timothy murphy asked,

    > And isn't it quite sensible to distinguish between text and maths?

and torsten bronger responded,

    XML doesn't do it and I find this very convenient.  In (La)TeX, for
    many characters you need different commands for text and math mode.
    I'd love to have a typesetting system to which I could pass a say
    'small Greek letter alpha', and it would just work in every context.
    No font families, no encodings, no active/special characters, no
    babel settings, and no modes to worry about.  Wonderful ...

not quite.  i'm with tim here.

for math publication, it's traditional to have variables in
italic.  it's also traditional to have theorems in italic.
unless it's marked as math, how can you tell that "a" in a
theorem is a variable or the english indefinite article?
knuth tried to avoid this by
 - using a slanted font instead of italic for theorems
 - making a math italic that is ever so slightly wider than
   text italic
but it still definitely requires marking a math "a" as math.
                                                        -- bb