A further thought...

My too-clever (not clever enough?) hack can cause problems downstream,
if you have

1. auth (DNSSEC) -> 2. cache (DNSSEC) -> 3. cache (insecure) -> 4. resolver


3 sends a DO=0 ANY query to 2

2 sends a DO=1 version of the query to 1

1 does ANY minimization and responds with a signed NSEC (and nothing else)

2 does not know about ANY minimization

What does 2 return to 3? It can't send a signed NSEC because DO=0.

If 2 returns noerror/nodata then 3 will cache this (for all RRtypes) and
will fail to forward subsequent queries from 4 for other RRtypes.


If 1 uses Evan's ANY minimization then 2 will be able to safely drop any
DNSSEC rubric when passing the response to 3, without implying there are
no records at the QNAME. So 3's cache entry will not suppress queries for
other RRtypes.

To summarize: one constraint on a minimal ANY response is that if there
are non-DNSSEC records at the QNAME then the answer must include at least
one of them, that is, it must not look like a NODATA response.

Another way of saying this is that you can't just translate an ANY query
into a query for a fixed RRtype: as Evan said, you have to look at the
data at the QNAME so you can pick one of the existing RRsets and return
that.

Tony.
-- 
f.anthony.n.finch  <[email protected]>  http://dotat.at/
Fair Isle, Faeroes: Southeast becoming cyclonic 6 to gale 8, occasionally
severe gale 9 in Faeroes, becoming west 4 or 5 later. Moderate or rough,
becoming very rough for a time. Rain or showers. Good, occasionally poor.

_______________________________________________
DNSOP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsop

Reply via email to