Numerics

So I started playing with 2.1+ this weekend when I came across an
interesting little change, REAL is 64-bits. Coming from a scientific
background, I like the idea of 64-bits everywhere, in fact, we've talked
with B. Meyer a number of times over the last 6-8 years over full 64-bit
support in Eiffel. By full, I mean the basic REAL and INTEGER types are
64-bit, and that is supported the entire class library (i.e, array
indicies are 64-bit, ...)

Given that both ETL2 and ETL3 state that REAL is 32-bits what is the
motivaton for the change? This change has messed up a lot of my code
that has an external interface to OpenGL which requires single precision
floats in many of the calls (the symptom that led to the discovery was
an object that was previously lit with white light (3.0, 3.0, 3.0) had
turned purple (3.0, 0.0, 3.0).

Is this a test to try a new implementation and will change (or not) once
we have an ECMA standard?

I'm not bashing, I'm just hoping for an explanation so I can decide how
much effort to require in porting to a newer version of SE.

Mark

P.S. Please fix Bug 471 so EWG can support SE 2.1