fair%arpa.ucb-vax.arpa@BRL.ARPA (Erik E. Fair) (10/16/85)
I just saw Mark Crispin's quick flame in Header-People, and I applaud the sentiment, but thought the message needed to be expanded upon slightly. So I have extracted the relevant paragraph from the page break between pages 5 & 6 of RFC921, to wit: There are two communities that are taking slightly different courses in this transition. The DARPA research community is making the full transition. The DDN operational community is making the change in naming on the same schedule, but is not requiring hosts in the DDN operational community make the change to using servers at the same time (they can if they want to). The DDN PMO will establish a schedule for that change at a later time. The NIC will maintain a central table of all DDN operational hosts. This is restated a number of times throughout the document. The way I read this, the NIC really has to continue to maintain the entire host table until such time as the DDN PMO decrees conversion to domains for the MILNET also, in order for MILNET hosts to be able to reach hosts on the ARPANET. In turn, ARPANET hosts must continue to register their names with the NIC Hostmaster when they make changes, which includes the change to the new domain names. Further, I have a quick awk script that counts how many hosts are in each top level domain (according to the official name of the host, from hosts.txt), and here is the count from host table # 487 (10-Oct-85): arpa 1481 edu 317 com 33 uk 19 gov 11 org 8 To be fair, a number of hosts have their new domain name in as an alias, which indicates that they're in process of converting, and some of those `arpa' hosts are really MILNET sites, but this still ain't so hot, considering that the `arpa' domain was supposed to vanish on 15-Sep-85... glad I'm not a postmaster or liaison, Erik E. Fair ucbvax!fair fair@ucbarpa.BERKELEY.EDU
hedrick@topaz.RUTGERS.EDU (Charles Hedrick) (10/16/85)
The policy we are following at Rutgers is to register with NIC any host that we expect to send a sigificant amount of mail traffic to the Arpanet. I agree that it is not going to be practical for NIC to know about all of our PC's. Fortunately, most of them will not be authorized to send Arpanet mail, or we will be able to provide return addresses for them via some registered hosts. (For example, mail sent from a diskless Sun will show as coming from the file server it is using.) Even this cannot work forever, but we hope it will tide us over the period until resolvers are available for all of our major operating systems. We cannot always guess which hosts should be in the table. So I am prepared to accept requests from postmasters at other sites to register Rutgers hosts that are not currently registered. I agree with Mark Crispin that person@rutgers would be a good way around having to register all of our hosts. We are hoping to do that, but I suspect it will be a year before we can come up with a network-wide user database. And even then, such a database will probably only include hosts that are willing to cooperate. The problems in supporting organizational mailboxes are at least as much organizational as technical.