Single Origin (Re: Order in the working-group!)

[hardie@equinix.com]

Masataka,

> I'm now having an impression that you are talking about a set
> of default resolvers sharing a single unicast address locally
> in a single organization, which is already covered by:
> 
> 	draft-catalone-rockell-hadns-00.txt
> 
> Do you understand the difference between the root servers and
> local resolvers?

Yes.  Thank you again for your continuing concern.  

> Anyway, your proposal does not make the shared address have a single
> origin, which is an Internet-wide issue.

Yes, it does.  That's the point of section 1.5.  Is it possible that
this passage has confused you?

  The organization's border routers must then deliver the
  traffic destined for the root name server to the nearest
  instantiation.  To avoid internal routing difficulties, a static
  route to that network is recommended.

Reading it, I can see that the mention of a static route might cause
someone to believe that a static route from external peers to the root
server is required or recommended.  One of the reviewers of a
pre-release version of the draft indicated that it might be easier for
the organization to manage their internal routing if a static route
from the border router to the nearest instance of the root server
network were used.  That's certainly not the only way to do it,
though, and if it is confusing the larger issue, it should be cut from
the draft.

For those who did not pick up a copy of draft
draft-ietf-dnsop-shared-root-servers-00.txt before it was replaced,
the text is at draft-hardie-dnsop-shared-root-servers-00.txt .

				Ted Hardie