Re: [office-formula] CODE and CHAR should not be Unicode aware,proposing UNICODE and UNICHAR

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

Me:
> > I believe we still should include CODE and CHAR; they'll simplify round-tripping.

Eike:
> That derived from the Excel behavior of using either Windows or Mac
> character sets, and OOo extended it to system encodings. But for our
> file format spec that behavior doesn't make sense, that's why I proposed
> to spec that CODE and CHAR use cp1252. So CODE and CHAR exist for
> interoperability with Excel, but we loose functionality for Excel
> documents that originated on a Mac when we spec cp1252.

Hmm.  I understand your rationale, but I disagree with it.

As you note, we lose functionality on the Mac, and I don't think Macs are disappearing this year.  I also think this fails even on Windows.  Windows usees the Windows-1252 charset in the U.S. and most West European countries, but I think it is NOT the default on Windows in other countries.

So I think we should NOT require cp1252, though I don't have any problem with _encouraging_ the use of cp1252 in this case if there's no obvious alternative. Instead, I think this should be the "standard" way to gain access to the underlying platform and locale-specific information, whatever that is.

> > The term "Unicode" is probably a trademark of the Unicode Consortium,
> > but I expect that this would be considered a fair use in any country.
> 
> Sigh.. see http://unicode.org/policies/logo_policy.html#4
> Could you please clarify with them whether having a function named
> UNICODE in our spec would touch that legal stuff? In descriptive text
> I guess we'll have to write Unicode^TM wherever we mention it and follow
> those other rules of that page..

I've separately sent an email to a lawyer, to find out how to work this out.  I'm not a lawyer... I want the lawyers to work that out.

If we have to, we can rename the function :-(.  But let's see what the lawyers work out first; we can spec the function now, and change its name later if we need to.  Gnumeric implements it as "UNICODE", but if we can't use that name in the spec due to legal reasons, Gnumeric will probably want to change that name pronto to avoid legal entanglements.

> Given the legal blah I guess the correct wording would be "Returns the
> character code according to the Unicode^TM Standard / ISO 10646".

Probably.  Let's do that for now, and if we can't resolve the legal junk, we'll change the name.

--- David A. Wheeler