xml-dev - RE: XLink transformations

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

Not asking would be a harsher problem.  No, the 
schemas are not secretly presupposed.  They may 
exist and are useful, but they do not of necessity exist.
The existence of schemata and the usefulness of 
declaring links are related.
  
There are more mathematically precise explanations 
than what follows (invariance under transforms).  
The importance of invariance under transforms is to determine if 
information lost by transform invalidates declared relationships.  
(Ie, did we snap the links?).  There is a lot of 
fun stuff in the bits on manifolds... but... 
I think I can confuse you without all of that. :-)

1.  XLink isn't an instance of XSLT.  XLink 
is a declaration of a relationship, not a transform 
or stylesheet application.  There is some cognitive 
overlap in that one way to make a link functional 
is to apply a transform.

2.  Unlike SGML, XML does not automatically 
presuppose the existence of a declaration (eg, 
DTD or schema).  In common practice, these 
probably exist because they are useful in 
several operational contexts.  There are issues in which 
one declares a link which depends on the target 
being in a particular position of the tree.  
In that case, a validation is useful.  One 
can more safely create persistent links given 
a schema, but again, it is left up to the 
designer to determine what safety is.  The 
issue is how much information is declared 
versus how much is derived in the operational 
context of resolving a link or transform.

3.  For a transform, one certainly does have 
a target in mind, but again, the existence of 
the schema is not predetermined since the target 
is an instance.  When thinking 
about transforms, it is useful to keep the 
Champagne/Urbana concepts of down and uptranslation 
in mind.  For example, a transform from an XML 
document to an HTML document is typically a 
downtranslation, meaning, information is lost 
in going to the target format.  Translating in 
the other direction is typically an up translation, 
meaning information must be added.

The reason to think about this is the namespace. 
If semantics are inferred by the namespace, that 
is, the namespace is a processing context, then 
ensuring the right names are used to create the 
correct context is at issue.  In this case, the 
existence of the schema can determine the context 
and from it, one can get the information needed for 
the target context.

1.  A schema declares an environment (a context).

2.  An environment can be thought of as a space 
of potential objects (any legal context).

3.  The schema declares the members for all of the 
possible operations that can be applied within that space.

4.  Operations which change that space can  
add or subtract members from the environment 
or instance.  If the environment is changed, 
the result is a loss or addition of potential 
members. 

5.  XLinks declare relationships that do 
not add or subtract objects from the space 
but can declare legal operations (constraints) over the 
objects within that space OR among different 
spaces by declaring what must exist in a given 
context for the link to be operationally valid.

6.  Operations may be reversible (non-lossy) 
or may not be (lossy).  (variant or invariant 
under transform).

7.  XSLT declares a transform operation which 
may use environmental constraints.  One use as 
stated is to determine by querying the declaration if 
the transform is valid in terms of the namespace 
context or by detemining if the transform operationally 
invalidates a link relationship.

The math majors can now proceed to chew me out for 
mistakes here.  :-)

Len Bullard
Intergraph Public Safety
clbullar@ingr.com
http://fly.hiwaay.net/~cbullard/lensongs.ram

Ekam sat.h, Vipraah bahudhaa vadanti.
Daamyata. Datta. Dayadhvam.h



From: Steve Boyce [mailto:SteveB@hbs.com]

This is: How is XLink anything other that a particular instance of XSLT?