Path: utzoo!utgpu!watmath!iuvax!ndcheg!uceng!kamat From: kamat@uceng.UC.EDU (Govind N. Kamat) Newsgroups: comp.mail.uucp Subject: Re: Domain Registration Message-ID: <1120@uceng.UC.EDU> Date: 2 Jun 89 01:08:46 GMT References: <1105@mailrus.cc.umich.edu> <7518@bsu-cs.bsu.edu> Distribution: na Organization: College of Engg., Univ. of Cincinnati Lines: 81 In article <7518@bsu-cs.bsu.edu> dhesi@bsu-cs.bsu.edu (Rahul Dhesi) writes: >In article >karl@triceratops.cis.ohio-state.edu (Karl Kleinpaste) writes: >>Then I would be able >>to take out the morbid special case in my mailer where I detect >>.BITNET addresses and foist such mail onto a gateway system of which I >>know little. >To send mail to a host on BITNET, no system outside BITNET should be >required to know any other than "hand this to the nearest BITNET >gateway and let it handle delivery". The domain scheme violates this >basic modularity principle by wanting hosts on BITNET conform to a >non-BITNET naming scheme. No one is forcing BITNET hosts to change their names within their net. However, if they wish Internet hosts to recognize who they are and what they are, evidently they have to conform to the domain naming system. The DNS violates no modularity principle; on the contrary, its zones encourage modularity and delegation of authority. >When you are talking about connectivity between two networks such that >each network is internally well-connected, and there are gateways >between the networks, it seems wrong to me to insist that an address >not contain information that will tell you which network it belongs to. Let's assume that tomorrow an "internally well-connected" network, XNET comes along with its own naming scheme. The only sane way to recognize an XNET address externally is to append an uncouth .XNET to it. And create yet another special case in the cf file to push such stuff on to a hardcoded XNET gateway -- one that I have to know all about before I can "let it handle delivery". If some day that gateway disappears, I can just see irate users: "How come my XNET mail is bouncing on this machine, and not on the one down the hall?" I completely agree with Karl's view on this point -- I don't wish to be at the mercy of XNET. >You can have "France" as the last line of the address on paper mail, >and it will get to France, where the French postal service will be 100% >responsible for figuring out where it goes. This is simple, efficient, >and requires no country to know anything about internal addresses in >any other country, or for addressees to register in some world-wide >database scheme. Your analogy is misleading. The name of the country has to be in a language the postal authorities in any country can understand, i.e., similar to the .XNET appendage above. Sure, no one needs to know about your internal addresses. To achieve that (I can't imagine why), just register the network as "XNET.NET." in the DNS, and broadcast a MX for *.XNET.NET pointing to your gateways. This is akin to telling the world how to get to the Central Post Office of France. That institution is not likely to move overnight but a hardcoded gateway very well might. With the domain system, all that needs to be done is update XNET's records, and everyone can see the change. The "*" represents your internal XNET addresses, which have thus not been registered in any world-wide database. Also, as per your desire, everyone everywhere can see that this is a XNET address. The idea of the DNS *is* to steer clear of a world-wide centralized database -- each zone decides what information it wants others to see. In the XNET you wish, seems to me that there is nothing much that others *can* see in any case, even if you wanted them to. That is, unless you get your hosts to change their over to the domain naming convention, at least for their transactions with the outside world. >Yes, the postal services could have come up with a scheme in which >addresses were entirely unrelated to location, but they are smart >enough not to try to maintain and update a world-wide database system >of domains when there *already* exists a well-defined set of names >(names of countries) that will do just fine. Domains are administrative, not topological. But that does not in any way preclude setting up a domain for XNET, using your own set of names as I mentioned above. >Rahul Dhesi >UUCP: ...!{iuvax,pur-ee}!bsu-cs!dhesi >Career change search is on -- ask me for my resume -- Govind N. Kamat College of Engineering kamat@uceng.UC.EDU University of Cincinnati Cincinnati, OH 45221, USA