For a while, I've been thinking that ESR's papers (tCatB, HtN, tMC) don't quite tell the whole story. The rest of the story is probably hidden in jwz's comment about the "magic pixie dust of open source" (bottom of the page). My idea here is something along the lines that, yes, open source might well have sound business reasons, as ESR is at pains to point out both in his papers and on opensource.org, but too many people are treating it as magic pixie dust, just like jwz said.
I mean, look at opensource.org itself. It says there that "This site is still evolving as we think through the implications of open source in the commercial world". When was it last updated, eh? When was there some genuinely new content? There was the huge flurry when it was set up, and now the one site which portrays itself as the flagship of the Open Source concept sits idle. Look at SourceForge. It seems to me that anyone who writes a nine-line Perl script to do anything now needs a SourceForge project page wrapped around it. Bah! There are dozens of great Open Source projects out there (I'm sure you can name them), but there are an awful lot of people who are just saying "look, here's my project, released under the GPL, with its own SourceForge project page: now, where's this army of co-developers that ESR talks about to help me out and make it the next category killer?"
ESR would say, probably (although I don't like putting words in people's mouths), that he's covered this when he spoke about getting significant mindshare for your project; it has to excite interest in other people in order to get co-development. However, the impression I get from his papers is a kind of "release it and they will come" mentality. Now, as I say, I doubt that ESR himself thinks that, but if I read the papers as having that kind of subtext, it's reasonable to assume that others do too. That's what's behind this explosion of unfinished beta projects at SourceForge which don't ever look like they'll be finished. For instance, yes, every project starts by scratching a developer's personal itch, as tCatB says. The implication from the paper, as I read it, is that there are that many people out there in the world that there are bound to be lots of others who share your personal itch. However, I don't think that's the case any more, not if you're talking about, say, MUAs, or IRC clients, or any one of the few other things where there are already hundreds of different options. To give an example, I've been looking around for a GUI mail client, because I like GUI stuff, and I'm currently using Mutt. Now, at work I use Outlook -- it may do many bad things, but it's a reasonably well presented GUI, as evidenced by the fact that a pretty large chunk of Linux GUI mail clients emulate its interface to a greater or lesser degree. However, there are dozens of them. Mahogany, rzMail, Pronto, Postilion, Sylpheed, TkRat, TkMail, Aileron, and those are just the ones that I can name off the top of my head. I'd like to see people pull together a bit more. ESR's theory seems to be that market forces will mean that lots of clients get started, but, as things move on, most will fall by the wayside and a few will come up tops. I do not believe this to be the case, and I see it as a central flaw in the OSS methodology. Instead of one great app, what you get is many, many underdeveloped apps, because you've fragmented your developer community to such an extent. I think it's different for, say, Linux, or the Gimp, because writing those things is hard. Writing a mail client and slapping a Tk or Gtk or wxWindows GUI around it is pretty easy, though, so everyone and his wife writes their own one. OSS works for large projects where you need a lot of people to pull together. I just don't see the expected market forces working toward a winner in areas such as IRC clients or MUAs, which anyone can code. This, I think, is my key point.
© Aquarius, November 2000