Timothy Murphy writes: : On Sat, Dec 12, 1998 at 06:05:25PM +0100, Chris Rowley wrote: : : There is a fundamental question about MathML/XML/OpenMath vs TeX/LaTeX : which does not seem to me to have been answered here. : : As I understand it, *ML _parses_ (or tries to parse) maths, No. : while Knuth in his wisdom decided this was impractical. : For example, if I write $AB = CD$ : this might refer to variables AB,CD (perhaps line segments) : or it might refer to products of 4 variables A,B,C,D; I would suggest that in a latex-like document preamble one declare, with something like "\mathsym" the symbols A, B, C, D either as having a type such as "vertex" or else having a type such as "element of an additive group". In either case to obtain recognition of the typed symbols, one should in the body use $A B = C D$ if either "AB" or "CD" is a symbol as declared by another \mathsym. Absent a mathsym declaration the string "AB" in a math zone would be equivalent, by tradition, to "A B". Although I argue that "A B" by default should stand for function composition of A and B when it makes sense, it does not make sense, absent symbol declarations, to assume that A and B are functions that can be composed. The next choice, absent other guidance, is to assume an arithmetic context with juxtaposition standing for multiplication. (But note that I need none of this, other than symbol separation if I don't process math zones specially, to get the representations that I put up in my last posting to latex-l under the subject "notational examples".) : Would I perhaps have to put in an \invisibletimes between A and B? I hope not. : The question is, then: : is it possible to parse mathematics, : and if so, is it wise to try? Don't readers do it? You think that it's based on "visual intelligence"? : [It may be impossible, : because it may be that mathematicians : would refuse to be bound by any particular formal system put forward.] Ay, there's the rub -- certainly for many. Note in particular that mathematical authors have incentive to provide MathML only if they perceive that the *most helpless* of their readers who have browsers also have MathML rendering. It would be helpful if some of the W3C MathML working group (whose emissary usually listens here) would ponder a format for a flexible author-specified type system and provide us with testbed code for generating MathML based on it. -- Bill