On Wed, 3 Sep 2003, Ville Syrjälä wrote: > > Isn't there some sort of delay involved here? I remember seeing something > about the kernel BK->CVS lagging by several days. If so moving to BK could > make the problem worse...
The main delay we've seen is when the central BK places have gone down due to disk issues or other maintenance. And one or two of those were indeed a few days. I think that is something you'll always end up having when you have one central site. Either the site is going to be _very_ expensive (major maintenance, redundancy, you name it), or the central site is going to have occasional hickups. This is clearly the problem with sourceforge anoncvs too: it's a central site that does _not_ get the big bucks for maintenance and hardware or software redundancy. So not only do things occasionally break, they stay broken for much longer than they should. Having a distributed setup works much better. Pure BK (without the CVS portal) is that. Read-write CVS on its own can't do it, but obviously it can be done with read-only anoncvs access by just replicating the repositories. It's a bit of a maintenance nightmare, but clearly doable. The _good_ news is that since there are only 31 committers, the writable repository itself probably doesn't need to be replicated - it just won't see the kind of load that requires it. Which is why the committers are ok already, and it's only anoncvs that has been broken. [ anoncvs has now been broken for so long that I think it's reason in itself for just moving away from sourceforge. Even if you don't change _anything_ else in the setup. Open source without access to actual users is pointless. ] Linus ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel