The old rule, "Be forgiving in what you accept, but rigorous in what you
send out," suggests adoption of the following algorithm (which I use in
my mail processor):
 
1. If filetype is NOTE, then treat as possibly RFC733.
2. If the filetype is NOTE, then discard any continuation lines of the
   field.
3. If the first line of the address field has no '@' anywhere and does
   not end with a comma, then treat as possibly RFC733.
4. If the file is possibly RFC733, and the address field has at least
   3 words, and the next-to-last word is "AT", "At", "aT", or "at",
   then replace the final phrase "word1 at word2" by "<word1@word2>".
5. If, after steps 1-4, the field is invalid, then give up.
 
This algorithm, of course, would present a problem if IBM NOTE changed
by placing quotes around its usual address fields, but I have a feeling
we'll never need to worry about that!
                               John