constant Bitkeeper bitching
Holy frelling frig-fest on a freighter of fruit-loops!
I've never seen such ridiculous behavior.. Well, except all the time on LK...
Anyway, folks, keep in mind the following. BK->CVS does not throw away any useful meta-data you would have had before using BK. You had patches, folks... That's it!! BK giveth, and giveth, and giveth, and all you see is taking away.
Whatever your thoughts on Larry, or the language in his license, you get to take the data out, sans only the metadata cvs cannot handle. Plus, you get much more granular history tracking you ever had with patches.
Now go sit and spin. I am neither a detractor or protractor... ;] I'm just sick of the bitching, as is everyone else.
Quoted 2 lines · Shawn
[..] You had
patches, folks... [..]
I agree with the rest but note that the above argument is silly too. Patches are missing a great deal of info. bitkeeper is more useful for a reason. Until today we had an information-loss problem, that was fixed only for a restricted number of people for the last year, so it was very far from a solution from my point of view. Today thanks to the kernel CVS that Larry thankfully provided, IMHO this is finally solved and I greatly appreciate that.
Of course if you don't develop the kernel you can live fine with monolithic undocumented patches, you're not going to audit those diffs anyways, do you? Few people will appreciate the difference between patches and bk, but for developers having the finer granularity helps a lot, so saying "go back to patches" is a no-way. Just try to extract stuff from the -ac tree and you get the idea, I'm stunned how can Alan submit stuff to Marcelo and Linus w/o major pain and leftovers (ok, some are driver changes and they're easy to extract, but not everything is that simple and self contained).
Andrea
Thanks for being a sane voice, btw...
Quoted 10 lines · Andrea Arcangeli
Quoted 2 lines · Shawn
[..] You had
patches, folks... [..]I agree with the rest but note that the above argument is silly too. Patches are missing a great deal of info. bitkeeper is more useful for a reason. Until today we had an information-loss problem, that was fixed only for a restricted number of people for the last year, so it was very far from a solution from my point of view. Today thanks to the kernel CVS that Larry thankfully provided, IMHO this is finally solved and I greatly appreciate that.
I don't disagree with you, it's just that this most recent thread was mainly about how the CVS gateway is/was lacking. (unless I'm missing the whole point). I would challenge any corporation to do what Larry has done, and tackle SCM to the best of his charitable ability for linux kernel development. Having said that, I'll say that of course concerns of the past had validity to them. So, if I'm not mistaken, I think we're agreeing.
Quoted 9 lines
Of course if you don't develop the kernel you can live fine with monolithic undocumented patches, you're not going to audit those diffs anyways, do you? Few people will appreciate the difference between patches and bk, but for developers having the finer granularity helps a lot, so saying "go back to patches" is a no-way. Just try to extract stuff from the -ac tree and you get the idea, I'm stunned how can Alan submit stuff to Marcelo and Linus w/o major pain and leftovers (ok, some are driver changes and they're easy to extract, but not everything is that simple and self contained).
Ditto on -ac... Actually, all you hard hitters kindof amaze me. I appreciate all y'all. And of course, I'm not saying go back to patches. It's just that when bk was offered as a solution, strings attached or no, it added a lot of sanity to the way changes are introduced into the kernel, and thankfully historical comments and whatever other metadata. Before that there was only lkml.
I'll try to clarify the point I was trying to make... Some all-too-vocal folks, even after the bk -> cvs gateway was created were saying "not good enough" because it was "lossy". It is the fact that bk -> cvs couldn't possibly lose any more than was lost inherently by not using bk. in other syntax...
lossiness("BK-to-CVS gw") == BK - "diff/patch"Basically, it's Linus's tree. He can get /all/ the meta-data out at will and convert it if he pleases, regardless of the discussions going on here. In any case, Larry is busy coding up "BitchKeeper" so threads like this can be managed more efficiently ;P