Open vs closed source software: The quest for balance (Voxeu.org)

You have this one pretty much backwards, actually. I was there at the beginning of monotone, and *not* copying BK was a high priority. (Not that there was much danger, since we were forbidden from downloading or using BK anyway. But we had some vague idea of how it did various things, and that made those approaches *less* attractive.). We didn't really learn much about how BK worked until after Tridge did his reverse-engineering, and even then it was a mix of "huh, interesting idea, but..." and "they do *what*? WHY?". For instance, the most important design idea in monotone was content-addressed snapshots + chained hashing, and BK doesn't use crypto at all. (The most important design idea in BK is the use of SCCS weaves, but (1) weaves are not as awesome as Larry thinks they are, and (2) FOSS folk reinvented the important parts before Tridge's work.) To the extent that BK and monotone are similar, I think it has more to do with convergent evolution than anything else. (There were other projects around then using other approaches, like Arch -- but you don't remember them now.)

Arguably git/mercurial took monotone's ideas about history representation and combined them with BK's approach to branch tracking (the whole clone-to-branch workflow), but then, darcs was already using that approach, and I don't know that they got it from BK either.

If anything, the development of modern VCS systems is a great success story for FOSS. It's not just an area where we developed better software, it's an area where we developed better theory.

But it doesn't make much sense to talk about whether "proprietary software" or "FOSS" are better at innovation -- neither of them is a development model! BK was basically Larry's project; monotone was Graydon's project, bzr was Martin's, arch was Tom Lord's, darcs was David Roundy's, codeville was Bram Cohen's, git was Linus's, hg was Matt Mackall's.... plus other interested people in all cases. Each group built the best software they could, and Larry's penchant for proprietary licensing did not really give him any advantages or disadvantages in doing good distributed systems design. The proprietary licensing + business model *did* allow him to attract contributors a different way (by hiring them) and get more hours per week of contribution... but OTOH, it also meant they were stuck with a product to support, while the rest of us were exploring a larger space of ideas and eventually found better ones. (While Larry raged, made misleading claims about how BK worked, and generally tried his very best to hold the whole field back!)

There are certainly advantages to an organizational structure that lets one person say "okay, we're throwing a thousand hours of developer time into this vision I just wrote on a napkin" -- it's true that they'll tend to get to some approximation of that vision faster than many other organizational models. But I suspect the main "advantages" of proprietary software with regard to innovation is (1) there's more proprietary software out there period, so no wonder if there are more innovative ones in *absolute* terms (if that's even true -- most proprietary software is not very innovative!), (2) if someone does have an innovative idea like, say, PageRank, they're still more likely to take it proprietary than to release it as FOSS. That doesn't reflect the superiority of the proprietary model, just its attraction.