wales@ucla-locus.ARPA (Rich Wales) (10/16/85)
In connection with the recent discussion on whether or not all the new
domain-style host names should be in the NIC table, my name was invoked
in such a way as to suggest that I supported a position which in fact I
oppose. I am certain this was simply an innocent misunderstanding. In
any case, let me set the record straight.
(1) I believe that EVERY host name which is used in a mailing address
should appear in BOTH the domain data base AND the NIC host name
table. Any host which doesn't have its name in both places is inev-
itably going to encounter problems getting mail from some portion of
the net.
When I posted my celebrated host-name study (and the accompanying
set of messages to individual errant hosts) last month, by the way,
among the kinds of hosts I flagged were those using domain-style
names that didn't appear in the NIC table.
I do NOT support the idea that hosts should be kept out of the NIC
host name table in order to put pressure on hosts which haven't yet
converted their software.
(a) Just because a given host is still using the NIC table does not
necessarily mean that its administrators are lazy, apathetic, or
incompetent.
(b) In the case of the MILNET, for instance, it has been correctly
pointed out that MILNET hosts are not required to convert (or is
it, "are required not to convert"?) to the domain system for
some time yet.
(c) And in any case, whenever we try to exert pressure in this way,
the real losers in the end are the end users who are unable to
get their mail through.
(2) If an organization has a second-level domain, I believe that they
should be allowed to assign that second-level name to one of their
hosts (so that it can act as a mail gateway for the organization)
and have that name listed in the NIC table in addition to the host's
regular third- or lower-level name.
I am aware of at least one organization whose mail guru told me that
the NIC had refused to list both "xxx.COM" and "yyy.xxx.COM" (for
the appropriate values of "xxx" and "yyy") as host names for its
mail gateway machine. Perhaps this request was refused for some
other, unrelated reason. If, however, it is in fact the NIC policy
to turn down such requests on princple, I believe this policy should
either be defended publicly and cogently, or else discarded.
Actually, I suspect this policy is no longer being enforced by the
NIC (if indeed it ever was) -- since a quick scan of the newest host
name table shows several instances of second- and third-level names
for the same host -- including at least one recent addition.
--
Rich Wales // UCLA Computer Science Department // +1 213-825-5683
3531 Boelter Hall // Los Angeles, California 90024 // USA
ARPA: wales@LOCUS.UCLA.EDU -or- wales@UCLA-LOCUS.ARPA
UUCP: ...!(ucbvax,ihnp4)!ucla-cs!walesnetinfo@bbn-jade.ARPA (Postmaster + BITINFO) (10/17/85)
In reply to: Date: Mon, 14 Oct 85 17:54:21 PDT From: Rich Wales <wales@LOCUS.UCLA.EDU> To: Header-People@MIT-MC.ARPA Subject: Re: Domain names in the NIC table ... (1) I believe that EVERY host name which is used in a mailing address should appear in BOTH the domain data base AND the NIC host name table. Any host which doesn't have its name in both places is inev- itably going to encounter problems getting mail from some portion of the net. ... At Berkeley we have run into the problem of not all hosts being on a packet (eg. TCP/IP) net. In addition to hosts on our local ethernets, we support mail service to hosts that are only on our BERKNET, BITNET, or linked by dialup UUCP connections. These hosts are in the Berkeley.EDU mail domain, but we cannot register them in the NIC host table because they do not have Internet network addresses, nor can they support SMTP. However, they can be supported by a mail domain nameserver. Prehaps we need a separate nameserver for mail domains, or away of indicating a domain name is not on the physical internet? Bill Wells
drockwel@CSNET-SH.ARPA (Dennis Rockwell) (10/17/85)
From: Postmaster + BITINFO <netinfo%ucbjade@ucb-vax.arpa> Date: Wed, 16 Oct 85 21:11:54 pdt Subject: Re: Domain names in the NIC table [ ... ] Prehaps we need a separate nameserver for mail domains, or away of indicating a domain name is not on the physical internet? Bill Wells There already is: the domain servers for those hosts should set up MF records pointing to the relay host that is on the Internet. CSNET plans to use this mechanism to support our hosts on PhoneNet (with the MF pointing to RELAY.CS.NET aka CSNET-RELAY.ARPA). Thus, people using resolvers can send mail to, for instance, JoeStudent@Foo-U.EDU, and their mailer should ship the message off to CSNET-RELAY. This implies that everybody who is rewriting their mailers to use a resolver should cause them to ask for type=MAILA records *first*, then type=A records. This way we can give the appearance of connectivity for mail hosts not directly on the Internet. Of course, this can used for UUCP, BITNET, MAILNET and whoever else wants to do this. Berkeley can put in an MF record for, say ERNIE.BERKELEY.EDU, which points to BERKELEY.EDU, or whatever. We (BBN) are considering using this mechanism to keep people from trying to send mail to TACs (yes, it happens). Our local TACs would have MF records pointing to BBN-UNIX (aka UNIX.BBN.COM), but PTR and A records giving their IP address, so that telnet servers can discover the name of the source of new connections. Let me repeat, because it's *important*: mailers should look for MAIL records first, address records second. Dennis Rockwell CSNET Technical Staff