On Sun, 8 Nov 2026, Peter Thomassen wrote:

I think there should be implementation flexibility (pre-filling cache, local auth, ...), but I do think that Paul's points about re-using the normal cache are pretty compelling.

I therefore think that either the rootcache document should be relaxed, to describe its current method and recommend it, but allow others, or alternatively the localroot document should spell out clearly that cache prefill has some safety belts you'd otherwise have to think of when using a different approach, and therefore recommend it. Both cases would lead to the same document. If I was the editor, I would start from Paul's coz succinctness.

If I understood it correctly, there are multiple solution in both
documents:

1) use DNS to AXFR the root zone, pretending this is a hidden primary 
configuration.
2) download the zone over HTTPS (or external AXFR?) and feed it into the 
resolver cold cache.

Where 1) runs as kind of as authoritative and 2) runs as recursive

For DNS servers supporting running auth+recursive, 1) might be the best
solution. For DNS servers not supporting auth but only recursive, 2)
might be the best solution.

I would say merge these two into one document to prevent the duplication
of lots of text and explain both as valid implementation strategies.

With specific care to say ZONEMD should be used to properly authenticate
all the glue records.

Paul
_______________________________________________
DNSOP mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to