Re: xlxp-dev: Coffee Cup Ruminations on Behavior

Gabe Beged-Dov wrote:
> 
> I think that the values of When/Where need to relate to the Information
> Set concepts and something like a reference processing model. They need
> to be very precise. As an example (off the top of my head), the "where"
> axis could have values like below/top/external rather than
> embed/replace/new.

For the record, I am opposed to anything browser-specific being in the
core XLink spec. I am all for having specs on top of XLink that customize
it for different domains like hypertext browsing, print cross-references,
XML node inclusion, Java superclass referencing, etc. There are a hundred
different applications and it will actually confuse things more to
privilege one in the spec. 

People will spend their whole lives trying to shoehorn their applications
into the spec's mold -- that's what lead us to the "is not transclusion"
thread. Instead, we should keep it wide open and teach people that they
need to build layers on top, just as they need to build layers on top of
XML. auto/actuate or when/where strike me alot like having hard-coded
element names in XML for "paragraph" and "emphasis" because that's what
browsers need. XML is used for a lot more than browsers and so will XLink
be.

Also consider the need to experiment. If we leave the attributes OUT of
the spec and someone can implement them as a local convention or a
W3C-NOTE then we will have some experience with running code and will know
what we need. Right now I think that there are about 6 people on the
planet with experience in building generic markup-based browsers and that
isn't a sufficient foundation for hard-coding the behaviors we will be
stuck with for a Long Time.

Re: Infoset

I am hearing again and again that things should be done in terms of
information sets. Here, in XML-DEV, in XSL-List, etc. That's a good meme
to keep promoting. The next step is to make the concept of an infoset
*extensible* and *media-independent*. If you build a new specification for
behavioral attributes on top of my minimal XLink then you will probably
want those to be properties at the infoset level.

XML Structures' has started extending the infoset informally but I don't
think that the infoset's extension mechanism is formal yet. I haven't been
through it closely, though, so I could be wrong. XLink will also need to
extend the infoset.

I don't see how hypertext linking can work if there isn't an infoset for
the other media types being referenced. How can code or stylesheets deal
with a pointer into the middle of a PDF document if the mapping from PDF
documents to nodes is not expressed somehow?
-- 
 Paul Prescod  - ISOGEN Consulting Engineer speaking for only himself
 http://itrc.uwaterloo.ca/~papresco

"It's only a movie. People should get a life." 
 - George Lucas (http://www.nypost.com/news/9025.htm)

Received on Wednesday, 19 May 1999 13:09:49 UTC