Please, I think I made it very clear that whatever the agreed direction, the documentation needs to be explicit. I also looked in the AOO embedded help for POWER and it is not helpful. (By the way, it would be great if it remained possible to access the on-line help even if the embedded help is installed.)
That is the only point of the note being replied to below. Knuth and I will continue to disagree about what is an appropriate result for 0^0 (maybe even 0/0). Either way, I rest my case. 0 is already an exception in the laws of arithmetic. One must always make exceptions to avoid reasoning through a division by 0 (a common trick for proving that 0 = 1). And as far as the February 1900 calendar bug goes, please recall that it was to remain downward compatible with Lotus 1-2-3. At the international standard level that was not tolerated, as you know. Think of this as delayed karma for Apache OpenOffice. In this case, allowing any of 0, 1, or error to be implementation-defined is not helpful and the compelling case for me is alignment with Excel on "error." I see no point in reiterating on this. I am not going to change my recommendation. All of the considerations have been presented. Now it remains to determine how to go forward. I have no idea what the consensus will be. I await what others have to say. I have no need to say any more. - Dennis -----Original Message----- From: Rob Weir [mailto:robw...@apache.org] Sent: Saturday, February 09, 2013 16:57 To: dev@openoffice.apache.org; orc...@apache.org Subject: Re: Calc behavior: result of 0 ^ 0 On Sat, Feb 9, 2013 at 7:39 PM, Dennis E. Hamilton <orc...@apache.org> wrote: > It is not clear that OpenOffice-lineage software has returned the same value > for POWER(0,0) over the years. It seems that a third-party library has been > relied upon for the implementation and there was apparently not much > attention to edge cases. If that library changes or is different on > different platforms, there is the prospect of unexpected differences. It is > good that this is being nailed down. > > In any case, in order to produce *any* reliable result for POWER(0,0), it is > necessary to declare what that is as an implementation-defined (not > -dependent) commitment. So Sayeth ODF 1.2 OpenFormula. There is big difference between documenting the behavior ("nailing it down") and changing the behavior. [ ... ]
