On Sun, Nov 16, 2025 at 11:44:35AM +0100, Philip Homburg wrote:
> In my opinion this is a very strong anti-pattern. It is not good for the
> state of DNSSEC if a signed zone is considered secure by one validator
> and not by another. That leads to fragmentation, make security analysis
> harder.

That's how DNSSEC works, though. A zone signed by an unknown algorithm is
treated as if it were not signed.  Any new or experimental algorithm will
produce a signed zone that looks insecure to older validators, it's always
been that way.

> For this reason we carefully manage the RECOMMENDED and MUST Implement status
> of algorithms to make sure that validators support an algorithm before it
> is recommended to use it in production.

Mark said "used over the public internet", not "recommended for use in
production".

An experimental algorithm obviously can't deployed for production use
yet, but that doesn't mean experiments should fail.  Currently, PRIVATE*
algorithms do not work correctly: they require klugey workarounds that
change the way validators behave, throwing unnecessary variables
into the testing process. It was a design flaw to have DS records that
fail to identify the signing algorithm in the keys they digest.

-- 
Evan Hunt -- [email protected]
Internet Systems Consortium, Inc.

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

Reply via email to