For all the scorn Tridgell gets, it would appear that Torvalds is actually more guilty of violating the license, he just had the clout to get away with it.
I was curious about this myself! I’ve been trying to find the actual license text, because this article just repeats what Wikipedia says, and Wikipedia doesn’t have any citations for that “one year” bit. Apparently the BK license had many versions, and none of the ones I’ve seen so far mention a non-compete period.
Update: Here’s an article from April 21, 2005. Git is brand-new, it officially exists… and McVoy doesn’t seem to consider it a violation at all. Linus simply stopped using BitKeeper and started writing his own thing, and everyone seemed okay with that. In general, it seems very much like the split happened on as good terms as possible (between Linus and McVoy, at least).
Also, I still haven’t seen any evidence for a one-year non-compete period. I’m increasingly skeptical that one existed.
It feels as though McVoy was attempting to concoct some IP concept by fiat that acted like a Patent without being one. This notion that he had total ownership over his particular data model and nobody was allowed to RE it is pretty objectionable to me, and I doubt it would ever hold up in a lawsuit.
The sad thing is that BitKeeper is now Apache licensed and does a few things much better than git (the submodule equivalent on BitKeeper looks fairly sensible, for example). Unfortunately, it was too late to dethrone git.
I mean, the same could be said for Mercurial for many years. Not that Git wasn’t superior than Subversion for Linux’s workflow, but at least Monotone, Mercurial, and BitKeeper have done things better than Git, at various points, in an objectively major way, and it never mattered, because software is ultimately social. BitKeeper could do things 1000% better than Git in every single possible axis, but who cares if I can’t interact with the LKML, right? And that’s played out in dozens of different ways since GitHub began The De Facto Source Repository.
I do not say that in a way to disparage Git. Git was meaningfully better than the mainstream from inception to now. But it was never the best at what it was doing until quite recently, if at all. What it had going were network effects. And for anything to ever displace Git, it’s going to have to deal with Git, because it’ll need those network effects. And that applies to BitKeeper as much as anything else.
One of my favourite parts of this story (not mentioned in the article) is that Tridge’s “reverse engineering” began by telnet-ing to the BK server and typing help.
My recollection is, Linus looked at Hg, but decided it was stupidly slow for his use case(and to be fair, it was basically brand new and mostly un-optimized python at that point in time). Of course Hg did a lot more stuff, was a lot more user friendly, objectively better in many ways, etc. Hg got optimized, but by then the writing was on the wall, Github existed and the war was over before there was even a battle really.
The one thing Git did, it was very fast for Linus’s use-case. It eventually got sped up for most other use-cases.
For all the scorn Tridgell gets, it would appear that Torvalds is actually more guilty of violating the license, he just had the clout to get away with it.
I was curious about this myself! I’ve been trying to find the actual license text, because this article just repeats what Wikipedia says, and Wikipedia doesn’t have any citations for that “one year” bit. Apparently the BK license had many versions, and none of the ones I’ve seen so far mention a non-compete period.
I have found lots of non-compete stuff in general, like this email where McVoy clarifies that a Subversion developer’s BitKeeper license is revoked. No mention of him still being bound by a non-compete; just a hey, you’re not allowed to use BitKeeper anymore. Though this was in 2002, so perhaps the non-compete period wasn’t in the license back then.
Here’s a more recent one: in 2005, a developer stops contributing to Mercurial because his day job uses BitKeeper. Again, no mention of a non-compete period, just that using BitKeeper and contributing to Mercurial are mutually exclusive.
I really want to see the actual license text that Linus would have been subject to at the time he wrote Git.
Update: Here’s an article from April 21, 2005. Git is brand-new, it officially exists… and McVoy doesn’t seem to consider it a violation at all. Linus simply stopped using BitKeeper and started writing his own thing, and everyone seemed okay with that. In general, it seems very much like the split happened on as good terms as possible (between Linus and McVoy, at least).
Also, I still haven’t seen any evidence for a one-year non-compete period. I’m increasingly skeptical that one existed.
Interesting detective work!
It feels as though McVoy was attempting to concoct some IP concept by fiat that acted like a Patent without being one. This notion that he had total ownership over his particular data model and nobody was allowed to RE it is pretty objectionable to me, and I doubt it would ever hold up in a lawsuit.
Honestly, good riddance.
He wasn’t replacing a version control system, he was providing patch management!
Right?
The sad thing is that BitKeeper is now Apache licensed and does a few things much better than git (the submodule equivalent on BitKeeper looks fairly sensible, for example). Unfortunately, it was too late to dethrone git.
I mean, the same could be said for Mercurial for many years. Not that Git wasn’t superior than Subversion for Linux’s workflow, but at least Monotone, Mercurial, and BitKeeper have done things better than Git, at various points, in an objectively major way, and it never mattered, because software is ultimately social. BitKeeper could do things 1000% better than Git in every single possible axis, but who cares if I can’t interact with the LKML, right? And that’s played out in dozens of different ways since GitHub began The De Facto Source Repository.
I do not say that in a way to disparage Git. Git was meaningfully better than the mainstream from inception to now. But it was never the best at what it was doing until quite recently, if at all. What it had going were network effects. And for anything to ever displace Git, it’s going to have to deal with Git, because it’ll need those network effects. And that applies to BitKeeper as much as anything else.
One of my favourite parts of this story (not mentioned in the article) is that Tridge’s “reverse engineering” began by
telnet-ing to the BK server and typinghelp.Source for that? Would appreciate!
Here’s a LWN article about a keynote Tridge gave at linux.conf.au in April 2005 (perfect timing!). I’ve only read the summary so I can’t be sure, but it sounds like he might have given a live demonstration of bringing up the help menu.
Thanks!
I think this post should at least briefly mention Mercurial as well.
My recollection is, Linus looked at Hg, but decided it was stupidly slow for his use case(and to be fair, it was basically brand new and mostly un-optimized python at that point in time). Of course Hg did a lot more stuff, was a lot more user friendly, objectively better in many ways, etc. Hg got optimized, but by then the writing was on the wall, Github existed and the war was over before there was even a battle really.
The one thing Git did, it was very fast for Linus’s use-case. It eventually got sped up for most other use-cases.
This is definitely part of the story! But also both git and hg came out as reactions to bk, within a few days of each other.