Hiya,

On 10/03/2012 04:11 PM, Sam Hartman wrote:
> OK.
> Let me rephrase.
> 
> I claim for a given message the following reactions are reasonable:

I'm assuming we're talking generalities here and not
just about abfab.

> 
> 1) require signature
> 2) ignore any present signature
> 
> and that
> 
> 3) verify if present and fail if verified
> is unreasonable.

Did you mean:

3) verify if present and fail transaction if not good-signature
   but handle transaction if no signature present as if a
   good signature was present

If so, yes that'd almost certainly be unreasonable.

> It may be that verify if present and tell a human what verification
> yielded is sometimes helpful but I'm dubious.
> 
> Are you arguing for 3?

I'm arguing that in general (perhaps not for abfab, not
sure), your three choices are too binary:-)

Signature checking can fail for various reasons (hash
compare fails, wrong key, don't have key, don't have path
for key, revocation, naming, ...) and the reaction to
failure of a signature check can also vary (fail the
transaction, subject transaction to more checks as in
DKIM, re-route transaction, log the transaction
somewhere special, ...).

The appropriate reaction to not being able to verify
a signature on a signed message can reasonably be
different to the reaction to an unsigned message.

For example, if I were using key continuity, then
the first time I see a signed thing I might allow
the transaction even though I've no good reason to
accept the public key. But the 2nd time different
checks will apply, with possibly different results.
For example, I might start rejecting transactions
that claim to be from that source but don't have
signatures that verify with the key I saw on the
first signed transaction from that source.

And again, the above are generalities. I'm not
saying how the argument might or might not apply
to abfab. (And I think we've wandered off topic
somewhat.)

> Or are you arguing that we should add a new security requirement to
> RADIUS that there are some messages that MUST be signed?

No, I've made no argument about RADIUS at all.

S.

> 
> 
_______________________________________________
abfab mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/abfab

Reply via email to