mo@seismo.UUCP (01/28/84)
The current proposal to use telephone area codes in uucp host addresses is a reasonable idea, assuming we deal with the North American chauvanism by including country codes or some other thing. However, calling this "domains" is probably incorrect; it is most-likely an example of what is generally called "area addressing" or "sub-area routing." While it is easy to see that "Domains" are related and similar, that word already has very specific meanings in a larger context and I suggest we not overload it or further blurr its meaning. The difference between domains and area addressing lies in the notion that domains are, in general, distinct universes, either because of the mail protocols used and/or because of administrative partitions. Domains are the basis whereby a small number of distinct entities can agree to interconnect mail universes with only a very few people directly involved. I would suggest that Usenet and the UUCP network on which it is based want to present a single domain to the rest of the mail world if interconnectivity is to be fostered. On to area addressing: the idea is simple, like the one proposed. (The name "area code" did not come from nowhere, nor did the name "area addressing.") If the namespace of the network is so large that maintaining routing information everywhere is onerous (especially like in implictly or explictly source-routed networks), the notion is to adopt a heirarchical partitioning of the address: an area address, and an address within the area. Then hosts only have to deal with two cases: (1) direct destination addressing and routing for destination addresses within its area, and (2) knowing how to forward traffic for areas other than its own. If that this point you ask "How is this different from domains?" I have to answer "Technically, very little. Politically, a great deal." So, I support using a pre-existing metric for adding area addressing to UUCP host names, and the telephonic country and area codes are well-established choices. They are even in-line with CCITT guidelines!! But I also strongly suggest abandoning the use of the word "domain" except in very explicit contexts and making such an addressing decision an UUCP network internal matter. Yours for more reliable mail, -Mike O'Dell
jhh@ihldt.UUCP (01/31/84)
It is questionable whether telephonic codes are appropriate for data networking. CCITT, in Recommendation X.121, has specified Data Country Codes (DCC) for each nation. For example, the United States has DCC of 310, the Netherlands has 204, the U.K. has 234, and Canada has 302. The number is further broken up as a 4 digit Data Network Identifier Code for nations with more than 1 data network. In the case of numbers referenced relative to the public telephone network, the number is 9 + Telephone Country Code + National Telephone Number. I suspect that DCCs will be more important than POTS numbers in the not-too-distant future. Something that locks us into a numbering plan that will superceeded should be avoided. John Haller AT&T Bell Laboratories
honey@down.UUCP (code 101) (02/01/84)
c'mon mike, use your imagination. we should be using ZIP codes, not area codes! in fact, if we use the 9 digit versions, we can all throw away our terminals and let USPS deliver the mail. peter honeyman
sew@minn-ua.UUCP (02/04/84)
#R:seismo:-56300:minn-ua:8900001:000:2182 minn-ua!sew Feb 3 14:21:00 1984 Those who are in favor of using area codes for domain/subdomain addressing seem to want to aim a message at a geographic region, rather than addressing to an organization or site. Should problems be expected if sites can be in more than one subdomain? I think not. Assuming that the net usually is expanded and backbones are not removed very often, the route-sensitive user.site.site.subdomain.domain addressing will work most of the time. If backbones or major links within a geographical region are altered somewhat more frequently then user.site.region.domain will function often, and probably should be available for when major changes do occur. Although having too many options could make the addressing too complicated, I think that there should be some kind of geographical addressing. Allowing both kinds of addressing should be possible by having aliases in routing tables and by a site knowing which geographical areas it is included in. Area codes can be too limited for geographical addressing. Example: If you wanted to contact xyz someplace inside Digital Equipment Corporation near Boston, you might after many attempts find xyz is actually in Rhode Island or in west Massachusetts' area code. Being able to send to that general region would result is some more effort by routing software but a greater chance of getting through with one message. How to address? nw | nc | ne ------------------------ wc | cent | ec ------------------------ sw | sc | se The above nine subdomains would cover a geographical subdomain. Overlapping at the edges is recommended. (Should the four cardinal points be allowed as addresses?) A site which was involved in transferring a message would know what subdomains it was in. Example: a site in Harrisburg, PA would be in ec.northamerica, se.northamerica, ec.usa, ne.usa and ec and ne. Giving only a geographical subdomain code would probably assume the address was in the current country. A site in alaska would be in both alaska.usa and nw.northamerica, but not nw.usa; a distance of separation rule probably needs to be made. from the analog digits of Scot E. Wilcoxon ...ihnp4!umn-cs!minn-ua!sew
smk@axiom.UUCP (Steven M. Kramer) (02/07/84)
No, Peter! ZIP codes are out. Since everyone will soon have a personal 370 or 3Bxx (just WHEN will the anouncement be made??), we should go with SSN. Then my creditors can send me electronic warnings that I can & d away. No fuss! No messy phone calls! Who DOESN'T have my SSN? --Just thinking of where the trends are going ... -- --steve kramer {allegra,genrad,ihnp4,utzoo,philabs,uw-beaver}!linus!axiom!smk (UUCP) linus!axiom!smk@mitre-bedford (MIL)