- From: Thomas Roessler <tlr@w3.org>
- Date: Mon, 9 Jul 2007 15:33:39 +0200
- To: Juan Carlos Cruellas <cruellas@ac.upc.edu>
- Cc: XMLSec <public-xmlsec-maintwg@w3.org>
On 2007-07-09 15:07:45 +0200, Juan Carlos Cruellas wrote: > It could initially thought that ds:Reference's Type operates at a > higher level than the ds:Object's mimeType, the first identifying > that what is signed is a ds:Object and the second identifying > that the media type of what is signed si for instance a pdf > document. If this is true then these attributes could be seen as > somehow orthogonal. That sounds like a plausible interpretation to me. > But this interpretation has one drawback: > 1. If the signature is dettached and the signed data object is not a child > of a ds:Object, then how to report its media type? The intent of the spec seems to be that an Object be used in these cases; see the discussion toward the end of 2.2. I don't see any obvious way, though, to integrate a media type with a reference to an external resource; that's probably an open point. (In fact, from an architectural point of view, there's an interesting tension between resources and their representations in here as well...) > If I am not wrong (and please forgive me if I am) the text in a > MIME media type identifier could also be seen as a relative URI > reference (a one having a relative-part= path-noscheme without > query and fragment). If we may set the ds:Reference's Type > attribute to a MIME media type then we may assert the media type > of the dettached data object to be signed, but then we should > make it clear that both attributes overlap in their purposes. That sounds a bit too much like shoe-horning an additional feature into existing markup. I'd suggest we note this in the wiki as an issue for further work, and not make any changes in the PER. Cheers, -- Thomas Roessler, W3C <tlr@w3.org>
Received on Monday, 9 July 2007 13:33:43 UTC