	Linus demurred that the person who understands and fixes the problem is not necessarily or even usually the person who first characterizes it.
11	1	0.017607	0.035023	0.002004	``Somebody finds the problem,'' he says, ``and somebody else understands it.
15	0	0.039601	0.218103	0.000295	And I'll go on record as saying that finding it is the bigger challenge.''
22	0	0.041905	0.195463	0.000063	That correction is important; we'll see how in the next section, when we examine the practice of debugging in more detail.
18	1	0.048282	0.256161	0.008352	But the key point is that both parts of the process (finding and fixing) tend to happen rapidly.
17	1	0.040943	0.251955	0.013492	In Linus's Law, I think, lies the core difference underlying the cathedral-builder and bazaar styles.
16	1	0.025102	0.242982	0.005445	In the cathedral-builder view of programming, bugs and development problems are tricky, insidious, deep phenomena.
19	0	0.034924	0.037792	0.000023	It takes months of scrutiny by a dedicated few to develop confidence that you've winkled them all out.
16	0	0.028099	0.310456	0.000447	Thus the long release intervals, and the inevitable disappointment when long-awaited releases are not perfect.
39	0	0.098339	4.472656	0.002013	In the bazaar view, on the other hand, you assume that bugs are generally shallow phenomena—or, at least, that they turn shallow pretty quickly when exposed to a thousand eager co-developers pounding on every single new release.
29	0	0.070668	6.172084	0.000683	Accordingly you release often in order to get more corrections, and as a beneficial side effect you have less to lose if an occasional botch gets out the door.
4	0	0.005747	0.005342	0.000083	And that's it.
3	0	0.005859	0.004794	0.000083	That's enough.
43	1	1.237083	6.304179	2.830688	If ``Linus's Law'' is false, then any system as complex as the Linux kernel, being hacked over by as many hands as that kernel was, should at some point have collapsed under the weight of unforeseen bad interactions and undiscovered ``deep'' bugs.
28	0	0.065635	1.563663	0.002809	If it's true, on the other hand, it is sufficient to explain Linux's relative lack of bugginess and its continuous uptimes spanning months or even years.
10	1	0.016966	0.044951	0.001194	Maybe it shouldn't have been such a surprise, at that.
38	1	0.123831	0.701834	0.010239	Sociologists years ago discovered that the averaged opinion of a mass of equally expert (or equally ignorant) observers is quite a bit more reliable a predictor than the opinion of a single randomly-chosen one of the observers.
6	1	0.015767	0.015253	0.000744	They called this the Delphi effect.
35	1	0.181849	7.050158	0.183766	It appears that what Linus has shown is that this applies even to debugging an operating system—that the Delphi effect can tame development complexity even at the complexity level of an OS kernel.
27	1	0.082451	0.387586	0.021793	One special feature of the Linux situation that clearly helps along the Delphi effect is the fact that the contributors for any given project are self-selected.
45	1	0.139674	194.007050	3.191810	An early respondent pointed out that contributions are received not from a random sample, but from people who are interested enough to use the software, learn about how it works, attempt to find solutions to problems they encounter, and actually produce an apparently reasonable fix.
15	1	0.034153	0.460832	0.016739	Anyone who passes all these filters is highly likely to have something useful to contribute.
10	1	0.031423	0.082380	0.002697	Linus's Law can be rephrased as ``Debugging is parallelizable''.
17	0	0.029853	0.115462	0.000127	Although debugging requires debuggers to communicate with some coordinating developer, it doesn't require significant coordination between debuggers.
18	1	0.039620	1.536210	0.062311	Thus it doesn't fall prey to the same quadratic complexity and management costs that make adding developers problematic.
25	1	0.040734	0.427887	0.012690	In practice, the theoretical loss of efficiency due to duplication of work by debuggers almost never seems to be an issue in the Linux world.
20	1	0.047677	0.367839	0.022963	One effect of a ``release early and often'' policy is to minimize such duplication by propagating fed-back fixes quickly.
39	1	0.155923	97.651489	13.102148	Brooks (the author of The Mythical Man-Month) even made an off-hand observation related to this: ``The total cost of maintaining a widely used program is typically 40 percent or more of the cost of developing it.
11	1	0.021266	0.072760	0.002859	Surprisingly this cost is strongly affected by the number of users.
5	1	0.009552	0.019361	0.001397	More users find more bugs.''
2	1	0.012152	0.003077	0.000687	[emphasis added].
17	1	0.032935	0.407822	0.015878	More users find more bugs because adding more users adds more different ways of stressing the program.
9	1	0.017390	0.047531	0.002099	This effect is amplified when the users are co-developers.
23	1	0.039358	0.376265	0.024987	Each one approaches the task of bug characterization with a slightly different perceptual set and analytical toolkit, a different angle on the problem.
11	1	0.016782	0.140603	0.005851	The ``Delphi effect'' seems to work precisely because of this variation.
15	1	0.021384	0.078679	0.001840	In the specific context of debugging, the variation also tends to reduce duplication of effort.
49	1	0.200821	70.393532	14.213466	So adding more beta-testers may not reduce the complexity of the current ``deepest'' bug from the developer's point of view, but it increases the probability that someone's toolkit will be matched to the problem in such a way that the bug is shallow to that person.
5	1	0.008143	0.015153	0.005026	Linus coppers his bets, too.
45	1	0.183442	214.528076	15.444814	In case there are serious bugs, Linux kernel version are numbered in such a way that potential users can make a choice either to run the last version designated ``stable'' or to ride the cutting edge and risk bugs in order to get new features.
27	1	0.059011	0.596700	0.011700	This tactic is not yet systematically imitated by most Linux hackers, but perhaps it should be; the fact that either choice is available makes both more attractive.
5	1	0.006291	0.036934	0.001262	How Many Eyeballs Tame Complexity
19	0	0.025111	0.327053	0.000474	It's one thing to observe in the large that the bazaar style greatly accelerates debugging and code evolution.
24	0	0.044360	0.370688	0.000244	It's another to understand exactly how and why it does so at the micro-level of day-to-day developer and tester behavior.
32	0	0.104136	1.331843	0.000126	In this section (written three years after the original paper, using insights by developers who read it and re-examined their own behavior) we'll take a hard look at the actual mechanisms.
11	0	0.017530	0.046847	0.000022	Non-technically inclined readers can safely skip to the next section.
30	1	0.082752	2.986279	0.145975	One key to understanding is to realize exactly why it is that the kind of bug report non–source-aware users normally turn in tends not to be very useful.
34	1	0.093195	2.116336	0.029696	Non–source-aware users tend to report only surface symptoms; they take their environment for granted, so they (a) omit critical background data, and (b) seldom include a reliable recipe for reproducing the bug.
35	1	0.092390	0.949984	0.073758	The underlying problem here is a mismatch between the tester's and the developer's mental models of the program; the tester, on the outside looking in, and the developer on the inside looking out.
24	0	0.060863	2.735229	0.000836	In closed-source development they're both stuck in these roles, and tend to talk past each other and find each other deeply frustrating.
31	0	0.069034	0.525179	0.000396	Open-source development breaks this bind, making it far easier for tester and developer to develop a shared representation grounded in the actual source code and to communicate effectively about it.
41	1	0.084013	28.012074	2.225027	Practically, there is a huge difference in leverage for the developer between the kind of bug report that just reports externally-visible symptoms and the kind that hooks directly to the developer's source-code–based mental representation of the program.
24	1	0.054824	0.212506	0.012413	Most bugs, most of the time, are easily nailed given even an incomplete but suggestive characterization of their error conditions at source-code level.
51	0	0.150990	58.063221	0.028037	When someone among your beta-testers can point out, "there's a boundary problem in line nnn", or even just "under conditions X, Y, and Z, this variable rolls over", a quick look at the offending code often suffices to pin down the exact mode of failure and generate a fix.
27	1	0.045081	0.347518	0.018260	Thus, source-code awareness by both parties greatly enhances both good communication and the synergy between what a beta-tester reports and what the core developer(s) know.
19	1	0.055616	2.186655	0.124464	In turn, this means that the core developers' time tends to be well conserved, even with many collaborators.
20	1	0.032573	0.081515	0.004015	Another characteristic of the open-source method that conserves developer time is the communication structure of typical open-source projects.
46	0	0.111726	0.222124	0.000024	Above I used the term "core developer"; this reflects a distinction between the project core (typically quite small; a single core developer is common, and one to three is typical) and the project halo of beta-testers and available contributors (which often numbers in the hundreds).
24	1	0.056054	0.666980	0.015024	The fundamental problem that traditional software-development organization addresses is Brook's Law: ``Adding more programmers to a late project makes it later.''
30	1	0.061591	1.796646	0.037575	More generally, Brooks's Law predicts that the complexity and communication costs of a project rise with the square of the number of developers, while work done only rises linearly.
42	1	0.095918	38.343689	6.797430	Brooks's Law is founded on experience that bugs tend strongly to cluster at the interfaces between code written by different people, and that communications/coordination overhead on a project tends to rise with the number of interfaces between human beings.
39	1	0.090376	2.496434	0.803693	Thus, problems scale with the number of communications paths between developers, which scales as the square of the number of developers (more precisely, according to the formula N*(N - 1)/2 where N is the number of developers).
39	1	0.109850	9.823833	0.434316	The Brooks's Law analysis (and the resulting fear of large numbers in development groups) rests on a hidden assummption: that the communications structure of the project is necessarily a complete graph, that everybody talks to everybody else.
48	0	0.158777	88.221222	0.000025	But on open-source projects, the halo developers work on what are in effect separable parallel subtasks and interact with each other very little; code changes and bug reports stream through the core group, and only within that small core group do we pay the full Brooksian overhead.
17	0	0.039975	1.002758	0.000023	There are are still more reasons that source-code–level bug reporting tends to be very efficient.
28	0	0.049474	1.762340	0.000545	They center around the fact that a single error can often have multiple possible symptoms, manifesting differently depending on details of the user's usage pattern and environment.
50	1	0.261754	72.556458	13.964704	Such errors tend to be exactly the sort of complex and subtle bugs (such as dynamic-memory-management errors or nondeterministic interrupt-window artifacts) that are hardest to reproduce at will or to pin down by static analysis, and which do the most to create long-term problems in software.
63	0	0.290888	135.287323	0.000025	A tester who sends in a tentative source-code–level characterization of such a multi-symptom bug (e.g. "It looks to me like there's a window in the signal handling near line 1250" or "Where are you zeroing that buffer?") may give a developer, otherwise too close to the code to see it, the critical clue to a half-dozen disparate symptoms.
33	0	0.127303	3.719486	0.002268	In cases like this, it may be hard or even impossible to know which externally-visible misbehaviour was caused by precisely which bug—but with frequent releases, it's unnecessary to know.
17	1	0.040470	0.364517	0.012432	Other collaborators will be likely to find out quickly whether their bug has been fixed or not.
22	1	0.067916	8.526412	0.840243	In many cases, source-level bug reports will cause misbehaviours to drop out without ever having been attributed to any specific fix.
19	1	0.033677	0.581798	0.044607	Complex multi-symptom errors also tend to have multiple trace paths from surface symptoms back to the actual bug.
33	0	0.075481	3.708883	0.000027	Which of the trace paths a given developer or tester can chase may depend on subtleties of that person's environment, and may well change in a not obviously deterministic way over time.
25	1	0.070927	1.588447	0.061683	In effect, each developer and tester samples a semi-random set of the program's state space when looking for the etiology of a symptom.
22	1	0.030074	0.551443	0.019786	The more subtle and complex the bug, the less likely that skill will be able to guarantee the relevance of that sample.
32	0	0.139644	0.410316	0.000033	For simple and easily reproducible bugs, then, the accent will be on the "semi" rather than the "random"; debugging skill and intimacy with the code and its architecture will matter a lot.
11	0	0.017681	0.054706	0.000022	But for complex bugs, the accent will be on the "random".
31	0	0.069431	2.459160	0.000025	Under these circumstances many people running traces will be much more effective than a few people running traces sequentially—even if the few have a much higher average skill level.
35	1	0.096824	4.060534	0.075783	This effect will be greatly amplified if the difficulty of following trace paths from different surface symptoms back to a bug varies significantly in a way that can't be predicted by looking at the symptoms.
25	1	0.052665	1.337626	0.070692	A single developer sampling those paths sequentially will be as likely to pick a difficult trace path on the first try as an easy one.
17	1	0.043068	0.888070	0.038414	On the other hand, suppose many people are trying trace paths in parallel while doing rapid releases.
22	1	0.039044	0.520684	0.011521	Then it is likely one of them will find the easiest path immediately, and nail the bug in a much shorter time.
36	1	0.212437	78.751366	7.224583	The project maintainer will see that, ship a new release, and the other people running traces on the same bug will be able to stop before having spent too much time on their more difficult traces .
7	1	0.016689	0.038964	0.001832	When Is a Rose Not a Rose?
33	1	0.251861	1.093895	0.071472	Having studied Linus's behavior and formed a theory about why it was successful, I made a conscious decision to test this theory on my new (admittedly much less complex and ambitious) project.
13	0	0.025958	0.023892	0.000022	But the first thing I did was reorganize and simplify popclient a lot.
19	1	0.047075	0.543053	0.010567	Carl Harris's implementation was very sound, but exhibited a kind of unnecessary complexity common to many C programmers.
15	1	0.021058	0.150072	0.006481	He treated the code as central and the data structures as support for the code.
28	1	0.055261	1.167365	0.015674	As a result, the code was beautiful but the data structure design ad-hoc and rather ugly (at least by the high standards of this veteran LISP hacker).
16	1	0.039724	1.018869	0.038738	I had another purpose for rewriting besides improving the code and the data structure design, however.
10	1	0.018711	0.078109	0.002676	That was to evolve it into something I understood completely.
16	0	0.027388	0.217631	0.000289	It's no fun to be responsible for fixing bugs in a program you don't understand.
19	1	0.048422	1.293495	0.017938	For the first month or so, then, I was simply following out the implications of Carl's basic design.
11	1	0.038423	0.111931	0.002447	The first serious change I made was to add IMAP support.
21	1	0.030905	0.852940	0.095291	I did this by reorganizing the protocol machines into a generic driver and three method tables (for POP2, POP3, and IMAP).
30	0	0.066670	5.880875	0.000328	This and the previous changes illustrate a general principle that's good for programmers to keep in mind, especially in languages like C that don't naturally do dynamic typing:
16	1	0.021893	0.692325	0.088311	9. Smart data structures and dumb code works a lot better than the other way around.
19	1	0.043385	0.398844	0.013999	Brooks, Chapter 9: ``Show me your flowchart and conceal your tables, and I shall continue to be mystified.
15	0	0.027263	0.046226	0.000035	Show me your tables, and I won't usually need your flowchart; it'll be obvious.''
14	0	0.025656	0.089107	0.000452	Allowing for thirty years of terminological/cultural shift, it's the same point.
33	1	0.101720	5.998852	0.959649	At this point (early September 1996, about six weeks from zero) I started thinking that a name change might be in order—after all, it wasn't just a POP client any more.
14	1	0.027038	0.123214	0.004046	But I hesitated, because there was as yet nothing genuinely new in the design.
13	1	0.023984	0.080098	0.002047	My version of popclient had yet to develop an identity of its own.
15	1	0.056588	0.839628	0.031705	That changed, radically, when popclient learned how to forward fetched mail to the SMTP port.
8	0	0.014768	0.034423	0.000160	I'll get to that in a moment.
25	0	0.078141	3.689323	0.004114	But first: I said earlier that I'd decided to use this project to test my theory about what Linus Torvalds had done right.
9	1	0.022760	0.046499	0.001522	How (you may well ask) did I do that?
4	1	0.005143	0.005471	0.000810	In these ways:
21	1	0.031949	0.408642	0.006541	I released early and often (almost never less often than every ten days; during periods of intense development, once a day).
15	1	0.021902	0.120962	0.004440	I grew my beta list by adding to it everyone who contacted me about fetchmail.
15	1	0.048382	0.523585	0.018945	I sent chatty announcements to the beta list whenever I released, encouraging people to participate.
22	1	0.038626	0.328294	0.016243	And I listened to my beta-testers, polling them about design decisions and stroking them whenever they sent in patches and feedback.
8	1	0.015636	0.014156	0.000679	The payoff from these simple measures was immediate.
23	1	0.056747	0.581128	0.025915	From the beginning of the project, I got bug reports of a quality most developers would kill for, often with good fixes attached.
13	0	0.050342	0.060237	0.000035	I got thoughtful criticism, I got fan mail, I got intelligent feature suggestions.
4	0	0.006129	0.004741	0.000015	Which leads to:
24	0	0.039427	0.151122	0.000073	10. If you treat your beta-testers as if they're your most valuable resource, they will respond by becoming your most valuable resource.
18	1	0.031448	0.263978	0.015464	One interesting measure of fetchmail's success is the sheer size of the project beta list, fetchmail-friends.
23	1	0.039603	0.608769	0.328145	At the time of latest revision of this paper (November 2000) it has 287 members and is adding two or three a week.
28	1	0.091127	2.644707	0.020384	Actually, when I revised in late May 1997 I found the list was beginning to lose members from its high of close to 300 for an interesting reason.
26	1	0.075535	1.636983	0.093728	Several people have asked me to unsubscribe them because fetchmail is working so well for them that they no longer need to see the list traffic!
15	1	0.021653	0.052648	0.002125	Perhaps this is part of the normal life-cycle of a mature bazaar-style project.
3	0	0.003488	0.001802	0.000042	Popclient becomes Fetchmail
26	1	0.092209	3.217245	0.104074	The real turning point in the project was when Harry Hochheiser sent me his scratch code for forwarding mail to the client machine's SMTP port.
22	1	0.040751	0.689436	0.018071	I realized almost immediately that a reliable implementation of this feature would make all the other mail delivery modes next to obsolete.
32	1	0.156427	32.755501	6.270876	For many weeks I had been tweaking fetchmail rather incrementally while feeling like the interface design was serviceable but grubby—inelegant and with too many exiguous options hanging out all over.
22	1	0.042561	4.392895	0.078087	The options to dump fetched mail to a mailbox file or standard output particularly bothered me, but I couldn't figure out why.
18	0	0.043330	0.510581	0.001803	(If you don't care about the technicalia of Internet mail, the next two paragraphs can be safely skipped.)
20	1	0.104814	4.754477	0.238287	What I saw when I thought about SMTP forwarding was that popclient had been trying to do too many things.
18	1	0.046531	0.208897	0.006096	It had been designed to be both a mail transport agent (MTA) and a local delivery agent (MDA).
29	1	0.135350	0.690891	0.011208	With SMTP forwarding, it could get out of the MDA business and be a pure MTA, handing off mail to other programs for local delivery just as sendmail does.
42	1	0.116944	40.716908	10.458251	Why mess with all the complexity of configuring a mail delivery agent or setting up lock-and-append on a mailbox when port 25 is almost guaranteed to be there on any platform with TCP/IP support in the first place?
23	1	0.104565	5.651106	0.312421	Especially when this means retrieved mail is guaranteed to look like normal sender-initiated SMTP mail, which is really what we want anyway.
6	0	0.013211	0.025642	0.000502	(Back to a higher level....)
15	1	0.024451	0.090676	0.003176	Even if you didn't follow the preceding technical jargon, there are several important lessons here.
20	1	0.103850	0.395832	0.011429	First, this SMTP-forwarding concept was the biggest single payoff I got from consciously trying to emulate Linus's methods.
17	1	0.045413	0.088179	0.002725	A user gave me this terrific idea—all I had to do was understand the implications.
16	1	0.040314	0.210879	0.009739	11. The next best thing to having good ideas is recognizing good ideas from your users.
5	0	0.009952	0.010495	0.000078	Sometimes the latter is better.
49	0	0.298890	1.796454	0.000032	Interestingly enough, you will quickly find that if you are completely and self-deprecatingly truthful about how much you owe other people, the world at large will treat you as though you did every bit of the invention yourself and are just being becomingly modest about your innate genius.
10	1	0.023574	0.121388	0.003009	We can all see how well this worked for Linus!
22	1	0.043605	2.876268	0.806263	(When I gave my talk at the first Perl Conference in August 1997, hacker extraordinaire Larry Wall was in the front row.
19	0	0.074150	0.647341	0.000023	As I got to the last line above he called out, religious-revival style, ``Tell it, tell it, brother!''.
16	0	0.041722	0.309277	0.000407	The whole audience laughed, because they knew this had worked for the inventor of Perl, too.)
34	1	0.071941	4.740320	0.302766	After a very few weeks of running the project in the same spirit, I began to get similar praise not just from my users but from other people to whom the word leaked out.
27	0	0.063123	0.597696	0.000037	I stashed away some of that email; I'll look at it again sometime if I ever start wondering whether my life has been worthwhile :-).
18	1	0.028722	0.390316	0.015982	But there are two more fundamental, non-political lessons here that are general to all kinds of design.
19	1	0.052818	0.358595	0.013922	12. Often, the most striking and innovative solutions come from realizing that your concept of the problem was wrong.
28	1	0.091525	3.582524	0.066701	I had been trying to solve the wrong problem by continuing to develop popclient as a combined MTA/MDA with all kinds of funky local delivery modes.
25	1	0.074786	1.005424	0.024071	Fetchmail's design needed to be rethought from the ground up as a pure MTA, a part of the normal SMTP-speaking Internet mail path.
43	0	0.149386	20.160866	0.029341	When you hit a wall in development—when you find yourself hard put to think past the next patch—it's often time to ask not whether you've got the right answer, but whether you're asking the right question.
7	1	0.020542	0.048986	0.003149	Perhaps the problem needs to be reframed.
6	1	0.020419	0.016068	0.001161	Well, I had reframed my problem.
43	0	0.288360	11.152893	0.000027	Clearly, the right thing to do was (1) hack SMTP forwarding support into the generic driver, (2) make it the default mode, and (3) eventually throw out all the other delivery modes, especially the deliver-to-file and deliver-to-standard-output options.
21	1	0.042706	1.474070	0.049653	I hesitated over step 3 for some time, fearing to upset long-time popclient users dependent on the alternate delivery mechanisms.
19	1	0.033272	0.525187	0.008687	In theory, they could immediately switch to .forward files or their non-sendmail equivalents to get the same effects.
8	1	0.012820	0.028916	0.001105	In practice the transition might have been messy.
9	1	0.018104	0.080693	0.006132	But when I did it, the benefits proved huge.
8	1	0.008502	0.017568	0.001184	The cruftiest parts of the driver code vanished.
28	1	0.089930	1.952273	0.206605	Configuration got radically simpler—no more grovelling around for the system MDA and user's mailbox, no more worries about whether the underlying OS supports file locking.
8	1	0.018649	0.048363	0.002429	Also, the only way to lose mail vanished.
16	1	0.055328	0.112764	0.002523	If you specified delivery to a file and the disk got full, your mail got lost.
26	1	0.059471	0.978743	0.027577	This can't happen with SMTP forwarding because your SMTP listener won't return OK unless the message can be delivered or at least spooled for later delivery.
14	0	0.036281	0.121159	0.000396	Also, performance improved (though not so you'd notice it in a single run).
16	1	0.039145	0.307122	0.015011	Another not insignificant benefit of this change was that the manual page got a lot simpler.
25	1	0.099538	4.954374	0.387548	Later, I had to bring delivery via a user-specified local MDA back in order to allow handling of some obscure situations involving dynamic SLIP.
10	1	0.024043	0.169833	0.002740	But I found a much simpler way to do it.
2	1	0.002374	0.000686	0.000215	The moral?
16	1	0.026859	0.572127	0.011741	Don't hesitate to throw away superannuated features when you can do it without loss of effectiveness.
21	1	0.086406	1.729382	0.146048	Antoine de Saint-Exupéry (who was an aviator and aircraft designer when he wasn't authoring classic children's books) said:
24	1	0.084018	1.516212	0.042429	13. ``Perfection (in design) is achieved not when there is nothing more to add, but rather when there is nothing more to take away.''
17	1	0.049642	2.744854	0.328515	When your code is getting both better and simpler, that is when you know it's right.
18	1	0.020556	0.057485	0.002892	And in the process, the fetchmail design acquired an identity of its own, different from the ancestral popclient.
7	1	0.016613	0.035397	0.000891	It was time for the name change.
31	1	0.131475	1.875720	0.094794	The new design looked much more like a dual of sendmail than the old popclient had; both are MTAs, but where sendmail pushes then delivers, the new popclient pulls then delivers.
10	1	0.021551	0.036333	0.002320	So, two months off the blocks, I renamed it fetchmail.
16	1	0.026875	0.163391	0.008294	There is a more general lesson in this story about how SMTP delivery came to fetchmail.
21	0	0.089739	0.507486	0.000024	It is not only debugging that is parallelizable; development and (to a perhaps surprising extent) exploration of design space is, too.
30	1	0.060917	4.175567	0.177169	When your development mode is rapidly iterative, development and enhancement may become special cases of debugging—fixing `bugs of omission' in the original capabilities or concept of the software.
26	0	0.052624	0.139958	0.000023	Even at a higher level of design, it can be very valuable to have lots of co-developers random-walking through the design space near your product.
31	1	0.068360	10.496246	1.285374	Consider the way a puddle of water finds a drain, or better yet how ants find food: exploration essentially by diffusion, followed by exploitation mediated by a scalable communication mechanism.
32	1	0.169634	10.858189	0.570918	This works very well; as with Harry Hochheiser and me, one of your outriders may well find a huge win nearby that you were just a little too close-focused to see.
3	1	0.005732	0.009104	0.000938	Fetchmail Grows Up
26	0	0.060056	1.300373	0.000260	There I was with a neat and innovative design, code that I knew worked well because I used it every day, and a burgeoning beta list.
26	1	0.050447	5.260732	0.083293	It gradually dawned on me that I was no longer engaged in a trivial personal hack that might happen to be useful to few other people.
23	1	0.045408	1.089231	0.051785	I had my hands on a program that every hacker with a Unix box and a SLIP/PPP mail connection really needs.
41	1	0.269284	4.429494	0.347780	With the SMTP forwarding feature, it pulled far enough in front of the competition to potentially become a ``category killer'', one of those classic programs that fills its niche so competently that the alternatives are not just discarded but almost forgotten.
13	1	0.020986	0.349505	0.011398	I think you can't really aim or plan for a result like this.
22	1	0.072236	1.526244	0.192868	You have to get pulled into it by design ideas so powerful that afterward the results just seem inevitable, natural, even foreordained.
37	1	0.129331	78.266479	17.109007	The only way to try for ideas like that is by having lots of ideas—or by having the engineering judgment to take other peoples' good ideas beyond where the originators thought they could go.
25	1	0.094452	5.037283	0.189819	Andy Tanenbaum had the original idea to build a simple native Unix for IBM PCs, for use as a teaching tool (he called it Minix).
21	1	0.073382	1.283534	0.027173	Linus Torvalds pushed the Minix concept further than Andrew probably thought it could go—and it grew into something wonderful.
23	1	0.062538	0.655051	0.027451	In the same way (though on a smaller scale), I took some ideas by Carl Harris and Harry Hochheiser and pushed them hard.
13	1	0.033425	0.119850	0.002951	Neither of us was `original' in the romantic way people think is genius.
19	1	0.038795	0.678377	0.074967	But then, most science and engineering and software development isn't done by original genius, hacker mythology to the contrary.
21	1	0.033391	0.153478	0.003350	The results were pretty heady stuff all the same—in fact, just the kind of success every hacker lives for!
12	1	0.035073	0.314763	0.004182	And they meant I would have to set my standards even higher.
36	0	0.116544	59.782837	0.015554	To make fetchmail as good as I now saw it could be, I'd have to write not just for my own needs, but also include and support features necessary to others but outside my orbit.
10	1	0.021166	0.087248	0.004641	And do that while keeping the program simple and robust.
44	1	0.227426	116.261696	8.295957	The first and overwhelmingly most important feature I wrote after realizing this was multidrop support—the ability to fetch mail from mailboxes that had accumulated all mail for a group of users, and then route each piece of mail to its individual recipients.
40	1	0.245938	84.829269	3.519824	I decided to add the multidrop support partly because some users were clamoring for it, but mostly because I thought it would shake bugs out of the single-drop code by forcing me to deal with addressing in full generality.
4	1	0.010434	0.009585	0.000865	And so it proved.
32	1	0.155114	0.887569	0.027479	Getting RFC 822 address parsing right took me a remarkably long time, not because any individual piece of it is hard but because it involved a pile of interdependent and fussy details.
13	1	0.051847	0.212744	0.006724	But multidrop addressing turned out to be an excellent design decision as well.
6	1	0.010900	0.013293	0.000754	Here's how I knew:
22	1	0.048120	0.136639	0.002973	14. Any tool should be useful in the expected way, but a truly great tool lends itself to uses you never expected.
27	1	0.202319	17.149027	0.400555	The unexpected use for multidrop fetchmail is to run mailing lists with the list kept, and alias expansion done, on the client side of the Internet connection.
25	1	0.054436	0.909675	0.068240	This means someone running a personal machine through an ISP account can manage a mailing list without continuing access to the ISP's alias files.
19	1	0.045470	0.307329	0.011685	Another important change demanded by my beta-testers was support for 8-bit MIME (Multipurpose Internet Mail Extensions) operation.
40	1	0.216284	31.915100	1.183881	This was pretty easy to do, because I had been careful to keep the code 8-bit clean (that is, to not press the 8th bit, unused in the ASCII character set, into service to carry information within the program).
17	1	0.028096	0.256576	0.015776	Not because I anticipated the demand for this feature, but rather in obedience to another rule:
31	1	0.083015	34.356567	2.628199	15. When writing gateway software of any kind, take pains to disturb the data stream as little as possible—and never throw away information unless the recipient forces you to!
16	1	0.042608	0.161051	0.003047	Had I not obeyed this rule, 8-bit MIME support would have been difficult and buggy.
24	1	0.110611	0.413035	0.020662	As it was, all I had to do is read the MIME standard (RFC 1652) and add a trivial bit of header-generation logic.
28	1	0.052631	1.714231	0.306632	Some European users bugged me into adding an option to limit the number of messages retrieved per session (so they can control costs from their expensive phone networks).
16	0	0.025517	0.151589	0.000103	I resisted this for a long time, and I'm still not entirely happy about it.
28	0	0.061800	1.191969	0.001716	But if you're writing for the world, you have to listen to your customers—this doesn't change just because they're not paying you in money.
6	1	0.009198	0.021358	0.002114	A Few More Lessons from Fetchmail
22	1	0.044147	0.661958	0.037908	Before we go back to general software-engineering issues, there are a couple more specific lessons from the fetchmail experience to ponder.
7	1	0.008116	0.024226	0.000859	Nontechnical readers can safely skip this section.
16	1	0.032508	0.100885	0.007123	The rc (control) file syntax includes optional `noise' keywords that are entirely ignored by the parser.
25	1	0.051522	0.811087	0.043299	The English-like syntax they allow is considerably more readable than the traditional terse keyword-value pairs you get when you strip them all out.
24	1	0.072927	2.079617	0.090485	These started out as a late-night experiment when I noticed how much the rc file declarations were beginning to resemble an imperative minilanguage.
13	1	0.030526	0.119689	0.011666	(This is also why I changed the original popclient ``server'' keyword to ``poll'').
20	1	0.064574	2.878491	0.087917	It seemed to me that trying to make that imperative minilanguage more like English might make it easier to use.
37	0	0.224185	0.172531	0.000116	Now, although I'm a convinced partisan of the ``make it a language'' school of design as exemplified by Emacs and HTML and many database engines, I am not normally a big fan of ``English-like'' syntaxes.
20	1	0.040680	1.391183	0.037138	Traditionally programmers have tended to favor control syntaxes that are very precise and compact and have no redundancy at all.
23	1	0.073282	0.688410	0.010143	This is a cultural legacy from when computing resources were expensive, so parsing stages had to be as cheap and simple as possible.
13	1	0.036565	0.189559	0.010339	English, with about 50% redundancy, looked like a very inappropriate model then.
19	1	0.036942	0.068268	0.003267	This is not my reason for normally avoiding English-like syntaxes; I mention it here only to demolish it.
13	0	0.015655	0.049458	0.000023	With cheap cycles and core, terseness should not be an end in itself.
20	0	0.030853	0.406682	0.000122	Nowadays it's more important for a language to be convenient for humans than to be cheap for the computer.
8	1	0.020884	0.175728	0.002272	There remain, however, good reasons to be wary.
32	0	0.058094	0.946954	0.000037	One is the complexity cost of the parsing stage—you don't want to raise that to the point where it's a significant source of bugs and user confusion in itself.
44	1	0.205452	6.475545	0.600585	Another is that trying to make a language syntax English-like often demands that the ``English'' it speaks be bent seriously out of shape, so much so that the superficial resemblance to natural language is as confusing as a traditional syntax would have been.
18	0	0.048546	0.221236	0.000483	(You see this bad effect in a lot of so-called ``fourth generation'' and commercial database-query languages.)
16	1	0.028105	0.184603	0.007740	The fetchmail control syntax seems to avoid these problems because the language domain is extremely restricted.
38	0	0.097112	1.697940	0.000352	It's nowhere near a general-purpose language; the things it says simply are not very complicated, so there's little potential for confusion in moving mentally between a tiny subset of English and the actual control language.
10	1	0.013270	0.049772	0.001820	I think there may be a broader lesson here:
15	1	0.024336	0.320621	0.006766	16. When your language is nowhere near Turing-complete, syntactic sugar can be your friend.
7	1	0.010428	0.032949	0.001431	Another lesson is about security by obscurity.
26	1	0.064344	2.890915	0.292139	Some fetchmail users asked me to change the software to store passwords encrypted in the rc file, so snoopers wouldn't be able to casually see them.
10	1	0.015343	0.027350	0.001527	I didn't do it, because this doesn't actually add protection.
47	0	0.265920	141.457336	0.013915	Anyone who's acquired permissions to read your rc file will be able to run fetchmail as you anyway—and if it's your password they're after, they'd be able to rip the necessary decoder out of the fetchmail code itself to get it.
21	1	0.097195	0.166471	0.006711	All .fetchmailrc password encryption would have done is give a false sense of security to people who don't think very hard.
6	1	0.010883	0.020966	0.000948	The general rule here is:
11	1	0.023643	0.038381	0.001419	17. A security system is only as secure as its secret.
4	1	0.004663	0.002893	0.000560	Beware of pseudo-secrets.
6	1	0.009752	0.018807	0.001673	Necessary Preconditions for the Bazaar Style
47	1	0.112673	84.986214	18.093262	Early reviewers and test audiences for this essay consistently raised questions about the preconditions for successful bazaar-style development, including both the qualifications of the project leader and the state of code at the time one goes public and starts to try to build a co-developer community.
15	0	0.026020	0.469866	0.000527	It's fairly clear that one cannot code from the ground up in bazaar style.
22	1	0.032521	1.419232	0.010870	One can test, debug and improve in bazaar style, but it would be very hard to originate a project in bazaar mode.
4	1	0.005401	0.006942	0.000560	Linus didn't try it.
3	1	0.004771	0.002683	0.000428	I didn't either.
14	1	0.023763	0.657492	0.021368	Your nascent developer community needs to have something runnable and testable to play with.
17	1	0.055321	0.210777	0.008855	When you start community-building, what you need to be able to present is a plausible promise.
8	1	0.016911	0.070868	0.002037	Your program doesn't have to work particularly well.
9	1	0.017319	0.068165	0.002140	It can be crude, buggy, incomplete, and poorly documented.
28	1	0.083071	3.494708	0.192697	What it must not fail to do is (a) run, and (b) convince potential co-developers that it can be evolved into something really neat in the foreseeable future.
11	1	0.017549	0.093741	0.002111	Linux and fetchmail both went public with strong, attractive basic designs.
39	1	0.136904	14.908679	0.570188	Many people thinking about the bazaar model as I have presented it have correctly considered this critical, then jumped from that to the conclusion that a high degree of design intuition and cleverness in the project leader is indispensable.
7	1	0.018687	0.037696	0.001599	But Linus got his design from Unix.
23	1	0.070536	0.663489	0.011244	I got mine initially from the ancestral popclient (though it would later change a great deal, much more proportionately speaking than has Linux).
30	1	0.070792	0.744787	0.017398	So does the leader/coordinator for a bazaar-style effort really have to have exceptional design talent, or can he get by through leveraging the design talent of others?
34	1	0.074729	0.576270	0.012405	I think it is not critical that the coordinator be able to originate designs of exceptional brilliance, but it is absolutely critical that the coordinator be able to recognize good design ideas from others.
10	1	0.014175	0.099932	0.007990	Both the Linux and fetchmail projects show evidence of this.
26	1	0.100564	0.301774	0.005488	Linus, while not (as previously discussed) a spectacularly original designer, has displayed a powerful knack for recognizing good design and integrating it into the Linux kernel.
20	1	0.044598	0.276805	0.014175	And I have already described how the single most powerful design idea in fetchmail (SMTP forwarding) came from somebody else.
34	1	0.130911	2.312610	4.009349	Early audiences of this essay complimented me by suggesting that I am prone to undervalue design originality in bazaar projects because I have a lot of it myself, and therefore take it for granted.
19	1	0.040712	0.208726	0.004458	There may be some truth to this; design (as opposed to coding or debugging) is certainly my strongest skill.
37	1	0.096350	13.708181	1.086379	But the problem with being clever and original in software design is that it gets to be a habit—you start reflexively making things cute and complicated when you should be keeping them robust and simple.
20	1	0.074143	0.799251	0.014561	I have had projects crash on me because I made this mistake, but I managed to avoid this with fetchmail.
29	1	0.052944	1.032996	0.019942	So I believe the fetchmail project succeeded partly because I restrained my tendency to be clever; this argues (at least) against design originality being essential for successful bazaar projects.
3	1	0.004904	0.004934	0.000592	And consider Linux.
38	1	0.464328	3.958498	0.306027	Suppose Linus Torvalds had been trying to pull off fundamental innovations in operating system design during the development; does it seem at all likely that the resulting kernel would be as stable and successful as what we have?
31	1	0.105443	3.042884	0.046551	A certain base level of design and coding skill is required, of course, but I expect almost anybody seriously thinking of launching a bazaar effort will already be above that minimum.
27	0	0.042979	1.011050	0.000389	The open-source community's internal market in reputation exerts subtle pressure on people not to launch development efforts they're not competent to follow through on.
9	1	0.027534	0.106689	0.004745	So far this seems to have worked pretty well.
31	1	0.087435	60.857231	8.558026	There is another kind of skill not normally associated with software development which I think is as important as design cleverness to bazaar projects—and it may be more important.
13	1	0.018317	0.759896	0.074148	A bazaar project coordinator or leader must have good people and communications skills.
4	1	0.006040	0.007570	0.000445	This should be obvious.
31	0	0.096035	11.989797	0.004621	In order to build a development community, you need to attract people, interest them in what you're doing, and keep them happy about the amount of work they're doing.
18	0	0.035333	0.574732	0.000879	Technical sizzle will go a long way towards accomplishing this, but it's far from the whole story.
6	1	0.008053	0.019952	0.001141	The personality you project matters, too.
21	1	0.061209	1.408075	0.058740	It is not a coincidence that Linus is a nice guy who makes people like him and want to help him.
29	0	0.075607	0.531273	0.000240	It's not a coincidence that I'm an energetic extrovert who enjoys working a crowd and has some of the delivery and instincts of a stand-up comic.
20	1	0.045636	0.199705	0.006074	To make the bazaar model work, it helps enormously if you have at least a little skill at charming people.
7	1	0.009565	0.029059	0.003134	The Social Context of Open-Source Software
35	1	0.104499	50.883213	1.126598	It is truly written: the best hacks start out as personal solutions to the author's everyday problems, and spread because the problem turns out to be typical for a large class of users.
18	1	0.047989	0.691282	0.007674	This takes us back to the matter of rule 1, restated in a perhaps more useful way:
16	1	0.043179	0.241593	0.016656	18. To solve an interesting problem, start by finding a problem that is interesting to you.
16	1	0.024860	0.147381	0.007177	So it was with Carl Harris and the ancestral popclient, and so with me and fetchmail.
9	1	0.020878	0.032927	0.001027	But this has been understood for a long time.
40	1	0.076309	2.309471	0.053530	The interesting point, the point that the histories of Linux and fetchmail seem to demand we focus on, is the next stage—the evolution of software in the presence of a large and active community of users and co-developers.
24	1	0.073490	1.558772	0.014696	In The Mythical Man-Month, Fred Brooks observed that programmer time is not fungible; adding developers to a late software project makes it later.
31	0	0.071477	2.296956	0.001711	As we've seen previously, he argued that the complexity and communication costs of a project rise with the square of the number of developers, while work done only rises linearly.
10	1	0.026973	0.078624	0.001328	Brooks's Law has been widely regarded as a truism.
39	0	0.112683	2.982869	0.000640	But we've examined in this essay a number of ways in which the process of open-source development falsifies the assumptionms behind it—and, empirically, if Brooks's Law were the whole picture Linux would be impossible.
22	1	0.046274	0.213221	0.008396	Gerald Weinberg's classic The Psychology of Computer Programming supplied what, in hindsight, we can see as a vital correction to Brooks.
38	0	0.118504	8.599705	0.000036	In his discussion of ``egoless programming'', Weinberg observed that in shops where developers are not territorial about their code, and encourage other people to look for bugs and potential improvements in it, improvement happens dramatically faster than elsewhere.
28	0	0.119775	4.674216	0.000245	(Recently, Kent Beck's `extreme programming' technique of deploying coders in pairs looking over one anothers' shoulders might be seen as an attempt to force this effect.)
30	0	0.082982	1.891023	0.000028	Weinberg's choice of terminology has perhaps prevented his analysis from gaining the acceptance it deserved—one has to smile at the thought of describing Internet hackers as ``egoless''.
11	1	0.024700	0.294468	0.001613	But I think his argument looks more compelling today than ever.
21	0	0.038234	0.056702	0.000022	The bazaar method, by harnessing the full power of the ``egoless programming'' effect, strongly mitigates the effect of Brooks's Law.
31	1	0.090113	15.222889	0.779614	The principle behind Brooks's Law is not repealed, but given a large developer population and cheap communications its effects can be swamped by competing nonlinearities that are not otherwise visible.
36	1	0.106316	14.752325	0.569098	This resembles the relationship between Newtonian and Einsteinian physics—the older system is still valid at low energies, but if you push mass and velocity high enough you get surprises like nuclear explosions or Linux.
31	0	0.077737	1.963452	0.000342	The history of Unix should have prepared us for what we're learning from Linux (and what I've verified experimentally on a smaller scale by deliberately copying Linus's methods).
23	1	0.060332	0.307167	0.006396	That is, while coding remains an essentially solitary activity, the really great hacks come from harnessing the attention and brainpower of entire communities.
51	1	0.169397	85.830734	5.773186	The developer who uses only his or her own brain in a closed project is going to fall behind the developer who knows how to create an open, evolutionary context in which feedback exploring the design space, code contributions, bug-spotting, and other improvements come from hundreds (perhaps thousands) of people.
17	1	0.040507	0.238283	0.004977	But the traditional Unix world was prevented from pushing this approach to the ultimate by several factors.
13	1	0.022690	0.099260	0.002881	One was the legal contraints of various licenses, trade secrets, and commercial interests.
11	1	0.028507	0.159998	0.003426	Another (in hindsight) was that the Internet wasn't yet good enough.
30	1	0.067535	2.347995	0.045033	Before cheap Internet, there were some geographically compact communities where the culture encouraged Weinberg's ``egoless'' programming, and a developer could easily attract a lot of skilled kibitzers and co-developers.
23	1	0.041588	1.976953	0.202808	Bell Labs, the MIT AI and LCS labs, UC Berkeley—these became the home of innovations that are legendary and still potent.
23	1	0.072855	0.307784	0.008303	Linux was the first project for which a conscious and successful effort to use the entire world as its talent pool was made.
53	0	0.102017	46.095512	0.000445	I don't think it's a coincidence that the gestation period of Linux coincided with the birth of the World Wide Web, and that Linux left its infancy during the same period in 1993–1994 that saw the takeoff of the ISP industry and the explosion of mainstream interest in the Internet.
20	1	0.053902	1.679919	0.081889	Linus was the first person who learned how to play by the new rules that pervasive Internet access made possible.
23	1	0.053608	0.563155	0.005050	While cheap Internet was a necessary condition for the Linux model to evolve, I think it was not by itself a sufficient condition.
30	1	0.080061	19.429279	0.785313	Another vital factor was the development of a leadership style and set of cooperative customs that could allow developers to attract co-developers and get maximum leverage out of the medium.
11	1	0.016783	0.082334	0.004784	But what is this leadership style and what are these customs?
24	1	0.042535	0.636197	0.004583	They cannot be based on power relationships—and even if they could be, leadership by coercion would not produce the results we see.
25	1	0.050950	1.094849	0.472316	Weinberg quotes the autobiography of the 19th-century Russian anarchist Pyotr Alexeyvich Kropotkin's Memoirs of a Revolutionist to good effect on this subject:
38	1	0.157436	10.126665	1.214347	Having been brought up in a serf-owner's family, I entered active life, like all young men of my time, with a great deal of confidence in the necessity of commanding, ordering, scolding, punishing and the like.
52	0	0.177471	116.182663	0.000023	But when, at an early stage, I had to manage serious enterprises and to deal with [free] men, and when each mistake would lead at once to heavy consequences, I began to appreciate the difference between acting on the principle of command and discipline and acting on the principle of common understanding.
33	0	0.245437	1.823357	0.000024	The former works admirably in a military parade, but it is worth nothing where real life is concerned, and the aim can be achieved only through the severe effort of many converging wills.
37	1	0.116778	4.466710	0.245694	The ``severe effort of many converging wills'' is precisely what a project like Linux requires—and the ``principle of command'' is effectively impossible to apply among volunteers in the anarchist's paradise we call the Internet.
35	1	0.167851	27.191008	1.401903	To operate and compete effectively, hackers who want to lead collaborative projects have to learn how to recruit and energize effective communities of interest in the mode vaguely suggested by Kropotkin's ``principle of understanding''.
8	1	0.014557	0.092850	0.002824	They must learn to use Linus's Law.
15	0	0.031343	0.251442	0.000219	Earlier I referred to the ``Delphi effect'' as a possible explanation for Linus's Law.
15	1	0.017600	0.284394	0.018173	But more powerful analogies to adaptive systems in biology and economics also irresistably suggest themselves.
46	1	0.106372	2.772658	0.073285	The Linux world behaves in many respects like a free market or an ecology, a collection of selfish agents attempting to maximize utility which in the process produces a self-correcting spontaneous order more elaborate and efficient than any amount of central planning could have achieved.
11	1	0.028433	0.066827	0.002542	Here, then, is the place to seek the ``principle of understanding''.
25	1	0.045312	0.426766	0.030556	The ``utility function'' Linux hackers are maximizing is not classically economic, but is the intangible of their own ego satisfaction and reputation among other hackers.
23	1	0.043864	0.618636	0.008884	(One may call their motivation ``altruistic'', but this ignores the fact that altruism is itself a form of ego satisfaction for the altruist).
49	1	0.503888	3.484601	0.089042	Voluntary cultures that work this way are not actually uncommon; one other in which I have long participated is science fiction fandom, which unlike hackerdom has long explicitly recognized ``egoboo'' (ego-boosting, or the enhancement of one's reputation among other fans) as the basic drive behind volunteer activity.
43	1	0.161763	2.212777	0.203884	Linus, by successfully positioning himself as the gatekeeper of a project in which the development is mostly done by others, and nurturing interest in the project until it became self-sustaining, has shown an acute grasp of Kropotkin's ``principle of shared understanding''.
17	1	0.041146	0.130960	0.005990	This quasi-economic view of the Linux world enables us to see how that understanding is applied.
39	1	0.286139	12.355110	0.815038	We may view Linus's method as a way to create an efficient market in ``egoboo''—to connect the selfishness of individual hackers as firmly as possible to difficult ends that can only be achieved by sustained cooperation.
21	1	0.045930	1.740274	0.062750	With the fetchmail project I have shown (albeit on a smaller scale) that his methods can be duplicated with good results.
14	0	0.029101	0.148702	0.000191	Perhaps I have even done it a bit more consciously and systematically than he.
25	1	0.055888	0.846383	0.038065	Many people (especially those who politically distrust free markets) would expect a culture of self-directed egoists to be fragmented, territorial, wasteful, secretive, and hostile.
21	1	0.042948	0.355622	0.024744	But this expectation is clearly falsified by (to give just one example) the stunning variety, quality, and depth of Linux documentation.
20	0	0.081016	0.107483	0.000035	It is a hallowed given that programmers hate documenting; how is it, then, that Linux hackers generate so much documentation?
25	1	0.058319	0.189108	0.013819	Evidently Linux's free market in egoboo works better to produce virtuous, other-directed behavior than the massively-funded documentation shops of commercial software producers.
45	1	0.105660	43.768658	8.917637	Both the fetchmail and Linux kernel projects show that by properly rewarding the egos of many other hackers, a strong developer/coordinator can use the Internet to capture the benefits of having lots of co-developers without having a project collapse into a chaotic mess.
10	1	0.021117	0.065077	0.004384	So to Brooks's Law I counter-propose the following:
31	1	0.081702	3.241883	0.276960	19: Provided the development coordinator has a communications medium at least as good as the Internet, and knows how to lead without coercion, many heads are inevitably better than one.
31	1	0.075539	4.161103	0.131179	I think the future of open-source software will increasingly belong to people who know how to play Linus's game, people who leave behind the cathedral and embrace the bazaar.
48	1	0.141252	24.123890	0.258682	This is not to say that individual vision and brilliance will no longer matter; rather, I think that the cutting edge of open-source software will belong to people who start from individual vision and brilliance, then amplify it through the effective construction of voluntary communities of interest.
11	1	0.019124	0.051633	0.001563	Perhaps this is not only the future of open-source software.
20	1	0.044092	0.237691	0.020715	No closed-source developer can match the pool of talent the Linux community can bring to bear on a problem.
23	0	0.040910	2.530085	0.000025	Very few could afford even to hire the more than 200 (1999: 600, 2000: 800) people who have contributed to fetchmail!
62	1	0.864859	129.638062	11.339336	Perhaps in the end the open-source culture will triumph not because cooperation is morally right or software ``hoarding'' is morally wrong (assuming you believe the latter, which neither Linus nor I do), but simply because the closed-source world cannot win an evolutionary arms race with open-source communities that can put orders of magnitude more skilled time into a problem.
6	1	0.006838	0.009902	0.001078	On Management and the Maginot Line
33	0	0.057069	1.083720	0.000023	The original Cathedral and Bazaar paper of 1997 ended with the vision above—that of happy networked hordes of programmer/anarchists outcompeting and overwhelming the hierarchical world of conventional closed software.
16	1	0.038697	0.155677	0.003291	A good many skeptics weren't convinced, however; and the questions they raise deserve a fair engagement.
25	1	0.055865	0.447201	0.005278	Most of the objections to the bazaar argument come down to the claim that its proponents have underestimated the productivity-multiplying effect of conventional management.
46	1	0.161823	2.506770	0.044490	Traditionally-minded software-development managers often object that the casualness with which project groups form and change and dissolve in the open-source world negates a significant part of the apparent advantage of numbers that the open-source community has over any single closed-source developer.
46	0	0.127584	50.589691	21.963150	They would observe that in software development it is really sustained effort over time and the degree to which customers can expect continuing investment in the product that matters, not just how many people have thrown a bone in the pot and left it to simmer.
36	1	0.100417	2.989793	0.249210	There is something to this argument, to be sure; in fact, I have developed the idea that expected future service value is the key to the economics of software production in the essay The Magic Cauldron.
21	1	0.032819	0.115610	0.004116	But this argument also has a major hidden problem; its implicit assumption that open-source development cannot deliver such sustained effort.
38	1	0.100337	6.421541	1.219746	In fact, there have been open-source projects that maintained a coherent direction and an effective maintainer community over quite long periods of time without the kinds of incentive structures or institutional controls that conventional management finds essential.
50	1	0.420093	5.811188	1.539466	The development of the GNU Emacs editor is an extreme and instructive example; it has absorbed the efforts of hundreds of contributors over 15 years into a unified architectural vision, despite high turnover and the fact that only one person (its author) has been continuously active during all that time.
10	1	0.029314	0.043476	0.001195	No closed-source editor has ever matched this longevity record.
27	1	0.046448	1.077749	0.095267	This suggests a reason for questioning the advantages of conventionally-managed software development that is independent of the rest of the arguments over cathedral vs. bazaar mode.
37	0	0.089284	2.400359	0.000558	Whatever it is certainly doesn't include reliable execution by deadline, or on budget, or to all features of the specification; it's a rare `managed' project that meets even one of these goals, let alone all three.
26	1	0.055936	0.269799	0.007829	One thing many people think the traditional mode buys you is somebody to hold legally liable and potentially recover compensation from if the project goes wrong.
31	1	0.090754	3.207777	0.167644	But this is an illusion; most software licenses are written to disclaim even warranty of merchantability, let alone performance—and cases of successful recovery for software nonperformance are vanishingly rare.
17	1	0.058884	0.815169	0.032223	Even if they were common, feeling comforted by having somebody to sue would be missing the point.
12	1	0.025420	0.071348	0.001666	You didn't want to be in a lawsuit; you wanted working software.
8	1	0.017410	0.063165	0.004362	So what is all that management overhead buying?
16	1	0.031552	0.780924	0.033340	In order to understand that, we need to understand what software development managers believe they do.
21	1	0.037502	0.490375	0.006635	A woman I know who seems to be very good at this job says software project management has five functions:
11	1	0.022509	0.198250	0.005203	To define goals and keep everybody pointed in the same direction
10	1	0.031317	0.118528	0.019333	To monitor and make sure crucial details don't get skipped
9	1	0.017473	0.128490	0.008429	To motivate people to do boring but necessary drudgework
9	0	0.008528	0.010252	0.000018	To organize the deployment of people for best productivity
8	1	0.013281	0.032832	0.002685	To marshal resources needed to sustain the project
25	1	0.040251	0.729529	0.019971	Apparently worthy goals, all of these; but under the open-source model, and in its surrounding social context, they can begin to seem strangely irrelevant.
7	0	0.016771	0.037573	0.000098	We'll take them in reverse order.
50	1	0.147979	18.221748	1.058762	My friend reports that a lot of resource marshalling is basically defensive; once you have your people and machines and office space, you have to defend them from peer managers competing for the same resources, and from higher-ups trying to allocate the most efficient use of a limited pool.
38	1	0.114247	23.768799	12.589278	But open-source developers are volunteers, self-selected for both interest and ability to contribute to the projects they work on (and this remains generally true even when they are being paid a salary to hack open source.)
23	1	0.069446	0.144909	0.007085	The volunteer ethos tends to take care of the `attack' side of resource-marshalling automatically; people bring their own resources to the table.
17	1	0.029822	0.466150	0.027053	And there is little or no need for a manager to `play defense' in the conventional sense.
24	0	0.049692	0.336539	0.000081	Anyway, in a world of cheap PCs and fast Internet links, we find pretty consistently that the only really limiting resource is skilled attention.
28	1	0.053811	0.546419	0.011426	Open-source projects, when they founder, essentially never do so for want of machines or links or office space; they die only when the developers themselves lose interest.
29	0	0.059324	1.182293	0.000994	That being the case, it's doubly important that open-source hackers organize themselves for maximum productivity by self-selection—and the social milieu selects ruthlessly for competence.
37	1	0.076081	4.566682	0.037000	My friend, familiar with both the open-source world and large closed projects, believes that open source has been successful partly because its culture only accepts the most talented 5% or so of the programming population.
41	0	0.167973	0.952449	0.000023	She spends most of her time organizing the deployment of the other 95%, and has thus observed first-hand the well-known variance of a factor of one hundred in productivity between the most able programmers and the merely competent.
35	0	0.069895	1.890702	0.000388	The size of that variance has always raised an awkward question: would individual projects, and the field as a whole, be better off without more than 50% of the least able in it?
38	1	0.096337	2.955424	0.090087	Thoughtful managers have understood for a long time that if conventional software management's only function were to convert the least able from a net loss to a marginal win, the game might not be worth the candle.
47	1	0.129801	20.857855	0.658154	The success of the open-source community sharpens this question considerably, by providing hard evidence that it is often cheaper and more effective to recruit self-selected volunteers from the Internet than it is to manage buildings full of people who would rather be doing something else.
9	1	0.016926	0.015863	0.001128	Which brings us neatly to the question of motivation.
33	1	0.100401	17.640104	1.525790	An equivalent and often-heard way to state my friend's point is that traditional development management is a necessary compensation for poorly motivated programmers who would not otherwise turn out good work.
53	0	0.280118	11.154900	0.000274	This answer usually travels with a claim that the open-source community can only be relied on only to do work that is `sexy' or technically sweet; anything else will be left undone (or done only poorly) unless it's churned out by money-motivated cubicle peons with managers cracking whips over them.
17	1	0.026481	0.156481	0.005018	I address the psychological and social reasons for being skeptical of this claim in Homesteading the Noosphere.
20	0	0.053439	2.173338	0.000417	For present purposes, however, I think it's more interesting to point out the implications of accepting it as true.
59	0	0.614530	95.158493	0.000023	If the conventional, closed-source, heavily-managed style of software development is really defended only by a sort of Maginot Line of problems conducive to boredom, then it's going to remain viable in each individual application area for only so long as nobody finds those problems really interesting and nobody else finds any way to route around them.
60	0	0.264292	12.117504	0.003015	Because the moment there is open-source competition for a `boring' piece of software, customers are going to know that it was finally tackled by someone who chose that problem to solve because of a fascination with the problem itself—which, in software as in other kinds of creative work, is a far more effective motivator than money alone.
30	1	0.082057	3.082522	0.071320	Having a conventional management structure solely in order to motivate, then, is probably good tactics but bad strategy; a short-term win, but in the longer term a surer loss.
34	0	0.115308	110.644081	0.000023	So far, conventional development management looks like a bad bet now against open source on two points (resource marshalling, organization), and like it's living on borrowed time with respect to a third (motivation).
44	1	0.560540	11.491027	0.200421	And the poor beleaguered conventional manager is not going to get any succour from the monitoring issue; the strongest argument the open-source community has is that decentralized peer review trumps all the conventional methods for trying to ensure that details don't get slipped.
16	1	0.023206	0.153873	0.019852	Can we save defining goals as a justification for the overhead of conventional software project management?
45	0	0.175752	163.966934	0.009198	Perhaps; but to do so, we'll need good reason to believe that management committees and corporate roadmaps are more successful at defining worthy and widely shared goals than the project leaders and tribal elders who fill the analogous role in the open-source world.
13	1	0.029569	0.043575	0.001511	That is on the face of it a pretty hard case to make.
36	1	0.152011	15.418553	11.286896	And it's not so much the open-source side of the balance (the longevity of Emacs, or Linus Torvalds's ability to mobilize hordes of developers with talk of ``world domination'') that makes it tough.
16	0	0.028312	0.059449	0.000094	Rather, it's the demonstrated awfulness of conventional mechanisms for defining the goals of software projects.
32	1	0.118997	3.030559	0.143693	One of the best-known folk theorems of software engineering is that 60% to 75% of conventional software projects either are never completed or are rejected by their intended users.
42	0	0.128155	26.731409	0.011759	If that range is anywhere near true (and I've never met a manager of any experience who disputes it) then more projects than not are being aimed at goals that are either (a) not realistically attainable, or (b) just plain wrong.
42	1	0.102901	5.765344	0.830656	This, more than any other problem, is the reason that in today's software engineering world the very phrase ``management committee'' is likely to send chills down the hearer's spine—even (or perhaps especially) if the hearer is a manager.
20	1	0.034525	0.166420	0.003574	The days when only programmers griped about this pattern are long past; Dilbert cartoons hang over executives' desks now.
37	1	0.094601	0.504374	0.008802	Our reply, then, to the traditional software development manager, is simple—if the open-source community has really underestimated the value of conventional management, why do so many of you display contempt for your own process?
22	1	0.043198	0.154091	0.005853	Once again the example of the open-source community sharpens this question considerably—because we have fun doing what we do.
18	0	0.038650	0.083170	0.000022	Our creative play has been racking up technical, market-share, and mind-share successes at an astounding rate.
17	0	0.034528	0.858551	0.000455	We're proving not only that we can do better software, but that joy is an asset.
49	0	0.145759	2.771466	0.000025	Two and a half years after the first version of this essay, the most radical thought I can offer to close with is no longer a vision of an open-source–dominated software world; that, after all, looks plausible to a lot of sober people in suits these days.
23	1	0.045702	0.691976	0.021107	Rather, I want to suggest what may be a wider lesson about software, (and probably about every kind of creative or professional work).
30	1	0.088157	2.697071	0.152854	Human beings generally take pleasure in a task when it falls in a sort of optimal-challenge zone; not so easy as to be boring, not too hard to achieve.
20	1	0.042926	1.197207	0.053719	A happy programmer is one who is neither underutilized nor weighed down with ill-formulated goals and stressful process friction.
3	1	0.003859	0.002431	0.000481	Enjoyment predicts efficiency.
36	0	0.224463	30.622860	0.012217	Relating to your own work process with fear and loathing (even in the displaced, ironic way suggested by hanging up Dilbert cartoons) should therefore be regarded in itself as a sign that the process has failed.
36	0	0.166423	1.112681	0.000068	Joy, humor, and playfulness are indeed assets; it was not mainly for the alliteration that I wrote of "happy hordes" above, and it is no mere joke that the Linux mascot is a cuddly, neotenous penguin.
33	1	0.081275	20.872286	0.216826	It may well turn out that one of the most important effects of open source's success will be to teach us that play is the most economically efficient mode of creative work.
6	1	0.006308	0.014295	0.001528	Epilog: Netscape Embraces the Bazaar
13	0	0.034339	0.375496	0.000561	It's a strange feeling to realize you're helping make history....
29	0	0.051071	117.325134	0.000025	On January 22 1998, approximately seven months after I first published The Cathedral and the Bazaar, Netscape Communications, Inc. announced plans to give away the source for Netscape Communicator.
16	1	0.050387	0.642714	0.018462	I had had no clue this was going to happen before the day of the announcement.
40	1	0.169941	12.838938	1.383502	Eric Hahn, executive vice president and chief technology officer at Netscape, emailed me shortly afterwards as follows: ``On behalf of everyone at Netscape, I want to thank you for helping us get to this point in the first place.
10	1	0.022876	0.042155	0.001256	Your thinking and writings were fundamental inspirations to our decision.''
32	1	0.072813	2.754572	2.733524	The following week I flew out to Silicon Valley at Netscape's invitation for a day-long strategy conference (on 4 Feb 1998) with some of their top executives and technical people.
10	1	0.016057	0.068961	0.003566	We designed Netscape's source-release strategy and license together.
9	1	0.017710	0.056470	0.005747	A few days later I wrote the following:
21	1	0.035167	0.363017	0.006323	Netscape is about to provide us with a large-scale, real-world test of the bazaar model in the commercial world.
33	1	0.063875	0.238074	0.004036	The open-source culture now faces a danger; if Netscape's execution doesn't work, the open-source concept may be so discredited that the commercial world won't touch it again for another decade.
10	1	0.017995	0.019667	0.000984	On the other hand, this is also a spectacular opportunity.
14	1	0.028110	0.089335	0.002249	Initial reaction to the move on Wall Street and elsewhere has been cautiously positive.
10	0	0.027496	0.140336	0.000103	We're being given a chance to prove ourselves, too.
22	1	0.046590	0.476118	0.010034	If Netscape regains substantial market share through this move, it just may set off a long-overdue revolution in the software industry.
11	1	0.016457	0.031835	0.001728	The next year should be a very instructive and interesting time.
4	1	0.013083	0.009067	0.001164	And indeed it was.
20	1	0.063957	0.438778	0.007732	As I write in mid-2000, the development of what was later named Mozilla has been only a qualified success.
18	1	0.036688	0.155631	0.003651	It achieved Netscape's original goal, which was to deny Microsoft a monopoly lock on the browser market.
17	1	0.030341	0.101424	0.003726	It has also achieved some dramatic successes (notably the release of the next-generation Gecko rendering engine).
21	1	0.070913	0.314230	0.028396	However, it has not yet garnered the massive development effort from outside Netscape that the Mozilla founders had originally hoped for.
38	1	0.158122	35.776623	1.300783	The problem here seems to be that for a long time the Mozilla distribution actually broke one of the basic rules of the bazaar model; it didn't ship with something potential contributors could easily run and see working.
19	0	0.039137	0.520590	0.000377	(Until more than a year after release, building Mozilla from source required a license for the proprietary Motif library.)
55	0	0.364485	87.716110	0.000024	Most negatively (from the point of view of the outside world) the Mozilla group didn't ship a production-quality browser for two and a half years after the project launch—and in 1999 one of the project's principals caused a bit of a sensation by resigning, complaining of poor management and missed opportunities.
10	0	0.030995	0.048775	0.000045	``Open source,'' he correctly observed, ``is not magic pixie dust.''
5	1	0.010034	0.013719	0.001291	And indeed it is not.
43	1	0.331542	13.646282	1.905009	The long-term prognosis for Mozilla looks dramatically better now (in November 2000) than it did at the time of Jamie Zawinski's resignation letter—in the last few weeks the nightly releases have finally passed the critical threshold to production usability.
36	1	0.096858	70.718109	5.779320	But Jamie was right to point out that going open will not necessarily save an existing project that suffers from ill-defined goals or spaghetti code or any of the software engineering's other chronic ills.
19	1	0.035303	0.266221	0.006805	Mozilla has managed to provide an example simultaneously of how open source can succeed and how it could fail.
16	1	0.043384	0.273853	0.008965	In the mean time, however, the open-source idea has scored successes and found backers elsewhere.
33	0	0.078951	0.771287	0.000581	Since the Netscape release we've seen a tremendous explosion of interest in the open-source development model, a trend both driven by and driving the continuing success of the Linux operating system.
11	1	0.027334	0.069094	0.001798	The trend Mozilla touched off is continuing at an accelerating rate.
1	1	0.002192	0.000549	0.000358	Notes
28	1	0.106209	16.968987	0.972543	In Programming Pearls, the noted computer-science aphorist Jon Bentley comments on Brooks's observation with ``If you plan to throw one away, you will throw away two.''.
5	1	0.011766	0.028254	0.001114	He is almost certainly right.
39	0	0.345238	8.358012	0.004337	The point of Brooks's observation, and Bentley's, isn't merely that you should expect first attempt to be wrong, it's that starting over with the right idea is usually more effective than trying to salvage a mess.
21	1	0.046702	0.562388	0.087395	Examples of successful open-source, bazaar development predating the Internet explosion and unrelated to the Unix and Internet traditions have existed.
20	1	0.032953	0.262759	0.023511	The development of the info-Zip compression utility during 1990–1992, primarily for DOS machines, was one such example.
47	1	0.353006	13.636877	4.847623	Another was the RBBS bulletin board system (again for DOS), which began in 1983 and developed a sufficiently strong community that there have been fairly regular releases up to the present (mid-1999) despite the huge technical advantages of Internet mail and file-sharing over local BBSs.
38	0	0.229611	6.762740	0.004491	While the info-Zip community relied to some extent on Internet mail, the RBBS developer culture was actually able to base a substantial on-line community on RBBS that was completely independent of the TCP/IP infrastructure.
24	1	0.073287	1.024260	0.052085	That transparency and peer review are valuable for taming the complexity of OS development turns out, after all, not to be a new concept.
22	1	0.051367	0.544240	0.014801	In 1965, very early in the history of time-sharing operating systems, Corbató and Vyssotsky, co-designers of the Multics operating system, wrote
16	1	0.051444	1.155692	0.018798	It is expected that the Multics system will be published when it is operating substantially...
56	1	0.256802	16.531179	1.074610	Such publication is desirable for two reasons: First, the system should withstand public scrutiny and criticism volunteered by interested readers; second, in an age of increasing complexity, it is an obligation to present and future system designers to make the inner operating system as lucid as possible so as to reveal the basic system issues.
25	1	0.087050	3.302183	0.032890	John Hasler has suggested an interesting explanation for the fact that duplication of effort doesn't seem to be a net drag on open-source development.
41	0	0.237738	2.550182	0.000029	He proposes what I'll dub ``Hasler's Law'': the costs of duplicated work tend to scale sub-quadratically with team size—that is, more slowly than the planning and management overhead that would be needed to eliminate them.
9	1	0.020058	0.033760	0.001447	This claim actually does not contradict Brooks's Law.
36	1	0.068160	5.146553	0.377222	It may be the case that total complexity overhead and vulnerability to bugs scales with the square of team size, but that the costs from duplicated work are nevertheless a special case that scales more slowly.
55	0	0.155192	114.023155	0.000023	It's not hard to develop plausible reasons for this, starting with the undoubted fact that it is much easier to agree on functional boundaries between different developers' code that will prevent duplication of effort than it is to prevent the kinds of unplanned bad interactions across the whole system that underly most bugs.
22	1	0.055733	3.398147	0.037387	The combination of Linus's Law and Hasler's Law suggests that there are actually three critical size regimes in software projects.
24	1	0.053047	0.242523	0.008452	On small projects (I would say one to at most three developers) no management structure more elaborate than picking a lead programmer is needed.
41	1	0.146976	15.540508	1.551630	And there is some intermediate range above that in which the cost of traditional management is relatively low, so its benefits from avoiding duplication of effort, bug-tracking, and pushing to see that details are not overlooked actually net out positive.
40	1	0.297712	13.842632	0.088784	Above that, however, the combination of Linus's Law and Hasler's Law suggests there is a large-project range in which the costs and problems of traditional management rise much faster than the expected cost from duplication of effort.
40	0	0.137272	1.747403	0.000574	Not the least of these costs is a structural inability to harness the many-eyeballs effect, which (as we've seen) seems to do a much better job than traditional management at making sure bugs and details are not overlooked.
21	1	0.034236	0.135025	0.004139	Thus, in the large-project case, the combination of these laws effectively drives the net payoff of traditional management to zero.
19	0	0.052804	0.352287	0.001039	The split between Linux's experimental and stable versions has another function related to, but distinct from, hedging risk.
10	1	0.011643	0.021382	0.001782	The split attacks another problem: the deadliness of deadlines.
31	1	0.164670	2.858081	0.058948	When programmers are held both to an immutable feature list and a fixed drop-dead date, quality goes out the window and there is likely a colossal mess in the making.
29	1	0.133314	2.913866	0.094492	I am indebted to Marco Iansiti and Alan MacCormack of the Harvard Business School for showing me evidence that relaxing either one of these constraints can make scheduling workable.
26	1	0.076646	10.426630	0.790524	One way to do this is to fix the deadline but leave the feature list flexible, allowing features to drop off if not completed by deadline.
44	1	0.198609	53.182678	7.207754	This is essentially the strategy of the "stable" kernel branch; Alan Cox (the stable-kernel maintainer) puts out releases at fairly regular intervals, but makes no guarantees about when particular bugs will be fixed or what features will beback-ported from the experimental branch.
20	1	0.078697	0.984360	0.016543	The other way to do this is to set a desired feature list and deliver only when it is done.
10	1	0.015623	0.029851	0.001190	This is essentially the strategy of the "experimental" kernel branch.
36	0	0.331684	32.987381	0.007464	De Marco and Lister cited research showing that this scheduling policy ("wake me up when it's done") produces not only the highest quality but, on average, shorter delivery times than either "realistic" or "aggressive" scheduling.
42	0	0.131714	1.575623	0.000588	I have come to suspect (as of early 2000) that in earlier versions of this essay I severely underestimated the importance of the "wake me up when it's done" anti-deadline policy to the open-source community's productivity and quality.
28	1	0.046925	5.413289	0.265937	General experience with the rushed GNOME 1.0 release in 1999 suggests that pressure for a premature release can neutralize many of the quality benefits open source normally confers.
37	0	0.146950	2.218641	0.000023	It may well turn out to be that the process transparency of open source is one of three co-equal drivers of its quality, along with "wake me up when it's done" scheduling and developer self-selection.
46	0	0.141260	21.217443	0.002742	It's tempting, and not entirely inaccurate, to see the core-plus-halo organization characteristic of open-source projects as an Internet-enabled spin on Brooks's own recommendation for solving the N-squared complexity problem, the "surgical-team" organization—but the differences are significant.
40	0	0.191644	0.399960	0.000022	The constellation of specialist roles such as "code librarian" that Brooks envisioned around the team leader doesn't really exist; those roles are executed instead by generalists aided by toolsets quite a bit more powerful than those of Brooks's day.
27	1	0.070056	2.330216	0.202144	Also, the open-source culture leans heavily on strong Unix traditions of modularity, APIs, and information hiding—none of which were elements of Brooks's prescription.
52	1	0.199459	65.462860	2.685329	The respondent who pointed out to me the effect of widely varying trace path lengths on the difficulty of characterizing a bug speculated that trace-path difficulty for multiple symptoms of the same bug varies "exponentially" (which I take to mean on a Gaussian or Poisson distribution, and agree seems very plausible).
21	1	0.039120	0.749721	0.014323	If it is experimentally possible to get a handle on the shape of this distribution, that would be extremely valuable data.
38	1	0.087204	3.856918	0.263069	Large departures from a flat equal-probability distribution of trace difficulty would suggest that even solo developers should emulate the bazaar strategy by bounding the time they spend on tracing a given symptom before they switch to another.
8	1	0.009545	0.068212	0.001232	Persistence may not always be a virtue...
27	1	0.066082	6.239296	0.255165	An issue related to whether one can start projects from zero in the bazaar style is whether the bazaar style is capable of supporting truly innovative work.
36	1	0.104913	3.290274	0.034769	Some claim that, lacking strong leadership, the bazaar can only handle the cloning and improvement of ideas already present at the engineering state of the art, but is unable to push the state of the art.
22	1	0.092183	3.347263	0.073790	This argument was perhaps most infamously made by the Halloween Documents, two embarrassing internal Microsoft memoranda written about the open-source phenomenon.
42	0	0.138823	0.431524	0.000024	The authors compared Linux's development of a Unix-like operating system to ``chasing taillights'', and opined ``(once a project has achieved "parity" with the state-of-the-art), the level of management necessary to push towards new frontiers becomes massive.''
10	1	0.015114	0.048060	0.002320	There are serious errors of fact implied in this argument.
32	1	0.067463	54.365078	3.999263	One is exposed when the Halloween authors themselves later observe that ``often [...] new research ideas are first implemented and available on Linux before they are available / incorporated into other platforms.''
17	1	0.050680	0.661138	0.030750	If we read ``open source'' for ``Linux'', we see that this is far from a new phenomenon.
47	0	0.390452	132.190628	0.000023	Historically, the open-source community did not invent Emacs or the World Wide Web or the Internet itself by chasing taillights or being massively managed—and in the present, there is so much innovative work going on in open source that one is spoiled for choice.
37	1	0.122886	125.445610	12.420542	The GNOME project (to pick one of many) is pushing the state of the art in GUIs and object technology hard enough to have attracted considerable notice in the computer trade press well outside the Linux community.
16	1	0.041503	0.151601	0.002801	Other examples are legion, as a visit to Freshmeat on any given day will quickly prove.
32	1	0.053658	3.161078	0.064372	But there is a more fundamental error in the implicit assumption that the cathedral model (or the bazaar model, or any other kind of management structure) can somehow make innovation happen reliably.
3	1	0.006480	0.004907	0.000401	This is nonsense.
33	1	0.098940	2.270971	0.026881	Gangs don't have breakthrough insights—even volunteer groups of bazaar anarchists are usually incapable of genuine originality, let alone corporate committees of people with a survival stake in some status quo ante.
4	1	0.009753	0.004914	0.000665	Insight comes from individuals.
31	1	0.074623	7.127655	0.246656	The most their surrounding social machinery can ever hope to do is to be responsive to breakthrough insights—to nourish and reward and rigorously test them instead of squashing them.
15	1	0.024103	0.058309	0.002784	Some will characterize this as a romantic view, a reversion to outmoded lone-inventor stereotypes.
39	1	0.263764	2.536021	0.060372	Not so; I am not asserting that groups are incapable of developing breakthrough insights once they have been hatched; indeed, we learn from the peer-review process that such development groups are essential to producing a high-quality result.
26	1	0.065995	1.745327	0.045269	Rather I am pointing out that every such group development starts from—is necessarily sparked by—one good idea in one person's head.
21	1	0.033267	0.568661	0.004517	Cathedrals and bazaars and other social structures can catch that lightning and refine it, but they cannot make it on demand.
39	0	0.096657	6.542124	0.003158	Therefore the root problem of innovation (in software, or anywhere else) is indeed how not to squash it—but, even more fundamentally, it is how to grow lots of people who can have insights in the first place.
25	1	0.037205	1.137993	0.027876	To suppose that cathedral-style development could manage this trick but the low entry barriers and process fluidity of the bazaar cannot would be absurd.
67	1	0.254002	111.074966	13.369572	If what it takes is one person with one good idea, then a social milieu in which one person can rapidly attract the cooperation of hundreds or thousands of others with that good idea is going inevitably to out-innovate any in which the person has to do a political sales job to a hierarchy before he can work on his idea without risk of getting fired.
24	0	0.056549	0.394040	0.000447	And, indeed, if we look at the history of software innovation by organizations using the cathedral model, we quickly find it is rather rare.
26	1	0.042334	4.295534	0.278676	Large corporations rely on university research for new ideas (thus the Halloween Documents authors' unease about Linux's facility at coopting that research more rapidly).
12	1	0.021450	0.290367	0.003322	Or they buy out small companies built around some innovator's brain.
34	1	0.078195	1.437787	0.060824	In neither case is the innovation native to the cathedral culture; indeed, many innovations so imported end up being quietly suffocated under the "massive level of management" the Halloween Documents' authors so extol.
6	1	0.015995	0.021859	0.001156	That, however, is a negative point.
10	1	0.018284	0.072702	0.001951	The reader would be better served by a positive one.
8	1	0.019377	0.032819	0.001252	I suggest, as an experiment, the following:
12	1	0.015057	0.068575	0.004672	Pick a criterion for originality that you believe you can apply consistently.
21	0	0.043287	0.456595	0.001866	If your definition is ``I know it when I see it'', that's not a problem for purposes of this test.
21	1	0.044771	0.572430	0.066309	Pick any closed-source operating system competing with Linux, and a best source for accounts of current development work on it.
8	1	0.009216	0.031390	0.002342	Watch that source and Freshmeat for one month.
15	1	0.029029	0.198907	0.057252	Every day, count the number of release announcements on Freshmeat that you consider `original' work.
15	1	0.025349	0.241641	0.013744	Apply the same definition of `original' to announcements for that other OS and count them.
7	1	0.012489	0.092135	0.006488	Thirty days later, total up both figures.
25	1	0.061612	7.918462	0.047165	The day I wrote this, Freshmeat carried twenty-two release announcements, of which three appear they might push state of the art in some respect.
29	1	0.105366	3.735479	0.035118	This was a slow day for Freshmeat, but I will be astonished if any reader reports as many as three likely innovations a month in any closed-source channel.
29	1	0.052753	1.688558	0.054457	We now have history on a project that, in several ways, may provide a more indicative test of the bazaar premise than fetchmail; EGCS, the Experimental GNU Compiler System.
28	0	0.053107	0.329520	0.000024	This project was announced in mid-August of 1997 as a conscious attempt to apply the ideas in the early public versions of The Cathedral and the Bazaar.
16	1	0.037693	0.306680	0.015633	The project founders felt that the development of GCC, the Gnu C Compiler, had been stagnating.
40	1	0.282721	69.927162	31.001379	For about twenty months afterwards, GCC and EGCS continued as parallel products—both drawing from the same Internet developer population, both starting from the same GCC source base, both using pretty much the same Unix toolsets and development environment.
34	1	0.218137	5.044241	0.363412	The projects differed only in that EGCS consciously tried to apply the bazaar tactics I have previously described, while GCC retained a more cathedral-like organization with a closed developer group and infrequent releases.
19	1	0.063764	0.689184	0.007998	This was about as close to a controlled experiment as one could ask for, and the results were dramatic.
19	1	0.060497	0.425895	0.033990	Within months, the EGCS versions had pulled substantially ahead in features; better optimization, better support for FORTRAN and C++.
28	1	0.078271	20.052759	1.064115	Many people found the EGCS development snapshots to be more reliable than the most recent stable version of GCC, and major Linux distributions began to switch to EGCS.
32	0	0.071465	6.193932	0.000672	In April of 1999, the Free Software Foundation (the official sponsors of GCC) dissolved the original GCC development group and officially handed control of the project to the the EGCS steering team.
19	1	0.037105	0.183834	0.014540	Of course, Kropotkin's critique and Linus's Law raise some wider issues about the cybernetics of social organizations.
33	0	0.159520	2.616144	0.000104	Another folk theorem of software engineering suggests one of them; Conway's Law—commonly stated as ``If you have four groups working on a compiler, you'll get a 4-pass compiler''.
26	1	0.055410	2.947460	0.081005	The original statement was more general: ``Organizations which design systems are constrained to produce designs which are copies of the communication structures of these organizations.''
18	0	0.049146	0.133593	0.000024	We might put it more succinctly as ``The means determine the ends'', or even ``Process becomes product''.
19	1	0.034782	1.094712	0.043269	It is accordingly worth noting that in the open-source community organizational form and function match on many levels.
34	0	0.091282	4.635660	0.000741	The network is everything and everywhere: not just the Internet, but the people doing the work form a distributed, loosely coupled, peer-to-peer network that provides multiple redundancy and degrades very gracefully.
19	1	0.032806	0.235781	0.008011	In both networks, each node is important only to the extent that other nodes want to cooperate with it.
13	1	0.026805	0.075256	0.001787	The peer-to-peer part is essential to the community's astonishing productivity.
42	1	0.621833	2.953454	0.080040	The point Kropotkin was trying to make about power relationships is developed further by the `SNAFU Principle': ``True communication is possible only between equals, because inferiors are more consistently rewarded for telling their superiors pleasant lies than for telling the truth.''
19	1	0.038054	0.194709	0.003489	Creative teamwork utterly depends on true communication and is thus very seriously hindered by the presence of power relationships.
29	1	0.067977	0.845085	0.088125	The open-source community, effectively free of such power relationships, is teaching us by contrast how dreadfully much they cost in bugs, in lowered productivity, and in lost opportunities.
32	0	0.097262	0.918356	0.000633	Further, the SNAFU principle predicts in authoritarian organizations a progressive disconnect between decision-makers and reality, as more and more of the input to those who decide tends to become pleasant lies.
26	1	0.051104	0.585276	0.023042	The way this plays out in conventional software development is easy to see; there are strong incentives for the inferiors to hide, ignore, and minimize problems.
9	1	0.013832	0.019731	0.001196	When this process becomes product, software is a disaster.
1	1	0.000716	0.000124	0.000207	Bibliography
26	0	0.055294	1.117358	0.000300	I quoted several bits from Frederick P. Brooks's classic The Mythical Man-Month because, in many respects, his insights have yet to be improved upon.
23	1	0.036896	0.517295	0.088242	I heartily recommend the 25th Anniversary edition from Addison-Wesley (ISBN 0-201-83595-9), which adds his 1986 ``No Silver Bullet'' paper.
34	1	0.122002	4.006871	0.172342	The new edition is wrapped up by an invaluable 20-years-later retrospective in which Brooks forthrightly admits to the few judgements in the original text which have not stood the test of time.
29	1	0.173182	10.451000	0.143285	I first read the retrospective after the first public version of this essay was substantially complete, and was surprised to discover that Brooks attributed bazaar-like practices to Microsoft!
10	1	0.078929	0.312592	0.014965	(In fact, however, this attribution turned out to be mistaken.
33	0	0.150034	29.089151	0.003076	In 1998 we learned from the Halloween Documents that Microsoft's internal developer community is heavily balkanized, with the kind of general source access needed to support a bazaar not even truly possible.)
24	0	0.060654	1.615300	0.000026	Gerald M. Weinberg's The Psychology Of Computer Programming (New York, Van Nostrand Reinhold 1971) introduced the rather unfortunately-labeled concept of ``egoless programming''.
34	1	0.073362	15.356260	1.213061	While he was nowhere near the first person to realize the futility of the ``principle of command'', he was probably the first to recognize and argue the point in particular connection with software development.
37	1	0.132481	24.392540	68.312660	Richard P. Gabriel, contemplating the Unix culture of the pre-Linux era, reluctantly argued for the superiority of a primitive bazaar-like model in his 1989 paper ``LISP: Good News, Bad News, and How To Win Big''.
16	1	0.036808	0.187183	0.002951	Though dated in some respects, this essay is still rightly celebrated among LISP fans (including me).
18	0	0.041159	0.325599	0.000024	A correspondent reminded me that the section titled ``Worse Is Better'' reads almost as an anticipation of Linux.
11	1	0.013454	0.079195	0.003343	The paper is accessible on the World Wide Web at http://www.naggum.no/worse-is-better.html.
37	1	0.345155	50.423195	2.438544	De Marco and Lister's Peopleware: Productive Projects and Teams (New York; Dorset House, 1987; ISBN 0-932633-05-6) is an underappreciated gem which I was delighted to see Fred Brooks cite in his retrospective.
50	1	0.123707	5.214486	0.323510	While little of what the authors have to say is directly applicable to the Linux or open-source communities, the authors' insight into the conditions necessary for creative work is acute and worthwhile for anyone attempting to import some of the bazaar model's virtues into a commercial context.
30	1	0.086188	7.682764	1.142580	Finally, I must admit that I very nearly called this essay ``The Cathedral and the Agora'', the latter term being the Greek for an open market or public meeting place.
45	1	0.366403	128.348694	6.074011	The seminal ``agoric systems'' papers by Mark Miller and Eric Drexler, by describing the emergent properties of market-like computational ecologies, helped prepare me to think clearly about analogous phenomena in the open-source culture when Linux rubbed my nose in them five years later.
9	1	0.010063	0.031078	0.003640	These papers are available on the Web at http://www.agorics.com/agorpapers.html.
1	1	0.001155	0.000303	0.000302	Acknowledgements
16	1	0.032598	0.190425	0.003285	This essay was improved by conversations with a large number of people who helped debug it.
22	0	0.061689	0.317300	0.000245	Particular thanks to Jeff Dutky <dutky@wam.umd.edu>, who suggested the ``debugging is parallelizable'' formulation, and helped develop the analysis that proceeds from it.
15	1	0.017962	0.075010	0.005132	Also to Nancy Lebovitz <nancyl@universe.digex.net> for her suggestion that I emulate Weinberg by quoting Kropotkin.
17	1	0.024284	0.293650	0.007656	Perceptive criticisms also came from Joan Eslinger <wombat@kilimanjaro.engr.sgi.com> and Marty Franz <marty\snet-link.net> of the General Technics list.
34	1	0.098692	1.620755	0.083005	Glen Vandenburg <glv@vanderburg.org> pointed out the importance of self-selection in contributor populations and suggested the fruitful idea that much development rectifies `bugs of omission'; Daniel Upper <upper\speak.org> suggested the natural analogies for this.
28	0	0.054314	1.553778	0.000226	I'm grateful to the members of PLUG, the Philadelphia Linux User's group, for providing the first test audience for the first public version of this essay.
11	1	0.011758	0.037540	0.001649	Paula Matuszek <matusp00@mh.us.sbphrd.com> enlightened me about the practice of software management.
22	0	0.020784	0.095088	0.000087	Phil Hudson <phil.hudson@iname.com> reminded me that the social organization of the hacker culture mirrors the organization of its software, and vice-versa.
13	1	0.030059	0.464417	0.011547	John Buck <johnbuck@sea.ece.umassd.edu> pointed out that MATLAB makes an instructive parallel to Emacs.
19	1	0.047059	1.434166	0.108511	Russell Johnston <russjj@mail.com> brought me to consciousness about some of the mechanisms discussed in ``How Many Eyeballs Tame Complexity.''
13	1	0.071911	0.069843	0.001698	Finally, Linus Torvalds's comments were helpful and his early endorsement very encouraging.