On Fri, Sep 23, 2005 at 09:53:12AM +0200, Niklas Nebel wrote: > > But even the description of a single version's behavior is unlikely to > be 100% accurate. If OpenFormula is supposed to specify something, > something like a standard even, such a discrepancy with Excel's behavior > wouldn't invalidate it. You'd have to choose between supporting the > specified formulas, or Excel interoperability (much like our requirement > to handle our own old documents, but added on top of it). > > A mere description, on the other hand, could just be determined as wrong > and thus be corrected right away.
There's an important distinction to be made between errors of omission and commission. While it's true that we are unlikely to make any standard 100% accurate, even for a particular version of MS Excel. It seems reasonable that any errors would be omissions, corner cases that we did not think of. In which case an error is an error in the spec and should be corrected. Implementors would then need to decide how to handle the change, potentially renaming the old implementation to OOO_func and adding a map in import. However, I'd bet that if we didn't think of something for the standard we will have missed it in the implementation too. In which case there is no significant backwards compatibility issue and the change can be done in situ. This standard should be backed by a test suite. A combination of the analytics tests from Google's Summer of Code, Gnumeric's tests, and some of the standardized statistics tests. --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
