xml-dev - Re: XLink transformations

[Ben Trafford <ben@legendary.org>]

At 09:21 AM 7/18/00 -0400, tpassin@home.com wrote:
>This is a good example.  I think the approach to take is to create a new XML
>file, based on the xlink file and the source file(s), then to do any
>transformations on it afterwards.

         There are cases where this may not work, specifically, cases where 
the XSL stylesheet transforms the original links out of existence. At what 
point in the process do you process the links, and how do you bypass the 
XSL stylesheet to make them appear? And if you do bypass the stylesheet, 
what happens if you intentionally -want- the links to disappear?

>The harder part is when you use an XPointer expressions in an XLink, then
>change the document that is pointed to.  Chances are, the XPointer

         An icky problem. We resolved this, to some extent, by creating a 
tertiary information set (not a true infoset, but a homegrown document 
model to keep track of links), and then used that to parse against the 
generated document. When links disappeared, the user is informed. The depth 
of that informing (from keel over and die to a simple message) is up to the 
rendering application.

         A nifty thing is to have this tertiary information set keep track 
of where things are throughout the transformation. Since XSL 
transformations are predictable, it's not terribly difficult to follow 
where things are shifting in the process.

>table.  If the  key is compound (like (lastname,firstname,company)), and one
>part changes (John Smith changes his employer), what happens everywhere
>there is a pointer to this instsance of the key?

         As you mention later, the solution is to give the instance it's 
own identity. This need not be done explicitly, but can be generated by the 
linking/transformative engine.

--->Ben Trafford