Re: Question for implementors: XPath model and CDATA sections

On Friday 15 February 2002 02:53, Karl Scheibelhofer wrote:
> i just read some of the XPath sections again. i cannot figure out how
> the behavior of Xalan can be compliant with the XPath 1.0 recommendation
> (i really do not care about any working draft for a future version).
> maybe it is compliant with the mentioned working draft, but i guess it
> is not compliant to XPath 1.0.
> perhaps, it is very hard or not possible to be compliant with XPath 1.0,
> if you are working with DOM as it comes out of a parser.
> however, this situation is really unsatisfying for me.

There are definitely some incongruities between xpath, infoset, and dom and 
that should be improving in the future, but as you allude to, that's not 
terribly helpful in the present. However the two big rough spots that I'm 
aware of seem addressable. The first being in c14n: some DOM 
implementations (particular those based on "level 1") don't have proper 
namespace axis nodes. Most of this can be addressed by separating the 
namespace attributes out of the attribute list and creating one's own 
namespace axis. This present issue is tricky with respect to understanding 
the issue, but once that is done I think, as Philippe mentioned, one can 
emulate XPath (or the level 3 "wholetext" DOM) by sucking up nodes until 
one reaches the appropriate boundary. Neither of these are pretty, but they 
don't aren't unsurmountable.

I can certainly undertake to respond to any issues related to the specs and 
forward requests on, but I'm not sure what one can do with respect to Xalan 
since I'm not closely involved with that work. But I think there are a few 
implementors here using Xalan, so perhaps they could share their 
experiences or come up with an approach?


Joseph Reagle Jr.       
W3C Policy Analyst      
IETF/W3C XML-Signature Co-Chair
W3C XML Encryption Chair

Received on Friday, 15 February 2002 15:02:55 UTC