RE: [xml-dev] Xlink Isn't Dead

[Ben Trafford <ben@prodigal.ca>]

         I have some comments on Michael Kay's response to my 
questions, but I'd like to state a point of position first:

         In my mind, the problems he's addressing (semantic value of 
links) are separate from the problems of rendering links. XLink's 
biggest problem is that it conflates both. The problem with things 
like RDF and Topic Maps is that they ignore display altogether. Thus, 
neither really works for link rendering, and XLink really doesn't 
work for semantically rich links.

         A lot of this comes down to the difficulties encountered in 
specifying good link modelling over the last twenty years. It's very 
difficult to come up with a language that's going to satisfy everybody.

         However, XLink had a solution to this, and I think it was a 
good one: linkbases. A well-designed linkbase with a "front-end" of 
sorts in the markup could be a good step to solving the problem. And 
languages like RDF and Topic Maps are a strong foundation for a good 
linkbase, methinks.

         I'm thinking we need to produce three things from these 
discussions, if they're to have any value beyond navel-gazing:

         1) A method of rendering links using existing stylesheet 
languages, starting with CSS. And this method needs to enable 
referencing to a linkbase.

         2) A decision on a good linkbase framework. This framework 
cannot preclude rendering of links.

         3) A bridge between items 1 and 2.

         Thoughts?

At 03:03 AM 9/25/2006, Michael Kay wrote:
>We shouldn't confuse a reference to a person with a reference to their home
>page on the web. Obvious, but I've read a lot of stuff in the "semantic web"
>space that appears to encourage this confusion: again, by trying to overload
>URIs to serve both purposes.

         I agree, and this leads me to wonder what the justification 
is for URI overloading. Any thoughts?

>In a document containing information about an employee, there's probably a
>reference to their manager. That needs to be a valid reference to a current
>employee of the company. In a press release issued by the company, there's
>probably a reference to the person who issued it. That can be an
>ex-employee. We need to be able to say declaratively what should happen to
>references when the objects they refer to change or disappear.

         Are you talking about change management, versioning, or 
error-handling, or some mix of the three?

>People often put versioning on the wrong side of the 80/20, and that
>accounts for a lot of our problems. The biggest problem with relationships
>is managing change (the "broken links" problem). You can't put that problem
>to one side and hope it goes away.

         Point taken. Again, I refer to my earlier three points, 
where I think most of what you're discussing needs to go into some 
kind of linkbase framework, and most of what I'm worried about needs 
to go into a styling language.

         I'm about 80% done a document on styling links via CSS. 
Anybody want to volunteer some time to work with me on the linkbase 
problem, assuming you folks don't think I'm out to lunch on the topic?

--->Ben

--->Ben