xml-dev - RE: XLink transformations

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

The problem of ID maintenance is usually operational.  For 
relational systems, generated IDs are preferred in the role 
of the primary key.  Location can always be established by 
content value and often, secondary keys are used to establish
relationships.  Madly morphing values together to create 
keys resembling OLAP dimensions is an old dodge too.  

Identity, location are troublesome to conflate and names aren't 
very robust unless inside a context.  Positional linking 
is about the weakest kind I can conceive of.  You are 
right about links as view dimensions and we took this 
up with a formula I can't remember at the Hytime conference 
in Vancouver as part of the advanced IETM work.   A view 
is usually a temporary thing unless its 
contents are persisted into a new thing with a new identity 
or its contents are copied as an update or appended.  The 
notion of coupling is introduced to enable dependent variations.

I am confused.  A schema cannot declare that a value is a 
link or the type of link?  Otherwise, keeping a list of 
links to links to enable a rule for transforms would seem 
to be trivial if a little expensive (ie, couplings).  

Haven't the TechnoTeacher folks taken up the problems of 
ID and link maintenance in some detail in their GroveMinder 
product?  Steve Newcomb would be the fellow to ask.


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


-----Original Message-----
From: Rick JELLIFFE [mailto:ricko@geotempo.com]

The thing that should survive a transformation is not the text
specifying a XPointer or XPath but the pointer or path itself.   If
anyone cares to work out the details of what is needed for this, they
would be doing everyone a great service.

I think this is why is important that a schema language be able to
declare which value are XPaths or XPointers or whatever. (XML Schema
does not currently provide this directly, but I would imagine that the
neccessary types would be defined by other WG or body in short order.) 
Otherwise, how would a smart XSLT system know when to regenerate the
link. (And what happens if one end of the link disappears during the
transformation: what exception-handling should be in place for that.)  


Rick Jelliffe