Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!shadooby!accuvax.nwu.edu!nucsrl!gore From: gore@eecs.nwu.edu (Jacob Gore) Newsgroups: comp.mail.uucp Subject: Re: Domain Registration (was Re: rewriting FROM: lines) Message-ID: <3400012@eecs.nwu.edu> Date: 27 May 89 22:24:47 GMT References: <5794@microsoft.UUCP> Organization: Northwestern U, Evanston IL, USA Lines: 68 / comp.mail.uucp / les@chinet.chi.il.us (Leslie Mikesell) / May 26, 1989 / >Smail passing off >to a smart host works fine for sending, but how does a recipient, perhaps >on BITNET, get a response back to you if you aren't in the maps? >... an address like: > a!x.y.z!b!you (replace ! with your favorite character) (That's not an address, it's the end of a path. Your correspondent would still need to figure out how to get to 'a'.) >would mean for the local machine to send to machine a (which it must >know about). Machine a forwards to x.y.z by its choice of methods. >Machine x.y.z forwards to the machine b that it knows about, HOLD IT -- how does 'x.y.z' "know about" 'b' if 'b' is not in the maps? >This approach would avoid the necessity to track the transient connections >of every PC running uupc in the world yet allow the stable machines >to optimize routing to each other. On the contrary. If 'b' is my (God forbid :-) PC, and I move its connection to a place where it's no longer forwarded to by x.y.z, but by l.m.n, then I have to tell everybody who writes to me to stop using you-figure-out-what-goes-here!x.y.x!b!me and start using you-figure-out-what-goes-here-instead!l.m.n!b!me If I call it 'b.c' instead and register it with the domain name service, I just need to tell whatever nameserver on the net is handling the '.c' domain the new information. 'me@b.c' before the move, 'me@b.c' after the move. >A FROM: line line x.y.z!a!me >could then be replied to from anywhere. Unless 'a' itself moves to where 'x.y.z' no longer forwards to it, and 'q.r.s' is doing it instead. >I suspect that it is intentionally not done this way because the domain >nameserver would automatically get stuck with providing name service and/or >forwarding for anything that connects downstream without having the >control process of putting the machine in the domain. "THE domain nameserver"? The Internet nameservers (that's plural, very plural) form a distributed database. Their only purpose is to store this "knowledge" that you seem to expect some "smart hosts" to acquire out of nowhere (about sites that are not listed in any registry, such as the name service database or the UUCP maps). Name servers don't forward mail. They only provide answers about how to deliver mail to a given address. They don't "get stuck" with providing name service -- they exist for that purpose. They have nothing to do with the actual transfer of mail -- they just provide information about which host will accept (and, if necessary, forward) mail for a specific address. The domain naming scheme is there to make it possible to have a distributed (as opposed to simply duplicated) database. You say that 'x.y.z' knows about 'b'? Fine. Then you have two options: give 'b' a domain name so that 'x.y.z' can share that knowledge with other sites, or you go and share it with every one of your correspondents. Personally, I prefer giving people an address that they can send to, without having to figure out the route for the message to take. All I ask of mail systems is to deliver it faithfully to me, and not muck with my return address when the message is from me to somebody else. Jacob Gore Gore@EECS.NWU.Edu Northwestern Univ., EECS Dept. {oddjob,chinet,att}!nucsrl!gore