[viff-devel] Floating point operation [Re. to Martin's message]

[Mikkel Krøigård]

> > I have read the "Cryptocomputing with rationals" paper, and although
> > it is a nice idea it does not really solve the whole thing.
>
> Could you outline the problems so that I wont have to read it? :-)
I think you have seen it, but you didn't recognize the title. It is the one
where you think of a/b as ab^{-1} mod N and use Paillier encryption. If you
promise not to exceed certain bounds, you can add and multiply these things and
by using the "Gauss algorithm" (sort of like the lattice version of Euclid's)
you can return to a representation where gcd(a,b)=1.