ROODE@BIONET-20.ARPA (David Roode) (12/23/87)
Discussing mail relay service per se wasn't my intent. The need or lack thereof for host name databases in name servers given the presumption of assorted relays was the concern. Suppose BITNET registered name service for domain BIT.NET, and named its hosts in the form x.BIT.NET . Affirmation would occur for all queries x.BIT.NET for all x, i.e. regardless of x. The response would be one or more valid records, each indicating a mail relay location. The point is the name server need know nothing about the domain other than the relay sites' identities. I guess I am making a hypothesis that says it is trivial to provide linkage to the Internet domain naming system for BITNET. No accomodation is needed in name resolvers. This obviates multiple BITNET kludges in effect in many local mail systems. It is upwardly compatible with any future evolution of BITNET. -------
OWENSJ@VTVM1.BITNET (John Owens) (12/30/87)
>Suppose BITNET registered name service for domain BIT.NET, and named >its hosts in the form x.BIT.NET . Affirmation would occur for all >queries x.BIT.NET for all x, i.e. regardless of x. There are a number of reasons that this is not a good idea, and that BITNET will (most likely) not adopt this scheme. First, it violates the NIC's standards that the only machines with domain names inside a ".NET" domain should be those belonging to the network administration, not those that are members of the net; this is the same reason that CSNET sites didn't just become site.CS.NET. Second, it means that mail to an invalid BITNET site name will have to be sent to one of the (very busy) relays and back for a failure indication, rather than having the domain name resolution indicate the failure before the message is ever sent. Last, but probably most important, it would mean that many BITNET sites would have more than one canonical domain name: the site.BIT.NET form and a regular organizational domain name. This site is VTVM1.BITNET, but is also VTVM1.CC.VT.EDU (in HOSTS.TXT and everything). Even BITNET sites not directly on the Internet might have MX records provided by their (otherwise Internet-connected) institution.... >It is upwardly compatible with any future evolution of BITNET. Even within BITNET, people are studying how to get away from the flat sitename model, using (organizational) domain names for all user-visible interactions and relegating the RSCS/NJE nodenames to the same status as numeric internet addresses. -John Owens +1 703 961 7827 Virginia Tech Communications Network Services OWENSJ@VTVM1.CC.VT.EDU OWENSJ@VTVM1.BITNET
jsol@BU-IT.BU.EDU (12/30/87)
Some of the BITNET software internal to BITNET is being converted to use domain specifications such as BUACCA.BU.EDU rather than BOSTONU.BITNET (to name the same host). I think this is better than using HOSTNAME.BITNET. --jsol
PAP4@AI.AI.MIT.EDU ("Philip A. Prindeville") (12/31/87)
That's fine, but a major transition like this will be slow and painful. If everyone cut over at the same time, then the effort of erecting firewalls her-and-there could be saved, but this is probably unrealistic... (I've noticed that the VM TCP/IP product from IBM has an SMTP mailer that doesn't like hostnames longer than 20 characters. This will have to be fixed before the transition... If it hasn't been done already.) Still, I guess it's the best course of action. -Philip