Hi,

Thus wrote matthew sporleder (msporleder@gmail.com):

> Actually, I like this idea a lot.  Having mirrors qualify before being
> put into the roller would be a good way to ensure quality.  Are there
> any metrics on the ratio of people who read/write with cvs versus
> people who are just reading to stay current with their branch? 

Anybody using anoncvs (or any of its mirrors) is doing that read-only.
They simply don't (can't) accept writes, since There Can Only Be One
cvs master server :)

> Metrics on how in-sync mirrors are?  Could they be dynamically moved
> from stratum levels?  Is svn less overhead?  rsync?  etc?  Maybe it
> would be easier to offer different methods into the repository.  (If
> that's possible.  I'm --not-- suggesting a move away from cvs for
> development and actual code storage, just asking if there is a
> faster/lighter-weight way to access a cvs repository)

All cvs mirrors by necessity use rsync to update their repositories, since
they need all the version diffs (i.e. the ,v files).
If one built a staggered mirror network one should take into account both
connectivity of the potential high stratum mirrors to the master site and
connectivity to 'neighborly' mirrors, as well as stability of the mirror
server and reactivity of the mirror admin. :)
In order to make for fast updates it should be a push and not a pull model
btw, but many mirror admins will balk at that suggestion.

just my 2 cents,
	spz
-- 
spz@serpens.de (S.P.Zeidler)