Re: Excessive layering in XML Schemas

The few times I've written XSD from scratch, I overlayered when I did  
not constrain the domain in advance by constraining either the use  
cases or by doing a bit too much research and trying to capture all  
that I read in my own distilled abstractions.  It is all too easy to  
wander into the marshes of side
issues or otherwise boil the ocean.

When using schemas from outside sources, it depends on my own role in the
organization.  I have a habit from the early years of reading the schema to
answer questions rather than trusting the schema-aware editor.  So  
when someone
comes to my office and asks if they can do something or why something isn't
working my first reaction is to open the schema and sort through the layers to
find the expression that actually determines an occurrence.  That gets gnarly
when there are lots of not-terribly-descriptive names and names of names.

It seems to come back to the too-many-incompatible-but-consensual  
points of view being accomodated in a single schema. I give the S1000D  
folks cred for breaking those up.

Object-layering usually implies running code.  A similar problem but  
not quite the same in that getting it wrong means rewriting a lot of  
functions or methods and that is never fun.  Here it seems one can  
stay simple as long as one is the only person writing that code.  In  
teams it gets harder as more classes are added and one has to explore  
to find the right one for the task at hand,

The real world is messy as Kay says and code is not the real world.   
On the other hand if the real world has to work too hard to use the  
code or process the results, it's time to go back to the design and  
rework it if possible. OTOOH, after a recent stint working in a  
tagging trench with a group attempting to dig out of a messy  
conversion (XML to XML), I am more than ever convinced that lack of  
training and actually resisting code is a bigger problem for those who  
procure XML services without evaluating the humans who provide them.   
I still see tagging armies taking months to do jobs that a properly  
configured and enabled XML system can do in minutes.   The costs are  
prohibitive and our governments pay them because they don't take the  
time to learn what is possible and what is silly.

len