On 05/22/2013 06:56 AM, Richard Fateman wrote:
> On 5/22/2013 3:40 AM, Barton Willis wrote:
>> There is some difference between when a tellsimp rule is applied.
>> Maybe the "and may not work"
>> warning is relevant?
> It certainly is relevant. I don't understand why people insist on using
> such rules even after such warnings.
I think in common usage "may not work" means to most people "has some
small possibility of breaking". If instead this technique is really
rarely going to work it should say "and is not recommended as it will
most likely break" rather than "and may not work"
perhaps "correct operation of rules on + or * would require maxima to
determine all possible ways in which a result can be partitioned into
two terms or factors, which requires considering an infinity of
alternative possibilities. Maxima instead tries only one specific
possibility, and therefore normally will not work with +,* rules"
I remember being told that psychology research shows that warnings which
give some indication of *why* something is a bad idea are more effective
than warnings which simply tell people to do something. Compare:
"Do not smoke near this building"
to
"Do not smoke near this building as flammable vapors may cause an
enormous explosion"