Paul Wouters via Datatracker <[email protected]> writes:

>         Zone owners currently making use of SHA-1 based algorithms should
>         immediately switch to algorithms
> 
> I would use "should immediately rollover to algorithms" to avoid the illusion
> some inexperienced DNS admins might have that they can just "switch" the
> algorithm without proper prep work of doing a real roll over.

I changed to "roll" instead as "rollover" is kinda an odd term (though
heavily used by investment firms).

>        As a result, SHA-1 is no longer fully interoperable in the context of
>        DNSSEC. As adequate alternatives exist, the use of SHA-1 is no longer
>        advisable.
> 
> That should be, "SHA-1 as part of a signature algorithm". Because the document
> isn't obsoleting SHA-1 from DS hashing algorithms right?

Fair point, done.

> In the Operational Considerations, one could add a sentence about the
> difference of not supporting SHA-1 versus having a system that does not 
> support
> SHA-1. The first results in an insecure validation, which is okay. The second
> can result in ServFail, which is not okay. Something along the lines of:
> 
>       When not supporting or disabling SHA-1, care should be given by
>       implementers that the DNS software itself is made aware not to consume
>       SHA-1. For example, disabling SHA-1 at the Operating System level could
>       result in SHA-1 cryptographic failures within the DNS system, which 
> would
>       result in those zones failing, instead of the zones being treated as
>       unsigned/insecure

I'm not sure that fully works.  I'd rather not try to find all the right
lines where SHA-1 could be implemented and/or removed.  It is certainly
possible, for example, for the OS to have removed SHA-1 but the
application itself has its own implementation that it brings in.  In
fact, in Net-SNMP we do just this: we have a portion of openssl with
just the algorithms we need in the code base that can be enabled with a
flag to specifically avoid linking to a bunch larger library with a
bunch of stuff we don't need.

How about:

  Implementations following this advice need to ensure the associated
  operating system and software libraries they depend upon have SHA-1
  support.

Though I'm not sure how much that is actually adding. But it does warn
about the odd situation of today's environment I suppose.

-- 
Wes Hardaker
USC/ISI

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

Reply via email to