Re: Comments on Opera's XHTML Module: Extensions to Form Controls

On Mon, 15 Sep 2003, Sebastian Schnitzenbaumer wrote:
>
> asking for input - I do have a question. You clearly say that you do
> not support this yet, and I wonder whether there is any latest thinking
> concerning real XML submission, ie. being able to control element
> names and even hierarchies for submission, and loading external
> data, presumably symmetrical in structure to submission.

My latest thinking on the XML submission is that having constructed
the submission document (the one with the predefined structure), the
UA would pass this document to the author's script, which can then
transform it via the DOM.

Note that this model doesn't depend on XPath (it doesn't use XSLT
either, as that would depend on XPath itself).

This is actually quite similar to how XForms works in the absence of
an explicit instance model, except that XForms' solution runs into
several problems:

 * It doesn't seem to allow changing the instance model prior to
   submission, so if there is no instance model, that's that (and even
   if there is, it's not clear how changing the instance model on the
   fly will affect bound form controls).

 * It doesn't allow names that aren't valid QNames.

 * Whether or not the implementation fails is dependent on the order
   of the controls. For example, the following works:

     <input ref="b"> ... </input>
     <input ref="/instanceData/b"> ... </input>

   ...but the following would raise an 'xforms-binding-exception'
   event:

     <input ref="/instanceData/b"> ... </input>
     <input ref="b"> ... </input>

   This seems counter-intuitive at best.

I've tried to avoid these problems in my design (it's one of the
reasons I haven't interpreted the "name" attribute as a polymophic
attribute, sometimes containing a QName and sometimes an XPath
expression).

Now, the suggestion of rewriting the submission XML using scripting is
of course not optimal, and I see the attraction of using an explicit
instance model. But the cost of doing this (XPath) is considered too
high for authors (and indeed some implementors) at this time.


> If this is the case I'd seriously have some input. :-)

Yes please! Your input is very much welcome.


> [...] the w3c-archive mailing list designated on the document for
> comments [seems] wrong for this kind of discussion, as we are e.g.
> turning the W3C archive into a forms working group style discussion
> [...] For now, I'm happy to discuss this directly with Ian, Howcome
> and Jonny Axelsson of Opera - which seems more appropriate - this
> being a document from Opera Software - and not a deliverable of any
> kind based on consensus and discussion within the membership of a
> chartered Working Group.

Because this document is not currently a deliverable of any W3C WG,
but directly affects W3C work, I think the w3c-archive mailing list is
the appropriate place to archive comments on this proposal. However, I
am quite happy to discuss this wherever people want, so long as it is
archived (so that I can refer to the discussion with URIs while
editing the document).

Cheers,
-- 
Ian Hickson                                      )\._.,--....,'``.    fL
U+1047E                                         /,   _.. \   _\  ;`._ ,.
http://index.hixie.ch/                         `._.-(,_..'--(,_..'`-.;.'

Received on Monday, 15 September 2003 09:28:06 UTC