RE: XML Schemas: Best Practices

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

After twisting this a few times (like a 
Rubik's Cube), my opinion.  If indeed you want the 
choices to come from separately developed 
sources, you are as you say probably down to the 
choice, aka, Or group, aka, Switch.   Otherwise, 
assume a separate non-Schema, say human readable 
(aggggghghgh!) control exists.   

Why?  The other two by definition involve 
a structural and by inference, perhaps, 
semantic commonality which to exist 
must have some common control/rule/contract  
to constrain them.  In other words, 
you have defined a chicken or egg problem 
that has no solution until you 

1.  Decide that there is a structural relationship 
and a type relationship, meaning, you 
have prior knowledge, or 

2.  Decide that this is a choice among black boxes 
with all private information, therefore, you 
need no prior knowledge; just that all 
choices are equally probable and the only 
semantic is "containment".

For the constraints of your ideal, you have 
to know what you mean by "disjoint" or 
"independent" development.  Just as 
systems like COM need a common and 
enforced contract for discovery of the 
interfaces (e.g., must implement IUnknown), 
you preclude this contract with "independent" 
and "disjoint" development.  Therefore, 
content which is perfectly substitutable 
requires the container and by implication 
the processing semantic to also "not care".

The Choice is the laissez-faire or "only 
local implementation knows" or "or results 
unspecified or see Documentation" practice. 

The rest have a semantic of type.  These 
are a separate Best Practice question.

Len 
http://www.mp3.com/LenBullard

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