Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Motion 6



Vincent LefÛvre wrote:
> With midrad typical intervals, the error is small, so that the
> interval keeps the sign information, unless the real value is 0.
> This is the only case where the triplex representation can be
> better.

"Sign" is just an instance of "on which side of a singular value?".
It is true that this is limited by the set of representable values,
but that set includes small integers and (in binary) simple dyadic
rationals.

> Anyway standardizing either midrad or triplex (or both) in
> addition to infsup would be better than requiring infsup alone.

One thing that is clear is that different datum types would be needed;
the distinction can't be hidden behind "implementer's choice", unlike
different concrete representations of a given type.  It is almost a
level-1 distinction.  Actually, my "two domains" proposal should
probably be phrased in a level-1 sense, namely whether an interval
represents a single value with uncertainty, or a range of independent
values subject to boundary constraints.


Michel.
---Sent: 2009-09-21 11:55:53 UTC