Should FixedPoint Derive From 'object'

[Tim Peters]

[Doug Fort]
> We got a feature request at the SourceForge FixedPoint project
> http://fixedpoint.sourceforge.net
> to make FixedPoint compatible with Python 2.1.
>
> This would be pretty easy: just don't derive from 'object' and
> eliminate the property for precision.

try:
    object
except NameError:
    class object:
        pass

would let you keep deriving from object regardless.  You can assign to
__slots__ regardless too (doing so will simply create a useless class attr
before 2.2).  Etc -- there's always a trick just waiting to make long-term
life hell <wink>.

> My initial assumption on this project was that, since FixedPoint is
> targeted for Python 2.3, we can and should go with new features.

That it works in 2.2.2 and beyond in the 2.2 line is likely worth ensuring,
especially since the Python Business Forum has targeted the 2.2 line as the
Py-In-A-Tie release line, and the PBF is the natural home of potential
FixedPoint users.  I don't know who the 2.1 market for this may be, but one
request does not a landslide make.

> OTOH, in ESR's Bazaar model, the more people who use this code, the
> better it's going to get. This especially applies to those who have
> enough interest to post feature requests.
>
> I don't think we can back out the new class stuff now, and then put
> it back later. That could have the effect of luring people into
> using FixedPoint, and then breaking their code.
>
> I really don't know about this one, so I'm throwing it out to hear
> from the community.
>
> One restriction. I consider Tim Peters to be the BDFL of FixedPoint,
> so he gets a veto.

Na, I'm not even a developer on the project -- if I could make time for
this, I would have.  2.1 support isn't worth anything to me, but I don't
object to it either.  I'm not willing to do any work to get there.  If
somebody else is, fine.  The real question for you is whether that someone
is you <wink>.