But they don't. Neither XSD nor DTDs have anything in their specifications that say 'This element (or whatever) expresses the intent or core of the constraint (e.g. in technical English or whatever.).'
The requirement/use case/intent/summary/purpose is not a first-class object. In fact, it is not an object at all in those systems. (A comment in XSD or DTD could contain a list of the developer's favorite Laotian restaurants; an assertion in Schematron could not, without being incorrect.)
DTDs and XSD are based on 1970s ideas of software engineering, where you make homemade little languages, and exchange is by private point-to-point treaty, and there is no mapping between what the user/editor uses and the document markup. They are languages good for unit testing not for pipelines/ETL where 'what was it?' is just as important as 'What is it?'; and not even for the common case where editing uses a COTS office tool that gets converted to XML and validated: there diagnostics couched in terms of the markup talk past what the user sees or knows.
So, if 'underspecified' is a euphemism for 'imaginary', then we agree :-)
Cheers
Rick
On Fri, 2016-05-06 at 12:47 +1000, Rick Jelliffe wrote:
>
[...]
> Also on Eliots point that dtds etc have limitations in their power:
All formal systems, of course, are limited, whether by the Arch of
Compassion or by the mathematical inevitability of completeness :-)
> one
> limitation is how to capture (without resorting to ad hoc comments)
> some constraints that are beyond the expressive power of the grammar
> or are contingent in some way. I believe Schematron is still the
> only system to take this seriously.
To some extent XSD does too, by allowing embedded Schematron rules :-)
or other extension mechanisms. DTDs do too, using processing
instructions, although it's terribly underspecified.
Best,
Liam
--
Liam R. E. Quin <liam@w3.org>
The World Wide Web Consortium (W3C)