- From: Scott Cantor <cantor.2@osu.edu>
- Date: Tue, 13 Apr 2010 15:40:27 -0400
- To: "'Meiko Jensen'" <Meiko.Jensen@rub.de>, "'Pratik Datta'" <pratik.datta@oracle.com>
- Cc: <public-xmlsec@w3.org>
> Don't you get such qname problems as well with rewriting prefixes to > "ns$i" or to base64(hash(ns-uri))? Yes, but we're handling that now (to the extent we can). The point is, what does "simplifying" the other cases buy us if we still have to address the prefixes anyway? > I mean, this is about > canonicalization, hence direct input to digest computation. If you'd do > a prefix rewrite in *any* way, such references to "xsd:string" will be > broken in the c14n'ed XML. Unless you rewrite them too. > Regarding attributes: you'll only need prefixes if an attribute is from > a different namespace than its element (rare case). It's not really that rare (xsi:type being one obvious case, SOAP being another). > Then you can define > an arbitrary prefix (e.g. using the "ns$i" approach) at that element, > which thus is only relevant for that single attribute(s) at this > particular element. Advantage is that you don't have to deal with issues > of identical prefixes mapping to different namespace uris. The point is that it defeats the simplification you're suggesting exists when the exceptions are quite common. -- Scott
Received on Tuesday, 13 April 2010 19:41:08 UTC