Peter Shenkin wrote:
> "May be" is the right term; L'Hopital's rule can be
> invoked in some cases to show that the value also "may
> not be" path-dependent.
>
And it may fail entirely. See
http://mathworld.wolfram.com/LHospitalsRule.html for a quite little
discussion.
> ...
> But if x is a real number, we can apply L'Hopital's rule
>
I think you also need to have assertions about the function, not just
the datatype (continuous, if memory serves, as an example), the cited
reference has some examples where it's wrong even then (too many sign
changes....)
> ....
> To me, this is pretty convincing evidence that the only
> reasonable definition of 0**0 for real x (at least) is 1.
>
Many wise people have suggested it over the years (e.g. Kahan). It
certainly would have it's advantages. The downside is for those
times/places where it isn't appropriate testing becomes pretty tedious
... but it's a question of which set of people to annoy ;>
--
Keith H. Bierman [log in to unmask]
Sun Microsystems PAE | [log in to unmask]
500 Eldorado Blvd UBRM05-169 | 650-352-4432 voice+fax
Broomfield, CO 80021 | sun internal 68207
http://blogs.sun.com/khb
<speaking for myself, not Sun*> Copyright 2006
--
|