XML Signature 6.6 needs careful review and clean-up

Working on the algorithms cross-reference, I'm going through section  
6.6 of XML Signature with a relatively fine comb.

There is some significant processing that is dispatched off whether a  
transform takes a node-set or an octet-stream as input.  However,  
section 6.6 is (to put it mildly) conflating special processing for  
node-sets (or octet-streams) with re-telling pieces of the reference  
processing model.

I suggest that, for XML Signature 1.1, we do the following:

- Call out acceptable inputs (and output) in the top of the  
subsections of 6.6, along with the Identifier.
- Be very clear whether or not we specify special behavior when given  
the "wrong" input.
- clean up language where it's all too tortured.  The enveloped  
signature transform comes to mind for that.

--
Thomas Roessler, W3C  <tlr@w3.org>

Received on Monday, 9 February 2009 13:34:19 UTC