[ogsa-rss-wg] Candidate Ordering Language

[Donal K. Fellows]

Andreas Savva wrote:
> After a quick glance this looks a bit like MathML. One thought is that
> this 'simple' language could be a profiled subset of MathML, picking out
> those terms which must be supported by all implementations. More
> sophisticated implementations could support other features (and there
> would not be a need to worry about definining extensions).

Alas, I don't think MathML provides any way to incorporate the <value>
and <exists> elements that I proposed. I believe this stems from the
fact that MathML is a language for describing expressions and not a
language for describing a (tiny) program. Since support for those
elements is utterly neccessary (since it's the only way to couple the
expression language to the candidate being ranked!) there's not much
that can be done otherwise. (The main way of coupling values in MML is
via id and ref attributes, but there's no practical way to constrain the
candidates to be cooperative in respect of such things.)

Well, I suppose MML could be used by describing everything using a
scheme based on using XSLT to map some other language into MML and then
running the implied processing to derive the value, but that doesn't
admit even an approximately efficient implementation due to the nature
of transformations that XSLT can (rightly) perform.

Donal.