[Python-Dev] RELEASED Python 3.0 final

Guido van Rossum guido at python.org
Sat Dec 6 21:51:55 CET 2008

On Sat, Dec 6, 2008 at 12:19 PM,  <glyph at divmod.com> wrote:
> I also don't think 3.0 is perfect, and five years on, there will be a
> temptation to make more "just this once" incompatible changes.  Of course,
> you've promised these changes won't be made, and *this* set of design
> mistakes will be with us forever.  It would be nice if there were a way for
> evolution to continue without another reboot of the world.

It would be nice indeed. But we (and any other language that's alive)
will need to walk a careful line between evolving too slow and too
fast. Hopefully we'll be able to evolve mostly through deprecation and
eventual removal of misfeatures rather than through a series of
hiccups like 3.0. But it will still be too slow for some and too fast
for others.

Since one of your favorite themes is that your team is too small, I
would like to reuse that idea. If we had as many Python core
developers as Sun and IBM have working on Java, we could most likely
have introduced all Python 3.0 features gradually, with compiler flags
and __future__ imports to support different versions. But despite
being a bit bigger than Twisted, we're still severely constrained by
resources. My estimation when we started was that it would be easier
for the core team to maintain two separate versions over a long time,
than to try and produce a single binary capable of running both
versions of the language. (Maybe Jython and/or IronPython provide a
better platform for doing that though.)

Hopefully by the time Python 4000 rolls along, technology will be
available to make the transition more smoothly. But we'll still have
to break some eggs...

--Guido van Rossum (home page: http://www.python.org/~guido/)

More information about the Python-Dev mailing list