I've started an article on this subject at Not a Number. Please contribute. We need a solution to this problem.
The discussion has already started
Have a look at what we analyzed at: http://dev.eiffel.com/Real_numbers.
I couldn't see equality
Thereś some useful stuff in there, but I didn't see anything on the equality issue, which is the killer for assertion monitoring.
I meant proposed solutions
Yes, Peter has mentioned the issue too.
But my remark was about the page presenting the proposed new interface for REAL_64.
It has some nice features, but lacks an analysis of how to deal with the NaN equality problem.
Until this is decided upon, writing the contracts is premature, as you can not see the impact.