"g" protocol implementation (giowvms):

If we get an INITC, and send an INITC which is lost, we'll go to run
state but they'll keep sending INITCs and not respond to our data.  
We timeout and drop the line.  Fix:  If we get an INITC in run state,
we should send an INITC, and then resend everything in our tx window.  
A mild problem, since all it really does is extend the "non error 
corrected" period (which begins when we send the username) into the
inita/b/c sequence.  

Sometimes we get an unexpected control packet at close time.  Don't
know what this is.  We still shut down neatly (exchange OOOOOO, etc.)
in every case.  Seems benign.  

Mailer:

A message that starts with a line of the form

	anything:

will confuse the mailer into thinking that the message already contains
RFC822 headers.  Avoid beginning mail with a line ending in a colon.  

Mail:

VMSmail is woefully inadequate for handling network mail.  Not our
problem, but...  gih is reportedly putting a Mail interface into News 6.0,
which will help immensely.