> [...]
> anoncvs is a particularly bad case because, to get even moderate
> performance given the number of users we have, it needs a huge amount
> of memory so the whole repository fits in buffer cache. This costs
> bucks. (There really is no other way to do it -- CVS thrashes on i/o
> very badly, so no real disk can possibly handle 100 simultaneous
> checkouts, and we deal with that and worse all the time.)
 > [...]

may I ask why we still use cvs to distribute the source and repository 
files? I mean cvs may be fine for developing software, but for anything 
else? FreeBSD already uses cvsup (I know, it's based on modula3 and not 
portable) and cvsync sounds also very nice (though it still doesn't 
support fetching plain source files :-(). Or even rsync?!

regards,
SK