Re: [xml-dev] Rick Jelliffe quotable quote on the purpose of schemas

[Rick Jelliffe <rjelliffe@allette.com.au>]

Paul Spencer wrote:
> I agree totally with Rick Marshall, but for slightly different reasons. Very
> often, the data being validated has been generated by some computer system
> that has extracted data from a database. When receiving the data, a schema
> will usually be used to validate both the structure of the document (which
> has been created by the software) and the data types (that rely on the
> database content). These are completely separate issues and should be
> treated separately. I might choose to reject the document if the structure
> is incorrect, but process those parts that I can if the data does not meet
> datatype constraints, while quarantining the rest for some manual
> inspection.
>   
I don't think anyone is suggesting that all validation occur at once 
with fail-on-first-error.
(Err, except, of course, that that is how some grammar systems work.)

Or that different folks have different ideas and requirements for 
grouping together
constraints: the division of structural (what elements and attributes 
can go where?)
and data types (what values can data content have?) being an old favourite.

For example, in Schematron you can have one phase that validates only 
structure
and one phase that validates only datatypes if you want.

So me saying that it could be feasible to substitute a path based 
technology for
grammars does not in any way imply throwing away structural validation.
The fact that talking of weaning ourselves off grammars makes some people
say that they still want to be able to validate structures shows how 
embedded
the structure=grammar habit is.

Cheers
Rick Jelliffe