I have been working on a program to validate email addresses (syntax only) to filter out obviously bad addresses from internal files/databases prior to loading into a listserv list.  My questions are "what is a valid email address?", and "where do I go to find the authorative definition?"   I have come across two mailbox address references in the Internet Standards:

RFC821 (pp29-31), part of STD0010 SMTP Service Extensions. J. Klensin, N. Freed, M. Rose, E. Stefferud & D. Crocker. November 1995. (Format: TXT=23299 bytes) (Obsoletes RFC1651) (Also RFC821, RFC1869)
and
RFC822 (pp 10-11, 27, 44-47) in STD0011 Standard for the format of ARPA Internet text messages. D. Crocker. Aug-13-1982. (Format: TXT=109200 bytes) (Obsoletes RFC1653) (Also RFC0822)

However, these definitions are different.  Since RFC822 is extensively mentioned in the Listserv manuals, I originally wrote the program according to the RFC822 simple address type of "addr-spec", and further limiting the domain part to 
"domain-ref", excluding the more complex "domain-literal".  This should allow the following special characters in the domain part:
  !#$%&'*+-/=?^_`{|}~
The test load via bulk op in the web interface happened to contain underscore in the domain part, which Listserv rejected as invalid.  I tried a non-existent domain in WHOIS at http://www.networksolutions.com/cgi-bin/whois/whois/
which states that only alphanumeric and hyphen are valid.  Where is this written?

RFC821 states that sub-domains (between the dots) must begin with alpha, end with alphanumeric, and may contain alphanumeric and hyphen.  So I tried that, and the program rejected addresses with sub-domains beginning with numeric ,which Listserv had previously accepted.  

I'm sorry about the length, but my two questions remain.  Can anyone point me in the right direction?