[DNSOP] Re: [Ext] Re: Sanity tl;dr for Multi-algorithm DNSSEC Requirements

[John Levine <[e-mail]>]

It appears that Paul Hoffman  <[e-mail]> said:
>On Jul 27, 2026, at 05:51, Shumon Huque <[e-mail]> wrote:
>> 
>> A new validator implementation could unilaterally impose that rule, sure.
>
>And it might. That would require that the validator had such a knob in its config, and would be implementation-specific.
>
>> But ideally, there should be some general rule for algorithm downgrade resistance that is signaled in the
>protocol, rather than special casing a rule only for a specific algorithm.
>
>THERE BE DRAGONS!

I agree but for a different reason: I doubt that most DNS operators know enough
about cryptography to set the signals in a sensible way.

More likely, a harried DNS operator will have read somewhere (or a consultant
will have told her) to say all of the signatures must validate because it's
"more secure." Then her customers start complaining because their web sites have
disappeared and their users report their browsers are saying SIR FALE or
something like that. Then the DNS operator turns off DNSSEC altogeher, and the
problems go away!

Considering that most of the Internet still doesn't think it's worth the hassle
to use DNSSEC. why point another gun at our foot? If you think that PQ is a
theat, add some PQ signatures. If you think that PQ crackers exist and you think
that one of them is aimed at you, stop signing with conventional algorithms.

R's,
John