Hi Steinar,
Thanks for keeping me honest.
I knew that NSI had been making changes
to seperate the .COM servers from the root
servers, I just hadn't been keeping up on
the details.
Jay.
At 2/20/99, 01:23 PM, [EMAIL PROTECTED] wrote:
>> >But even you would have to admit that by adding lots of TLDs, extra
>> >orders of magnitude are induced in the DNS process, because queries
>> >are going further up the tree than they would otherwise be.
>>
>> In theory, I would agree with Greg.
>>
>> In this case, however, .COM requests *always*
>> go to the top of the tree (i.e. the root servers)
>> because the .COM zone is being resolved by the
>> root servers.
>
>No, it's not that simple.
>
>Some of the root servers ([a-i].root-servers.net) are also servers for
>.com. Some are not: [j-m].root-servers.net. Some .com name servers are
>not root servers: [j-k].gtld-servers.net.
>
>DNS is based on caching and load balancing, and name servers estimate
>round trip times and try to use the "nearest" server - ie. the one
>that answers fastest.
>
>Somewhat simplified, this means that for those name servers where
>[j-k].gtld-servers.net answer fastest, a .com request does *not* have
>to go to the root name servers.
>
>It's probably more correct to say that in a *majority* of the cases,
>.com requests go to the root servers because those root servers also
>happen to be servers for .com.
>
>Steinar Haug, Nethelp consulting, [EMAIL PROTECTED]
>
