Backwards Compatibility of Python versions

Kragen Sitaker kragen at pobox.com
Sun Feb 3 03:25:29 CET 2002


Alex Martelli <aleax at aleax.it> writes:
> Kragen Sitaker wrote:
>         ...
> > and "x in somedict", which makes it impossible to have a class
> > instance faithfully emulate both dictionaries and lists.  I'm not
> 
> No "make it impossible", because it ALREADY was impossible
> before.  How "faithfully" am I emulating a dictionary if rebinding
> x[-1] changes, say, x[5]?!  Yet that is part of what I MUST do
> to "faithfully emulate" a list which has 6 items at this time.

Well, I had a filesystem mapping; indexing by strings was associative
and accessed files by name, while indexing by integers was sequential
and iterated over the files in a directory.

Faithfully emulating a dictionary doesn't mean you can't do things
dictionaries don't do --- it just means you can do everything
dictionaries can.

> We need to reach a new stable plateau reasonably fast, at least as
> badly as we need to keep backwards compatibility.

What do you mean by "stable"?  What's "unstable" about having to put
'from __future__ import division' at the top of your module to get
reasonable division behavior for the next two, three, five, or ten
years?

> Right now we're clearly smack in the middle of it -- the worst of
> times to start slowing down.  I suspect Guido has an excellent handle
> on the balance of THIS tradeoff just like he has on the huge number
> of tradeoffs that cumulatively make a language.

Well, he's certainly better at it than I am.




More information about the Python-list mailing list