A suggestion:  the nntplink error reporting can make for a fairly
large log when network failures occur.  For instance, I now have quite
a lot of lines in my log reporting that zaphod refused to talk to us
:-) Any way to report the first instance, and then report them in
batches of 10 or so, or making the backoff faster?  The most
sophisticated solution is to report the beginning of the error
condition and the end, but that means that a successful operation has
to check if the last operation was an error. (moraes@cs.toronto.edu)

should nntplink NOT start up if there is already a process running?  (should
it check???)

profile the code and add comments.  This is the next project I'll be working
on unless someone comes up with some other ideas (please?  :-).
