To:
dnsop@cafax.se
From:
Rob Austein <sra+dnsop@hactrn.net>
Date:
Thu, 03 Jul 2003 23:12:57 -0400
In-Reply-To:
<20030703225153.58514.qmail@cr.yp.to>
Sender:
owner-dnsop@cafax.se
User-Agent:
Wanderlust/2.10.0 (Venus) Emacs/20.7 Mule/4.0 (HANANOEN)
Subject:
Re: I-D ACTION:draft-ietf-dnsop-respsize-00.txt
At 3 Jul 2003 22:51:53 -0000, Daniel J. Bernstein wrote: > > The notion that we're stuck at 13 servers is silly. For example: > > com NS a.gtld-servers.com > com NS b.gtld-servers.com > com NS c.gtld-servers.com > com NS d.gtld-servers.com > a.gtld-servers.com A ... > a.gtld-servers.com A ... > a.gtld-servers.com A ... > a.gtld-servers.com A ... > b.gtld-servers.com A ... > b.gtld-servers.com A ... > b.gtld-servers.com A ... > b.gtld-servers.com A ... > c.gtld-servers.com A ... > c.gtld-servers.com A ... > c.gtld-servers.com A ... > c.gtld-servers.com A ... > d.gtld-servers.com A ... > d.gtld-servers.com A ... > d.gtld-servers.com A ... > d.gtld-servers.com A ... This has been suggested before, and it's an appealing idea, but it assumes that resolvers will try every address of every listed server. RFC 1035 is, unfortunately, not entirely consistant about this: some parts of the resolver description talk about putting all of the addresses of all of the listed name servers into the SLIST, but others talk about removing a "name server" (as opposed to one its addresses) from the SLIST. So there's an open question as to whether a resolver should treat your example as 16 distinct addresses to try, or should treat it as four name servers to try; one could find support for either choice in the spec. #---------------------------------------------------------------------- # To unsubscribe, send a message to <dnsop-request@cafax.se>.