Re: regarding the respsize draft: preferring glue of certain types

[Mohsen Souissi <Mohsen.Souissi@nic.fr>]

On 16 Jul, D. J. Bernstein wrote:
| Randy Bush writes:
| > this assumes that the transport available between the dns resolver and
| > the dns server is correlated to the transport between the application
| > client and the server.
| 
| No. The issue is delegations, particularly big delegations such as the
| DNS root servers' delegation of .com to 13 different DNS servers. Both
| levels are DNS; there isn't any ``application client'' here.
| 
| Suppose that a .com DNS server has both A and AAAA. I don't see why the
| root servers would ever want to provide both addresses. Specifically:
| 
|    * If the query shows up through IPv4, why bother sending the .com
|      server's AAAA address? Obviously the resolver is capable of
|      reaching the A address.
| 
|    * If the query shows up through IPv6, why bother sending the .com
|      server's A address? Obviously the resolver is capable of reaching
|      the AAAA address.

==> You know that the second scenario cannot happen today unless the client
can get some IPv6 address for the root server. As far as I know, even
if some root servers already run IPv6, their IPv6 address is not
present in the root zone. Putting AAAA glue records in the root zone
for .COM or whatever TLD is the only way to let iterative resolvers
reach TLD servers in both IPv4 and IPv6 as soon as they receive
referrals from root servers. Maybe you want to read the IESG
recommendation for RC Editor to add explicitely the note below to
RFC1886bis (becoming DS) in order to make things very clear:

"
RFC Editor note:

In the Abstract please change:
s/Domain Name System/Domain Name System (DNS)/
s/RFC1886/RFC 1886/

Please add this at the end of the Introduction section:
        The IP protocol version used for querying resource records is
        independent of the protocol version of the resource records; e.g.
        IPv4 transport can be used to query IPv6 records and vice versa.
"

In one word, the parent is not supposed to decide on behalf of the
client, which version of IP it has to use to reach the authoritative
nameservers for the child.

Hope this will help,

Mohsen.

| 
| Vixie was suggesting omitting AAAA's past the 512-byte limit in the
| first case, and A's past the 512-byte limit in the second, but I don't
| see any use of the extra records even if they all fit into 512 bytes.
| 
| ---D. J. Bernstein, Associate Professor, Department of Mathematics,
| Statistics, and Computer Science, University of Illinois at Chicago
| #----------------------------------------------------------------------
| # To unsubscribe, send a message to <dnsop-request@cafax.se>.
#----------------------------------------------------------------------
# To unsubscribe, send a message to <dnsop-request@cafax.se>.