Re: [office-formula] Constraints and infix ^

[robert_weir@us.ibm.com]

"David A. Wheeler" <dwheeler@dwheeler.com> wrote on 01/22/2009 10:55:46 
PM:


> robert_weir@us.ibm.com wrote:
> > Implementation-defined behaviors tend to stem from a few causes:
> ...
> 
> > 3) A standard created from divergent practice.
> ...
> > I think spreadsheet formulas are mainly divergent because of #3,
> 
> Agree.  But in the "0^0" case, the problem is that mathematics
> itself is divergent.  According to:
>   http://en.wikipedia.org/wiki/Exponentiation#Zero_to_the_zero_power
> "The debate has been going on at least since the early 1800s."
> 
> Even among spreadsheets, Excel returns an Error and OpenOffice.org 
> returns 1.  Personally, I think 1 is a "more useful" answer in the 
> spreadsheet context, but I completely agree that Error is a completely 
> justifiable result.
> 
> I think this is a divergence that we can't completely close at this 
> time, but we can at least limit the permissible behavior to a small set 
> of permitted results.   I think that is still termed 
> "implementation-defined", but if there's a better term for that, that's 
> fine.
> 

I wonder if in cases like this we can give a clearer direction, while 
still allowing some variation.  For example, saying that the function 
SHALL return one of a small set of results, but it SHOULD return one in 
particular.  Or say it SHALL return one of a small set of results, but 
that returning result X or Y is deprecated.

A lot of time has passed since your work on OpenFormula began.  ODF is 
much more mainstream now. We even have new implementations from Google, 
IBM and Microsoft that were not around when this work started.  What you 
have is a great compilation and analysis of spreadsheet practice at the 
time. But it might be worth seeing whether vendors attitudes have changed 
in their willingness to conform to a stricter standard.  In particular, is 
there a will to define 0^0 in a single way now? 

One way to do this is to have a list of all implementation-defined 
features in OpenFormula (or indeed all of ODF).  We could then have a 
quick discussion on each one and see if implementors are willing to 
compromise on some by changing their implementations.  I'm willing to put 
together such a list.

Note that some implementation-defined items, like numeric precision, 
spreadsheet sizes, etc., are intentionally left implementation-defined. We 
don't want to eliminate those.  But the others, we might see if there is a 
general will to define these in a narrower way.

-Rob