Infinite-precision arithmetic

[Cubist at aol.com]

sez eijkhout at cs.utk.edu:
>On May 24, 2005, at 5:02 AM, Cubist at aol.com wrote:
>>    I've just started to work on a suite of handlers for 
>> infinite-precision calculations.
>Wouldn't it be easier to interface to an existing bigint library?
   The whole point of what I'm doing with this is two-fold: (a) To have a 
suite of functions with truly *infinite* precision -- not 15 places, not 45 
places, but *as many places as you want or need* -- and (b) to have these functions 
exist as pure, native Transcript.
   So: Existing bigint libraries. Do you know of one that has truly 
*infinite* precision? If not, well, just going from 15 places to 30 ain't gonna pass 
muster. Not for anyone who needs *infinite* precision. Aside from that, I note 
that if anybody *has* interfaced to an existing bigint library, they don't seem 
to have mentioned this feat to the use-rev mailing list...

>Doesn't the Gnu Scientific Library have this stuff?
   No. While the GSL includes an admirably wide range of functions (and I may 
steal some of their algorithms...), every last one of them is of strictly 
*finite* precision, i.e. it's only good for a *finite* number of places of 
accuracy. See above for the whole point of what I'm doing here.