Re: [xml] Incorrect handling of non-hierarchical URIs?

[brendandixon mac com]

RFC 3986 introduced ABNF for URIs. In doing so, they resolved  
ambiguities in RFC 2396, one of which was the handling of query/ 
fragment components of a URI.

After again reading both RFCs, I can see the confusion and am not so  
certain that I would've coded URI parsing differently than LibXML2  
has. In any case, since they've clarified the issue (which results in  
much more consistent handling), I'll file a bug with a suggested fix.

Brendan Dixon
brendandixon mac com
425.922.8798

On May 16, 2005, at 2:36 AM, Daniel Veillard wrote:

> On Sun, May 15, 2005 at 04:31:07PM -0700, brendandixon mac com wrote:
>
> > In making some tests and scanning the code (LibXML2 2.6.7), it
> > appears that xmlParseURIOpaquePart (and possibly the routines that
> > build URIs) is slightly incorrect: RFC 3986 (and I don’t believe this
> > has changed since RFC 2396 etc.) states that, for non-hierarchical
> > URIs, the path component consists of all characters up to the ‘?’
> > after which it is parsed the same as hierarchical URIs. That is, what
> > LibXML2 calls the “opaque” part is really a non-hierarchical “path”
> > and it may be followed by a query and fragment component.
> >
> > I have two questions: First, has anyone else encountered and fixed
> > this? :) Second, if not, does anyone have a sense for the impact such
> > a change might have on LibXML2 and its community? Parsing out the
> > query and fragment components outside of LibXML2 for these opaque
> > paths might be problematic if they contain interfering escaped
> > characters (that LibXML2 translates).
>
>   There is no registered/known bug concerning URI parsingC in the
> current version 2.6.19. If you have a problem, please provide an  
> example
> showing up in the latest version, and it will certainly be fixed. If
> the bug is confirmed and there is a patch associated usually the  
> fix is
> even faster.
>   I'm a bit surprized that there would be a big error in the  
> conformance
> to 2396, this has been checked quite a lot, maybe it's in a not  
> frequently
> used part. In any case try to provide an example, see testURI.c in the
> distribution as a starting point.
>
> Daniel
>
> --
> Daniel Veillard      | Red Hat Desktop team http://redhat.com/
> veillard redhat com  | libxml GNOME XML XSLT toolkit  http:// 
> xmlsoft.org/
> http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/