logical: it would have to be that way (sup.netbsd.org not THE source
repository, mailing list on another machine), but the other questions go 
unanswered.

Where are the other level 1 machines (level 0 (root) being 'the' 
machine upon which the source tree is kept) to use when sup.netbsd.org
is down?  This sounds like a situation similair to ntp; why not think a 
little bit about the structure and allow more than one level 1 machine, 
then level 2, etc...  (and of course, publish the information)  It would 
seem a superior fan-out type of solution rather than just the one
sup.netbsd.org.  One could limit the load on the level N machines by having 
to be 'authorized' for use (or not), provide redundancy, and fault-tolerance.
It is kind of hard for us -current users (beta testers) to update if the 
connections are refused.

Of course, it would also be nice to be able to specify a list in one's
supfile (in order of preference) that would be tried if a connection
times out or fails for any reason so that the update is more likely to
complete.

chuck