> Rather than answering your question I'll explain to you why the master
> NetBSD AnonCVS repository does *not* let you do this.  We don't let you
> do it because of the extreme I/O load rsync places on the machine by
> busting the namei cache and causing tens of thousands of lookups to need
> to hit the actual inodes which are scattered all over the disk.

Is _that_ the reason? There are lots of ways to fix that!

1. There could be a separate rsync.netbsd.org which does one nightly rsync
from AnonCVS and serves that to everyone else via rsync. Configure a
ridiculously huge namei cache on the rsync server. End of problem.

2. The repository currently available by rsync du's at about 614 megs, tars
up to about 625 megs, and gzip -9's to about 140 megs. A daily process could
tar up the various anoncvs modules and put them up for FTP on another server.
A simple 7-day rotation scheme would protect against the occasional script
failure.

3. Tell SUP how to push the entire repository as a collection. A huge one,
but still just a collection.


Lately I'm finding that I really enjoy having a complete repository on my
internal network, but right now that limits me to the base system sources,
apparently.

Would hardware donations of some kind help here?

Todd Whitesel
toddpw @ best.com