Just as a curiosity, I was wondering why we use sup instead of running a
   little script on an NFS-mounted version of the source tree.

Huh?  What would such a script do?  Where would the source tree be
mounted from?  I don't really understand the question.

   Is there some
   kind of added functionality that could not be duplicated by the more standard
   Unix tools?

SUP makes it much easier for lots of pseudo-random people to request
updates.  You could do the same thing with, e.g., rdist(1), except
that it `pushes' rather than `pulls', so each user would have to
pre-arrange it with us.  You could do the same thing with FTP, but it
would be less efficient.

BTW, SUP isn't exactly new.  It's been in use at CMU for at least a
decade.