Re: [Maxima-discuss] SIGN1 strangeness, was: Problem with bfloat

[Henry B. <hb...@pi...>]

Shouldn't 'inf be converted to IEEE inf only when "numer" is active?

Shouldn't IEEE inf be converted back to 'inf by rat, just like it converts floats to rational numbers?

At 08:46 PM 9/4/2014, Richard Fateman wrote:
>On 9/3/2014 12:28 PM, Raymond Toy wrote:
>>>>>>> "Richard" == Richard Fateman <fa...@be...> writes:
>>      Richard> A solution to this problem of inf-inf (and a number of others)
>>      Richard> is to tag each inf uniquely.  So then inf-inf is zero  if those infs are
>>      Richard> identically tagged.
>>
>> I think this "inf" is the IEEE754 infinity value, not maxima's inf.
>>
>> --
>> Ray
>
>Seems to me that if we had a theory of how to deal with infinities, we should convert the IEEE
>infinities to "our" infinities.  I'm thinking of a hack that will do most of what I proposed (in conjunction
>with intervals), that can be read in as a patch.  If there are any people eager to look at this,
>send me a note.