A brief note on how this seems to work:

Invoking "track -w" simply writes out the statsfile.

Invoking track from a client invokes calltrack which connects to the trackd
on the librarian. The librarian trackd forks off a "track -g" that
communicates with the calltrack, passing it appropriate stuff.
[With -t, track invokes rsh to the librarian to run track -g]

Track error reporting is now much better, but if you really want to debug a
problem, you may find it easier to run calltrack, rsh or track -g on the
librarian directly and speak the track protocol at it.

track -v gives you some idea of what track is doing. The >> marks lines that
the client track is sending the server, and the << shows the reply header
(it doesn't show the body)
