[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[wg-c] reposted for Paul Mockapetris

>Date: Wed, 15 Dec 1999 17:47:55 -0800
>To: Eric Brunner <brunner@world.std.com>, vixie@mibh.net
>From: "Paul V. Mockapetris" <pvm@a21.com>
>Subject: Re: Adding (just under) a Million TLDs to the DNS Root
>Cc: brunner@world.std.com, wg-c@dnso.org
>At 06:58 PM 12/15/1999 -0500, Eric Brunner wrote:
>>Hi Paul (Vixie) and Paul (Mockapentris),
>>There is a person on a mailing list who periodically makes assertions
>>such as the following:
>> >Technically, adding a new TLD to the root means adding a few
>> >lines of text with the character string and pointers to two name
>> >servers. There are no technical issues whatsoever as long as the
>> >number stays below one million, which it certainly will do.
>I think the issues are not so much whether one could get the data into the 
>root server, but rather whether the performance of the overall system would 
>be an issue.  Most of the technical tradeoffs are related to the growth of 
>the name space, advances in processor and communications speeds, whether 
>the DNS itself evolves, and how closely implementations follow each other 
>and the specs, and whether root servers are also second level servers (e.g. 
>Arguments for a smaller root are basically fewer hits on root servers, 
>which are somewhat of a limited resource due to datagram size limits.  If, 
>in the limit, the root servers have only the TLDs, and each TLD covers X% 
>of the name space, then having X be larger means the root should be 
>smaller. Of course, if I change the number of root servers by removing the 
>datagram limit, or have local copies or half a dozen other scenarios I can 
>achieve the same effect.
>If the DNS and its implementations didn't change, then the right question 
>would be how many subdomains would motivate a million TLD root?  Maybe we 
>need a million TLDs when we get to a gigazone DNS?
>But remember, these are optimizations - if there is a compelling need for a 
>million TLDs, we should just get to work engineering it.
>I'd actually like to see a reasoned discussion of this issue sometime, I 
>always wonder if I'm missing something.  At a recent pre-ICANN meeting, I 
>said something like "I'd feel safe adding a bunch of new TLDs, 10s or maybe 
>100s, using a slow-start kind of algorithm as we gain experience.  I don't 
>think there's any danger."  Several people told me I was dead wrong, but 
>couldn't say why.
>[Please note that none of these comments have anything to say about the 
>ability of the political process to add new TLDs.  My only suggestion would 
>be of the form that we give the political process some fixed time limit 
>(say 3 months for concreteness), to create at least 10 new TLDs.  If it 
>fails to do so, give each of 5 people (say myself, Paul V, Eric, Vint Cerf, 
>and Dave Crocker - or any other set of five people that historically have 
>been unable to agree on anything) the option of creating up to 2 TLDs each 
>of their own design.  Repeat the process.
>Even if you believe that four of the five folks are completely wrong, its 
>better than the current stasis.]
>>Could either or both of you do me the kindness of commenting on the
>>second sentence (not the speculative comma subordinated clause)?
>>Technical discussion of stability, latency, etc. will be read by at
>>least myself, Dave Crocker and Kent Crispin. Operational discussion
>>may have a larger readership, and policy discussion will presumably
>>be read by at least one third of this list.
>>Thanks for your kindnesses and cheers,
>>The mystery guest is Milton Meuller, one of the many voluable ones the DNS
>>Wars have brought to the surface of our little puddle, look for him on the
>>DNSO NC in the near future, as the ACM rep, and the mailing list is WG-C, a
>>no-exit sort of thing two layers down in the ICANN mud bath.