| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Baker et al, P1788 On 10 Jul 2013, at 21:13, Ralph Baker Kearfott wrote: > Juergen, Ned, > > Do we need a simple motion stating "P-1788 will not specify > items dealing rational arithmetic" (or other clear phrasing)? > That would resolve the issue, but would take several weeks to > process... > > On 07/10/2013 04:47 AM, J. Wolff von Gudenberg wrote: >> P1788 >> I strongly support Ned's argument against rational endpoints in P1788 >> Jürgen This focuses our minds on a crucial issue. We need a standard that is easy (well, fairly easy for an experienced person) to implement in portable code, as well as easy to explain. We need an industrial-strength standard that has facilities required by demanding applications. These two aims unavoidably conflict. For the current motion, I attach final text to be voted on. I've done a bit of rewriting, but the main change is that I've hopefully lessened the above conflict by making "rational endpoints", i.e. rational number literals, optional. Besides infsup form [5.6,7.8], Neumaier's uncertain form 5.6?1 stays because I think it is useful, and easy to read once you get used to it. For the standard as a whole, I think we need to define a subset forming a "basic standard". - According to the OASIS document on conformance this is a respectable thing to do. - From the viewpoint of specifying this subset precisely it can be done within Christian's Conformance Clause. - However from the viewpoint of readability there needs to be a reduced standard document that explains the basic standard, and only that. I'm in discussion about appointing a "simplifier in chief" to define and document the basic standard, and hope an announcement will follow shortly. Best wishes John Pryce
Attachment:
20130711IntervalLiterals.pdf
Description: Adobe PDF document