headers
John R Levine | 1 Mar 2012 16:02

Re: Batch Multiple Query Packet

> It is also not very useful from the MTA point of view, since the MTA can't
> tell if A or AAAA records are missing because they don't exist or there
> wasn't enough space in the DNS reply etc. So in most cases (especially an
> IPv6 aware MTA getting no AAAA records in reply) the MTA still has to
> follow up an MX query with more queries to get address records.

Sounds like an AORAAAA record with flag bits would help there, too, so 
long as the authoritative server returned it in the addition section.  The 
client asks for the MX, and gets the AORAAAA so it doesn't have to do A or 
AAAA requests separately.

Alternate grosse hackque: some way to return an anti-result in the 
additional section that says there are no records of this type for this 
name, that can be negative cached.

Regards,
John Levine, johnl <at> taugh.com, Taughannock Networks, Trumansburg NY
"I dropped the toothpaste", said Tom, crestfallenly.
_______________________________________________
DNSOP mailing list
DNSOP <at> ietf.org
https://www.ietf.org/mailman/listinfo/dnsop
Permalink | Reply |
headers
Tony Finch | 1 Mar 2012 16:12
Picon
Favicon

Re: Batch Multiple Query Packet

John R Levine <johnl <at> taugh.com> wrote:
>
> Sounds like an AORAAAA record with flag bits would help there, too, so long as
> the authoritative server returned it in the addition section.  The client asks
> for the MX, and gets the AORAAAA so it doesn't have to do A or AAAA requests
> separately.

Doesn't solve the problem, because you can't infer anything about the
existence of a record based on it not appearing in an additional section.

> Alternate grosse hackque: some way to return an anti-result in the additional
> section that says there are no records of this type for this name, that can be
> negative cached.

That would help.

Tony.
--

-- 
f.anthony.n.finch  <dot <at> dotat.at>  http://dotat.at/
South-east Iceland: Westerly, backing southerly or southeasterly, 5 to 7,
perhaps gale 8 later. Rough or very rough. Wintry showers, occasional rain
later. Good, occasionally poor.
_______________________________________________
DNSOP mailing list
DNSOP <at> ietf.org
https://www.ietf.org/mailman/listinfo/dnsop
Permalink | Reply |
headers
John R Levine | 1 Mar 2012 17:13

Re: Batch Multiple Query Packet

>> Sounds like an AORAAAA record with flag bits would help there, too, so long as
>> the authoritative server returned it in the addition section.  The client asks
>> for the MX, and gets the AORAAAA so it doesn't have to do A or AAAA requests
>> separately.
>
> Doesn't solve the problem, because you can't infer anything about the
> existence of a record based on it not appearing in an additional section.
>
>> Alternate grosse hackque: some way to return an anti-result in the additional
>> section that says there are no records of this type for this name, that can be
>> negative cached.

New improved hackque: the AORAAAA record contains two fields, which 
contain the number of A and AAAA records at that name.  The server tries 
to put the A and AAAA records in the additional section, if the number of 
A or AAAA records is less than the count, the client knows they were 
truncated and has to ask for them explicitly, but in most cases won't.

The cache is allowed to synthesize a NODATA answer for A or AAAA if the 
AORAAAA value was zero, although I admit that the interaction with DNSSEC 
could be unpleasant.  Of course, the NSEC record also tells you what kind 
of records are present at a name, so it can equally well synthesize NODATA 
from that.

Regards,
John Levine, johnl <at> taugh.com, Taughannock Networks, Trumansburg NY
"I dropped the toothpaste", said Tom, crestfallenly.
_______________________________________________
DNSOP mailing list
DNSOP <at> ietf.org
(Continue reading)

Permalink | Reply |

Gmane