barmar (08/21/82)
Gill, you are missing a major point about this standard. It has nothing to do with what the user types or sees. Much of what you said went back and forth for the last several months on header-people (and other mailing lists, I'm sure). The RFC's define an interchange protocol, not an interface protocol. The user should be able to say "gill@ccc", and then the mailer software will figure out what the "right" thing to put in the envelope and header is. This process may involve contacting name a name server in order to find out the route to a particular host. The important point is that the user need not know about this. ~v (oh well, no editor escape) To continue, the domain scenario that you suggested is already in use. The full name for the Phoenix Multics site known as System M is M.PCO.LISD.HIS The domains in this are HIS - Honeywell Information Systems, LISD - Large Information Systems Division, PCO - Phoenix Computer Operations, M - System M computer. As an example of an implementation that does exactly what I said above, i.e. allowing the user to specify any allowable address, I just said send_mail barmar -at cisl.lisd.his to Multics (CISL is the Cambridge Information Systems Laboratory of the Large Information Systems Division of Honeywell Information Systems). After I finished typing the message, Multics responded Mail queued for barmar at MIT-DEVMULTIC.ARPA This is a perfect example of a sites that belong to multiple domains, and it works fine, and completely transparently to the user. As to header munging, what's wrong with that? It has been decided that the header will be the place that information is kept, so it is reasonable that mailers be able to put things there. If you don't want to see all the weird fields that are used, then tell your mail reading program not to print them; if it isn't smart enough to do that, then make it smarter. Don't complain about the transport system, though; the headers are mostly for them, although the protocol is nicely relaxed enough to allow arbitrary header fields.
thomas (08/21/82)
Just an aside: RFC stands for "Request for Comment", and just because RFC nnn has been published, doesn't mean that it's cast in concrete. Look what happened to RFC 733, after all. Looks to me like we're not done yet. =Spencer