Re: [9front] [Patch] ndb/dns: DNSKEY and OPT RR types

[Jacob Moody <moody@mail.posixcafe.org>]

On 12/18/20 9:21 AM, cinap_lenrek@felloff.net wrote:
> i wonder if the opt record should really be an RR and exposed
> to the caching code. it is just a hack to communicate protocol
> extensions by query, no? its not really part of the answer (->ar)
> to the dns query? in that case attaching Topt records to the
> queried domains seems very wrong to me.

Thank you for the feedback.
 From my understanding, the opt record is only used to to communicate the protocol extension.
 From my testing with other DNS servers the opt RR is included in the additional section (->ar)
of the response when the client sends a request that includes it's own opt RR in the additional section.

Are your concerns regarding the caching the fact that opt RR's coming from upstream would be cached?

On 12/18/20 6:25 AM, hiro wrote:

> why not just one line maxudp = 4096 ?


I agree, I will fix that in a revised patch if this behavior is still desired.
If this whole EDNS stuff is undesired I can rip that part of the patch out, but still think
the code should catch the Tdnskey RR's, as having those get caught in the default case was causing some
parsing issues(or perhaps there is a another nice way of handling that, I am not sure)

Thanks,
Moody