[closed] Re: XInclude and xml:id processing in Javascript

At the 13 Oct telcon, we agreed that this was an informative message, not a comment on the spec.

Alex Milowski <alex@milowski.org> writes:
> I've been experimenting with the implementation in Javascript issues
> for both XInclude and xml:id.  I believe that one could make a useful
> implementation that could quite possibly be compliant but it would
> suffer from some issues:
>
>   * The DOM mutation events aren't consistently supported by the major
> browsers.  Specifically, because DOMAttrModified isn't support by any
> WebKit based browser and that is unlikely to change in the future,
> certain DOM operations would cause the xml:id implementation to be
> unable to track changes in ID attributes.  A native implementation
> would not suffer from this problem.
>
>   * Fragment identifiers typically scroll the document but would
> unlikely do so across browsers as expected.  This is because the
> script runs after or during the document rendering and the initial
> scroll has already been estimated.  A native xml:id implementation
> would not have this issue and I've demonstrated that with the patches
> I've made to WebKit to support xml:id.
>
>   * baseURI fix-up for XInclude works well via the xml:base attribute
> and so your browser must support that.
>
>   * invocation: How does this all get invoked?  Magical extensions?
> Script-tag inclusions?
>
> I haven't completed my implementation and so I may have more before
> tomorrow's call.
>
> I did notice several other issues we should consider relating to
> browsers and XML processing:
>
>    * If a document is modified via a script, what is the expected
> processing model?  For example, if I add an XInclude, is it automatic
> that inclusion happens?
>
>    * When content is included and xml:base attributes properly set the
> base URI of an element, relative references in host markup languages
> need take that post-processed base URI into account.  We should be
> explicit about that somehow.  WebKit based browsers do not do this.
> Firefox does.
>
> We need to be explicit about the interaction between post-processed
> infoset and the host application.

                                        Be seeing you,
                                          norm

-- 
Norman Walsh
Lead Engineer
MarkLogic Corporation
Phone: +1 413 624 6676
www.marklogic.com