Re: Piece-wise verification

Simon St.Laurent (SimonStL@classic.msn.com)
Mon, 1 Jun 98 14:04:53 UT


[Forwarded for Peter Murray-Rust]
[Paul Prescod]
>Presume XSchema rules are similar to XSL rules, and are of the form:

[... XSL-like rule/pattern clipped ...]

>Patterns would be very simple in the first version. (e.g. just the element
>type name) Later versions could be aligned with XSL so that that code
>could be reused. Constraints would also be simple and get more and more
>advanced with subsequent versions.

I like this proposal. As Paul hints, we shouldn't attempt more than the
equivalent of a DTD in V1.0, and I suspect we might end up making the rule
optional (and thereby defaulting to the element). It strengthens the
XSL-like language as being useful for more than stylesheets [i.e. schemas,
and presumably transformations.] I suspect that it gives us a good deal of
the power we shall need 'for free'. For example if we want an element to
occur between 5 and 8 times the pattern could be:
descendant(5,FOO) <!-- must be verfied -->
descendant(9,FOO) <!-- must be falsified _->

The attraction of XSL-like language is that (presumably) there will be
engines to process it (and probably author it). And people will be familair
with in. [I built the rules bit of XSL into JUMBO1 so would be quite happy
to see it re-used for this purpose]. I assume that at this stage we may
simply want to give a meta-specification for the rule/pattern ("whatever
syntax is/comes_into force from the W3C"). I am assuming that although the
*output* of XSL is clearly very much up for grabs there is a good chance
that we are still retaining the rule/pattern bits.

P.

Simon St.Laurent
Dynamic HTML: A Primer / XML: A Primer / Cookies