Precision and CD's

[Bruce R Miller]

J H Davenport wrote:
> 
> I think you're being harsh on what David set out to achieve.

There was no intention of insulting David, his code, nor your
elegant bigfloat CD proposal --- and I hope none was taken.
I simply pointed out that there is an important, _practical_
limitation to this scheme.

> What he has demonstrated is that XSL {\em alone} can produce
> relatively readable representations even in bases other than decimal.

I predict the `XSL {\em alone}' part will turn out to be important to OM 
 --- but we certainly differ on `readable' part :>

  Is it, or should it be, an OM goal that XSL is sufficient to 
  do moderately high quality rendition into MathML?

I take this as a "No":   :>

> Clearly an application that was deisgned to render floats reasonably would
> have a small applet (or even a real program!) to do better. If it did not
> have bignums, it might truncate via IEEE, so one might get
> 1.000000000000000000000000000000123x16^3 (c. 4096.0).

Or alternatively, the underlying OM representation would be a printable
string. (not perfect either, but...)

--
bruce.miller at nist.gov
http://math.nist.gov/~BMiller/