Closed fpalombini closed 8 years ago
Per discussion on the list, this has been changed from a request to modify the current existing counter signature attribute to defining a new one. The reason for the change in the request is that the input for the new counter signature computation process should reflect the ToBeSigned structure of the COSE_Sign1 rather than the full COSE_Sign structure. This better reflects the circumstances that it is being designed for.
It's not clear to me why another way of doing this is being requested. Are both really needed?
Hi Mike, the goal behind this is again to reduce the size of the COSE message: instead of using a COSE_Signature structure (which itself contains headers and bstr signature value), we would like to use the simple bstr signature value directly. This would reduce the size of 3B: 1 for the array, 1 for the unprotected header and 1 for the protected header.
(the discussion about this in the mailing list is [here]([COSE] Issue - add the value type 'bstr' to counter signature)).
We propose to give the choice between the whole COSE_Signature structure or only the bstr 'signature' as "counter signature" value type. In CDDL, that would give:
Generic_Headers = ( ... ? 7 => COSE_Signature / bstr, ; Counter signature ... )
on section 3.1 and in Table 2. If we only send the btstr 'signature', we assume that the alg for the counter signature is also pre-established between nodes and is included in the "alg" parameter.