[Om3] Definite Integrals (was Re: [Fwd: MathML CDs])

[James Davenport]

On Fri, 6 Jul 2007, Michael Kohlhase wrote:
> Professor James Davenport wrote:
> > On Thu, 5 Jul 2007, Michael Kohlhase wrote:
> Dear James, a piece of background for our discussions here. In content MathML3
> (I mean the published draft at http://www.w3.org/TR/2007/WD-MathML3-20070427/)
> we have distinguished "canonical MathML" and "legacy MathML" (names to be
> reconsidered), the former is CD-based in syntax, and is the basis for OpenMath
> alignment. The latter is a convenience extension to keep backwards
> compatibility and write down things in a easy-to-understand way. Its meaning
> is given in form of canonical MathML. <upperbound> and <lowerbound> are part
> of the latter, and we do not have to align them with MathML.
[ I assume you meant OpenMath, not MathML, at the end.
OK. So legacy-C to canonical-C is nothing to do with OM, and W need only
worry about canonical-C to OM. That makes life a lot easier (and calcmml1
redundant). I'll have another stab, concentrating on the defint-binder.

> Introducing constructor symbols for these seems like a bad idea, since they
> only have a meaning inside integral, sum, and product... They are purely
> representational, and I dislike that.
Agreed - I only added them because I thought we had to. Happily junked.

> > Oh - I see, this is a portmanteau of defint and interval, and I am not
> > sure i like that.
> >
> yes it is, and I agree that the only redeeming feature with an applicative
> integral operator is that it allows us to get rid of interval. I am not really
> fighting for this, but in the presentation process (the ntn files), we do not
> really have a way of distinguishing $\int_{[0,1]}\sin(x)dx$ and
> $\int_0^1\sin(x)dx$. If you say that we should not cater to presentation in
> cMathML/OM, then I would have to agree. Distinguishing defint and defintbounds
> is not one of my top priorities.
Right. On the lines of my MKM paper, I would argue that this is purely a
presentational issue, and is some-one wants to invent a
presentation-oriened decoration for it, that's fine by me.
I'll ignore defintbounds until further notice.

> I do want to stress that we need binding forms of the integral, and that the
> set (and possibly bounds) arguments should be part of a complex binding
> operator.
Not sure what you mean by 'set', unless you mean writing [0,1] as a set.
But I agree with your general point.

I'll go re-read MathML-C3.0 again with this enlightemnment.
James