Andrew Wilcox <xxxxxx@andrewwilcox.name> writes:

> To take an example of [Egner et al. 2004], (< x y) in this proposal
> returns an *inexact* boolean, if either X or Y is inexact.

A boolean is almost always used to choose control flow. Since you
can't make control flow inexact, inexactness is not really contagious.
It can't be. Inexact booleans don't add any real value.

[More comments later.]

--
   __("<         Marcin Kowalczyk
   \__/       xxxxxx@knm.org.pl
    ^^     http://qrnik.knm.org.pl/~qrczak/