> Do not write a parser which accepts only what is specified in
> RFC822 exactly and nothing else. Rather, write a parser which
> accepts and tries to understand everything which remotely looks
> like RFC822! Otherwise you will get into endless trouble.
 
Absolutely!  Sad to say, one of the peculiarities you should be very
patient with is IBM NOTE format -- not because it really comes close to
RFC822, but because so many people use it.
 
> Example of illegal RFC822 things you should accept:
>
> John Smith <JS@HOST>
> should correctly be
> "John Smith" <JS@HOST>
 
Actually, either of the above is correct "as is."
 
> Message-ID: <AAA&BBB@CCC>
> should correctly be
> Message-ID: <"AAA6BBB"@CCC>
 
Ampersand (&) is not a special character and need not be enclosed in
quotes.  I believe a strict interpretation of RFC822 forbids the angle
brackets < > around the message ID (because there is no "phrase" before
the brackets -- the ID should conform to the same rule as a "From" or
"To".)  Still, you should accept ANYTHING in the ID field.
Moreover you should accept:
 (John Smith) <JS@HOST>
which should be
 "John Smith" <JS@HOST>
 
> Message-ID: <AAA@BBB> (From my perfect mailer)
> should correctly be
> Message-ID: <AAA@BBB>
 
Again, neither form is truly correct because of the brackets, but the
text in parentheses is a COMMENT (acceptable absolutely ANYWHERE there
is white space) and would be ignored in parsing.