[TUHS] Re: TUHS Digest, Vol 12, Issue 2

Larry McVoy lm at bitmover.com
Wed May 26 23:52:46 AEST 2004

> From: Albert Cahalan <albert at users.sf.net>
> Subject: Re: [TUHS] cvsweb for BSD

> > One of the big problems is that they move files all over the place as BSD
> > developed and CVS doen't work too elegantly with those kind of changes.
> Bitkeeper handles this well. I suspect that Larry McVoy would
> at least be mildly interested in giving advice for such
> a project. Bitkeeper is SCCS-based.

Yes, I'd be interested.  Other than the renames I think we can automate
most of this.

> Bitkeeper also has a superior web interface. You can't beat
> standard unified diff format with a tiny bit of color added.

Thanks.  One day we'll get around to adding sub line highlighting - that 
would be an improvement.

> From: "M. Warner Losh" <imp at bsdimp.com>
> Subject: Re: [TUHS] cvsweb for BSD
> If you are going to use a proprietary system, you might as well use
> perforce, which has better branching and file movement support than
> bitkeeper.  But I guess I'm a little biased because I like p4 better
> than bk.

Both biased and incorrect.  There are over 10,000 branches of the linux
kernel floating around in BitKeeper (we know, we counted them) and we 
handle file movement much more nicely than perforce does (we have our
own concept of an inode, a pathname is a attribute of an inode just
like contents are an attribute of the inode - so you can move A to B,
I modify A, you pull from me and the changes apply to B in your tree.)

You can be as biased as you want, I don't want to turn this into a 
SCM discussion, but try and be accurate.  About the only thing that 
p4 does that we don't do is centralized locking; we don't need to
do that in a distributed/replicated system but people sometimes want it.
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

More information about the TUHS mailing list