XSD - new VALID schema

[Rob Seaman]

Matthew says:

> Ideally we would have a VOEvent Data Model (described by the spec)  
> and then the schema is just one valid instantiation of that data  
> model: another could be a RelaxNG based description or a Schematron  
> one.

The question still remains of which is the master.  A specification  
could be written in some data modeling language rather than in  
prose.  I would suggest that, at least for VOEvent, this would be  
unwise.  Our standard is one that is meant to venture forth into the  
world and to encounter diverse projects and individuals.  Prose  
("natural language") has shortcomings, but it is after all the most  
widely supported standard available.  On the other hand, if some  
esoteric modeling language is the master - perhaps because it permits  
a richer expression of details - the human documentation of that  
specification may fail utterly to capture precisely those nuances.   
This will interfere not only with everyday usage but with detailed  
standards discussions in the future working group.

A truly useful ontology will be one that humans can use to converse  
among themselves.  One might assert that HAL's homicidal tendencies  
were the result of his semantic processing occurring distinct from  
his natural language processing, not from humans having asked him to  
keep a secret.  At least one hopes so, or the VO's security  
initiatives may prove mighty risky.

Rob