Hello, I have recently released the following two Haskell packages, which model and reasonably efficiently implement exact real arithmetic: http://hackage.haskell.org/cgi-bin/hackage-scripts/package/AERN-Real implements arbitrary-precision interval arithmetic for approximating real numbers http://hackage.haskell.org/cgi-bin/hackage-scripts/package/AERN-RnToRm implements arbitrary-precision arithmetic of bounded-degree piecewise-polynomial enclosures for approximating functions of type R^n -> R^m The libraries are fairly stable and complete (AERN-Real more so than AERN-RnToRm) and members of the KEG research group at Aston University have used it for reliably solving differential equations and numerical constraint satisfaction problems. Also, the libraries are designed in multiple levels of abstraction that are clearly separated by type classes. Thanks to this it is very easy to switch between different implementations of the lower-levels, eg using a different floating point type or different representation of polynomials. There are many similarities between AERN-Real and Augustsson's numbers package. The main difference is in how the bit-size of arbitrary precision floating point numbers is determined. In numbers, the bit size is determined statically via types using a type-level encoding of natural numbers. In AERN-Real, the bit-size is part of the floating-point number, which slows the arithmetic down a bit but one can change precision dynamically. Best regards, Michal -- |-| Dr. Michal Konecny, Computer Science, Aston University |-| Room MB212D | Tel +44 121 204 3462 | Fax +44 121 204 3681 |-| http://www.aston.ac.uk/~konecnym |-| OpenPGP key http://www.aston.ac.uk/~konecnym/ki.aston