> Correct me if I'm wrong but its pretty common in all areas of mathematics to

This is true, but you have to watch for two things:

> have more than one definition, each definition can be proved as a theorem

> from the others and so one can choose his definition according to his needs

> and prove the rest as theorems.

>

> Do you take Axiom of Choice as an axiom or theorem provable from Zorn's

> Lemma?

> Are Sin and Cos defined as properties of triangles or infinite sums?

1. When you teach a part of mathematics, or write a book about it, you must

be consistant in the order of your definitions. By that I mean that you

can't define A assuming the reader already knows the definitions of B,

and then define B assuming the reading already knows the definitions A.

Not only will this confuse a beginner that doesn't have previous knowledge

of the subject, but most likely that using such a flawed exposition you

can "prove" pretty much anything.

This is why, if you teach the function e^x in a beginning of a calculus

course, you cannot define it based on taylor series or integrals.

Later, when you teach those subjects (taylor series and integrals) you're

free (and indeed expected) to show that a new definition, based on these

new concepts, is equivalent to the original definition. The benefit of

seeing equivalent defintions is that sometimes the alternative defintion

is easier to use in practice in many cases. E.g., except in the beginning

of the calculus course, nobody actually _uses_ the "lim (1+1/n)^n"

definition of e.

2. When something has an original definition (e.g., sin as a ratio of

triangle's sides), and you later come up with a new definition that you

show is equivalent (e.g., the taylor series), you sometimes find that the

new definition can suddenly be extended to a bigger domain, such as to

define sin() for complex numbers. If you choose to extend the domain, you

must be aware that this is in fact a new definition: some of the the old

properties may still be true (e.g., sin''=-sin is true also for complex

arguments), but some are not (e.g., sin of complex numbers has nothing to

do with triangles).

Whether this new extension becomes useful depends on whether it keeps the

_important_ properties of the original definition. For example, e^x is

important because e'=e, not because, say, e^x>0 forall x (the latter is

no longer true for complex: e^(pi*i)=-1, but nobody cares about that).

--

Nadav Har'El | Sunday, Feb 11 2001, 18 Shevat 5761

nyh@... |-----------------------------------------

Phone: +972-53-245868, ICQ 13349191 |Disclaimer: The opinions expressed above

http://nadav.harel.org.il |are not my own.