On 16 November 1999, Karl Auerbach <[EMAIL PROTECTED]> wrote:
>
>> ... but is there a good solid techincal reason why multiple roots
>> wouldn't work?
>
>The issues concern records that contain textual names -- most importantly
>NS and CNAME records.
>
>The putative problem arises when the zone file creator puts a name into an
>NS or CNAME record that uses a TLD that is a different version of that TLD
>used by the person who consumes that record.
>
>In other words, if I say in an NS record that one of my servers is
>foo.blap.web, and I use IOD's version of .web, then if somebody who
>uses CORE's version of .web comes along and uses my NS record then they
>will end up with a mis-resolution.
But, since one of my criteria was that these competing roots would be
mutually exclusive, this couldn't happen, as the proprietorship of .web
would have been decided before the contract was signed.
>
>By-the-way, the limit of 13 servers comes from an ancient limit of 512
>bytes on the UDP data size used by DNS packets. (With modern IP stacks
>with working IP reassembly engines, one ought to be able to get up to 64K
>bytes in a single UDP frame - unfortunately not all stacks have been
>adequately tested, so a number in the 16K byte range is more common. And
>MTU's less than 1500 are fairly rare in the net these days except on some
>PPP links where the MTU has been cranked way down.)
Yep. And I believe that Windows is still shipping with MTUs optimised
for phone-based connections.
Thanks, Karl.
--
Mark C. Langston
[EMAIL PROTECTED]
Systems Admin
San Jose, CA
