that would be nice, but it's the normal way people get into a big
mess. the problem is simply that it works great for trivial cases,
but grows horribly messy when you try to do anything serious.
If you're doing high school algebra, this is fine. Try doing tensors,
or even plain old calculus. You will, step by step, be forced to
add hundreds of special symbols, dozens of special formatting types
(like a matrix type for linear algebra), and on and on and....
Guess what? there's a *reason* people don't do it that way. Don Knuth
would certainly have used that method in TeX if it would work.
The cases that such as scheme *can* handle don't actually look much
different from that in TEX or SGML -- but they look a little different,
in order to keep the door open for doing all the more sophisticated
stuff when you need to, rather than locking the door entirely or
forcing 2 incompatible 'simple' vs. 'complex' things to learn.
It would be much like proposing that we standardize on gear and chain
drive mechanisms for all vehicles, because it works great for bicycles,
and anyway, for lots of people a bicycle is all the need anyway.
So then we end up building chain-drive Jaguars?
Steve