\"Infinite\" precision

[ark at alice.UUCP]

In article <4521 at brl-adm.ARPA>, TVR%CCRMA-F4 at SAIL.Stanford.EDU writes:
>     In Algol 68 there is a provision for infinite precision arithmetic
>     (theoretically anyway--I don't know if anyone ever implemented such
>     a compiler).  Each time one prepends the word "long" to a variable
>     declaration, the compiler doubles the number of bits reserved.  (Or 
>     perhaps adds a constant factor--I forget.)

Nope.  Each length is required to be no shorter than the previous one,
but the implementation is allowed to stop lengthening at some point.
That point is defined by a built-in variable called intlengths.