[Om3] OMS and OMV contents?

[James Davenport]

On Wed, 11 Jul 2007, Michael Kohlhase wrote:
> one thing we have to think about wrt. OM3/MathML3 alignment is whether
> we want to allow presentation MathML in the body of OMV and OMS
> elements. I know we have discussed this before (for OM2) and rejected
> it, but with the alignment goal this question raises it's ugly head
> again.
Ugly is indeed correct.
> Allowing full presentation MathML content would quadruple the size of
> the RNC schema (well we would just add a link of course) or add an
> external dependency.
>
> Going for a restricted set of presentation MathML would be nice (who
> needs mtable in a variable name), but would require furer coordination
> with the W3C Math group.
AND it wuld mean that we could not just add a link - we would really have
to specify the subset.
> The OM2 standard allows to add presentation MathML as a special
> attribution, but that is much less direct than MathML does make the
> alignment non-isomorphic.
There are conceptually two questions: symbols and variables, with
different issues.
OMS: Here the addition of presentation can only change the presentation.
     In MathML, this markup can only occur in csymbol as far as I can see
     (section 5.2.3), and I don't see what is too hard about mapping
     <csymbol cdbase="http://www.example.com"
     encoding="application/x-MathML-CD"
     cd="VectorCalculus" name="Christioffel">random presentation</csymbol>
     To
     <OMATTR
      <OMATP>
        ... random presentation
      </OMATP>
      <OMS cd="VectorCalculus" name="Christioffel"/>
     </OMATTR>
OMV: Here the question is more subtle - does the presentation create a NEW
     OMV, or not.
     If it does NOT, then OMATTR strikes me as still being the right way
     to go.
     The hard problem is what happens if 'gothicA' is a DIFFERENT OMV from
     'A'. Here there probbaly is a case for changing OMV to allow a
     'renderas' attribute, which would name a NEW OMV.
James