[std-interval] response from ISL

first.i.last at comcast.net first.i.last at comcast.net
Wed Apr 12 05:27:58 PDT 2006


 -------------- Original message ----------------------
From: George Corliss <George.Corliss at marquette.edu>
> Sylvain,

[snip] 


> I am most interested in the question of whether the
> standard should specify a mathematical model for intervals in the sense
> discussed here, and if so, what that model should be.  The attached response
> argues that there should be a mathematical model, but that it should be
> implementation-defined, since there has not been a consensus.  One of the
> threads over the past couple weeks seems to suggest that leaving such an
> issue to be defined by the implementation guarantees lack of portability.

It may be worthwhile to place explicit limits on the future 
interpretation of the standard as far as restricting the models implementations may use.  For example, nothing in the standard should restrict an implementor from selecting a model wherein the elements contained in the set representing an interval can be from the union of all values representable on the underlying machine and all values representable in IEEE-754/854.  In addition to placing nothing in the standard indicating such a restriction, it may be useful to state that there is no such restriction.

-- Lee Winter


More information about the Std-interval mailing list