Machine Independent Floating Point

[nicholas]

Has anyone any info, references and the like on work concerning machine
independent floating point implementations ? 

I'm especially interested in implementations that enforce a certain 
granularity wihout resorting to fixed point, so that floating point values
can easily be enumerated by the integers in a machine independent way.

Ada numerics seem to go part of the way with the model number concept so
any details on that area would be very welcome as would any details of
68K and 32K micro implementations.

Thanks Nick Walton