[coyotos-dev] OSDoc issue

[Jonathan S. Shapiro]

On Wed, 2008-03-26 at 22:50 +0100, Marcus Brinkmann wrote:
> I consulted with Annika.  You should probably just ignore the issue.
> The problem is that not only is MathML not very well suited to be read
> by visually impaired people, the same is true for other linearization
> methods as well, at least in general.  The issue is still pretty much
> unresolved.  Next time you watch a mathematical talk, close your eyes
> and try to figure out the equations displayed/read aloud...

Actually, I sometimes find that helpful...

> I suppose that for Coyotos, the actual mathematics required in the
> documentation is minimal.

Unfortunately, it really isn't minimal, and it is likely to become more
important. As an illustrative example, have a look at some of the BitC
papers, where we need to present various proofs about the type system
and the inference algorithm. As Coyotos moves in the direction of being
a formally specified system, the need for mathematical notation will
become pressing.

> Maybe you can just provide textual
> descriptions in addition to any formula.

We can sketch what the math says, but if the math itself could be
presented in English in a satisfactory and sufficiently rigorous way, we
wouldn't be using the mathematical notation in the first place.

> For the general problem, a gentle introduction seems to be
> http://www.snv.jussieu.fr/inova/villette2002/act5b.htm

Thank you. That is a very helpful pointer. From a hasty read of that
document, it seems that (a) there isn't yet any good solution, but (b)
all of the promising approaches can eventually be driven from MathML.


shap