[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

RE: XInclude [was: XLink transformations]

• From: Paul Grosso <pgrosso@a...>
• To: xml-dev@x...
• Date: Tue, 25 Jul 2000 15:32:26 -0500

At 14:35 2000 07 25 -0500, Bullard, Claude L (Len) wrote:
>Is the sense of a merge operation that information 
>merged into another document loses its original 
>identity such that reversing the operation is 
>not guaranteed?

For XInclude, yes, reversing the operation is not
necessarily guaranteed, and that was not a goal of
XInclude (or the Infoset on which XInclude is built,
though a fair amount of information in the original 
can be captured by the infoset, and one could build
an application based on XInclude that did what was
necessary to allow most include operations to be
reversible).  

In fact, in general, W3C XML-related specs--including the 
DOM, XSL[T], XInclude, Infoset, XPath, etc.--operate on some 
data model and/or infoset that defines what information from 
the original source is maintained and what isn't.  Even XML 1.0 
itself defines attribute normalization and line end normalization.  
So complete fidelity in reproducing the original is almost never 
a reasonable expectation.  (I'm sure you remember RAST and the 
ESIS versus MSIS discussions from back in the early SGML days, 
Len.)  One person's bit of information is another person's markup 
minimization or yet another's editing tool or file system or 
operating system artifact, so pretending that "retaining everything 
in the original document" is even a coherent statement is a myth.

Exactly how much of the original is maintained and how much
isn't depends on the data model on which any spec is built,
and XInclude is built on the Infoset.

paul

• Prev by Date: RE: SAX and stuff ...
• Next by Date: RE: XInclude [was: XLink transformations]
• Previous by thread: RE: XInclude [was: XLink transformations]
• Next by thread: RE: XInclude [was: XLink transformations]
• Index(es):
  • Date
  • Thread