Well, that solution have a problem not present in vanilla DNS -- nameservers no longer correspond to administrative boundaries. And it does not do anything to fix the name collision problem. --vadim

From nanog-owner@merit.edu Mon Apr 22 19:26 PDT 1996 Return-Path: <nanog-owner@merit.edu> Received: from postman.ncube.com by butler.ncube.com (5.0/SMI-SVR4) id AA16645; Mon, 22 Apr 1996 19:26:03 +0800 Received: from titan.sprintlink.net by postman.ncube.com (4.1/SMI-4.1) id AA28280; Mon, 22 Apr 96 19:27:42 PDT Received: from merit.edu (merit.edu [35.1.1.42]) by titan.sprintlink.net (8.6.9/8.6.9) with ESMTP id WAA01557 for <avg@titan.sprintlink.net>; Mon, 22 Apr 1996 22:27:31 -0400 Received: (from daemon@localhost) by merit.edu (8.7.5/merit-2.0) id VAA29036 for nanog-outgoing; Mon, 22 Apr 1996 21:38:51 -0400 (EDT) Received: from chops.icp.net (chops.icp.net [199.0.55.71]) by merit.edu (8.7.5/merit-2.0) with ESMTP id VAA29030 for <nanog@merit.edu>; Mon, 22 Apr 1996 21:38:48 -0400 (EDT) Received: by chops.icp.net id <20637>; Mon, 22 Apr 1996 21:38:40 +0100 From: Sean Doran <smd@icp.net> To: pferguso@cisco.com, smd@icp.net Subject: Re: NAVYJOBS.COM Cc: john@interport.net, nanog@merit.edu Message-Id: <96Apr22.213840+0100_edt.20637+44@chops.icp.net> Date: Mon, 22 Apr 1996 21:38:31 +0100 Sender: owner-nanog@merit.edu Precedence: bulk Content-Type: text Content-Length: 2174 Status: R

| Personally, I like the <eighty-subdomains>.BigCompany.COM. approach; | it just scales better. Less messy. Tastes great. Me too. However, there appears to be some counterpressure to move towards what I perceive as some kind of system which lacks any real hierarchy in the sense of domains. Perhaps instead of trying to "fix" the DNS by creating huge new bunches of things to the right of the last normally-seen dot, we should move towards something completely different. Maybe instead of: <whatever>.COM being considered as three tokens and being looked up as: ask a ROOT nameserver for COM NSes ask a COM nameserver for <whatever> NSes ask a <whatever> nameserver for appropriate info (all modulo caching) instead we treat it as (leaving out the '<' '>' stuff out of laziness): <w><h><a><t><e><v><e><r><.><c><o><m> and look up like this: ask a ROOT nameserver for M NSes ask a M nameserver for O NSes ... ask a <w> nameserver for appropriate info that will scale to a huge number of generally unformatted labels for things. This strikes me as a practical way of moving towards ".Earth" and ".Alt" and thousands of other "top-level-domains". Other than technical scalability, the other general design goal for any changes to the status quo wouuld have to be administrative scalability. As a too-easy example, in this model that's essentially making sure an overworked or evil delegator for $origin ever.com won't break the ability of people wanting $origin whatever.com or $origin never.com from going about obtaining those names. Perhaps some form of automagic registration scheme should be available to allow for others to initiate any rightwards growth of names unless they involve certain charaters (like '.', perhaps, so as not to overly confuse people who are used to foo.bar.com being under the administration of whoever is in charge of bar.com). Again, I worry about being wasteful of other people's time here; this could be a non-issue for operators big and small for all I know, so I'll drop back into read-only mode for a bit, and let smarter DNS people and folks more in the trenches talk it out (or ignore it). Sean.