Re: Containment is necessary but not sufficient...

["Nate Hayes" <nh@xxxxxxxxxxxxxxxxx>]

N.M. Maclaren wrote:
> On Aug 26 2011, J. Wolff von Gudenberg wrote:
> > I think that we do not have the time to work on more than 1 concrete types
> > and suggest that we proceed with the discussion of IEEE 754 (binary)
> > interval in inf-sup representation, and drop all the others for
> > specialists.
>
> Hmm.  There are many people who believe that an interval specification
> should be generic, and concrete bindings should be an addition.  That
> is the way to make it resistant to future developments, and will increase
> the chances of it being adopted by language standards.

In my view, P1788 succeeded doing this (so far) by passing motion 19 and
defining the explicit and implict abstract interval data types. These ADTs
specify the required mappings between level 1 and level 2 without getting
into concrete level 3 and level 4 details.

On the other hand, my impression is people expect there will be some
explicit inf-sup data type defined at level 3 and level 4 to be a pair of
IEEE 754 values, as this will likely be used in a large number of
implementations and applications.

So I *think* all Jurgen and Marco were suggesting is that -- in addition to
specifying the explicit and implicit ADTs -- P1788 might also wish to
consider specifying an explicit inf-sup data type at levels 3 and 4.
However, to the extent I understand correctly, this would build on top of
the generic ADT bindings and not replace them.

For example, implementers could still provide conforming implicit types such
as mid-rad and variable precision, even though IEEE 1788 might not give any
concrete specification for these types at level 3 and level 4.


> If a standard does not address a programming community's requirements,
> which includes covering the special functions that they need, it won't
> get used.

Definately.

Nate