Future compatibility of the Python C API

[Mark Hammond]

No one can speak for Guido, so here are my opinions and guesses :-)

"Keith Davidson" <kbd at iti-oh.com> wrote in message
news:lwEJ4.7980$Ib7.101092 at typhoon2.kc.rr.com...
> 1) Has there been any discussion about the upwards compatibility of the
C
> API in future versions of Python?

Yes.

>  Historically our libraries have had a
> 7-10+ year lifespan where we provide 95+% backwards compatibility (i.e.
we
> don't want our C API users to have to make more than a day or so worth
of
> changes between releases).

And this is the goal with the 1.x family.

> My big concern is Python 3000- what are the odds
> that its C API will be backwards compatibility with Python 1.x?

Not very good.  If it is a real problem, simply dont upgrade.  I would
however expect the changes to cause compiler errors that are simple to
fix - ie, very similar APIs, and unlikely to silently change behaviour.

> 2) We are considering using the Python C API heavily in a set of
libraries
> that will be used by both C applications and as Python extension
modules.
> Has anyone else done this?  How did it work out?

Sure.  Great.  Really, there is no difference between a set of libraries
and an application that embeds Python - just the entry points change!

Mark.