I've tested compliance with unknown types in known class. I tested for
both serving and resolving capabilities.
What I did do:
I considered implementations that return an existing 'unknown' TYPE
when queried with the same 'unknown' QTYPE compliant. To avoid false
negatives I've queried every instance for a known type as well. In
effect, this test was specifically for RFC 3597 sec 3 'Transparency'.
The bulk of popular (**) implementations are compliant:
Could we combine this with a use survey to see what the coverage is?
How much deployment of compliant servers is sufficient? I would
hazzard that 50% is way too low to be acceptable, 95% and it is a
non issue - the remaining servers are run by folk who would never
do Marid either.
I'd like to know where we are to +/- 10% at am 85% confidence level.
On the wildcards issue. We are only talking about restrictions on
cachable wildcards. Synthetic wildcards will still be possible.
Phill