Re: [office-formula] Proposal: Drop "huge" group; anycriticallymissing functions for "large"?

["David A. Wheeler" <dwheeler@dwheeler.com>]

Andreas J. Guelzow:
>Would it be acceptable or at least up for discussion to change the
>semantic of functions such as BINOMDIST so that they makes more
>mathematical sense? (The Gnumeric developers do receive requests to
>change the semantic of Gnumeric's implementation of function such as
>BINOMDIST to that end. So far we refused since we viewed it as
important
>to have the same semantic as the other spreadsheets using that name.)

Yes, absolutely! In fact, let's start now.  If you can
identify functions which "don't make mathematical sense"
in the "large" set, I think that's ESPECIALLY critical, because
they might cause trouble when adding the "huge" group.
I already know that the xxx2yyy functions and most of the rounding
functions are completely busted from a cleanliness point-of-view.

The OOo folks have had the same problem; there are several
functions with _ADD names in OOo, because they implemented both
"more reasonable" and "compatible with someone else" versions
(for the latter, usually for Excel or Lotus 1-2-3).

The challenge (I think) is when people's documents start DEPENDING
on the oddities of a particular implementation.  Indeed, it's
hard to fault them; often the "nasty" function is the ONLY
mechanism available to the document creator.  The obvious
solution is to provide the "clean" function and the
"ugly compatibility" function.  We should probably do that
in some cases, and not in others, and it'll take judgement.
I would be delighted to identify a general solution for
"nasty" functions.  Maybe a "legacy" group? Perhaps their
names all begin with "LEGACY." so that their names can be
used for reasonable functions?

What would you recommend?

--- David A. Wheeler