Hi Petr, (no hats)
On 10/1/26 05:25, Petr Špaček wrote:
On 01. 10. 26 14:12, Mukund Sivaraman wrote:
On Thu, Oct 01, 2026 at 12:43:11PM +0200, Petr Špaček wrote:
b) should it trigger FORMERR (= whole message ignored)
I missed this part: perhaps this is a BIND-specific question, i.e.,
Please let's not delve into implementation details here. My question stands at
the protocol level:
Does appearance of such RR mean the whole message is invalid or not?
I would say that "no use case" doesn't make it invalid in the protocol sense.
For example, for CSYNC it's conceivable that some special meaning for the empty
type bitmap would be defined at some point. That does not even seem too crazy,
as you wrote:
I have found zero samples of CSYNC/NSEC with empty bitmaps in the traffic capture between resolver and auths.
So I don't think it should be considered "intransmissible", and it would be
reasonable for an auth to not interpret that bitmap, and just serve the CSYNC record
uninterpreted.
For NSEC, it may be harder to imagine that an empty type bitmap could make
sense, even eventually in the future, because it seems to be an inconsistent
claim (the NSEC record disclaiming its own existence). But my impression is
that's for the validator (if anyone) to flag, and as far as the auth is
concerned, I see no reason to apply logic that differs from the CSYNC argument
above. (Same for other invalid combinations, e.g. NSEC w/o RRSIG or vice versa,
or NXNAME with MX etc.)
Curiosity: What do validators do with such NSEC records -- ignore the
inconsistency, or SERVFAIL on the contradictory proof?
Best,
Peter
Might be in AXFR/IXFR context, or even some sort of crazy UPDATE (or anything
else, really).
_______________________________________________
DNSOP mailing list -- [email protected]
To unsubscribe send an email to [email protected]