Python 2.0

Paul Boddie paulb at infercor.no
Fri Jun 4 11:28:36 CEST 1999


Graham Matthews wrote:
> 
> Can I suggest you read the Java spec. As someone else has pointed out
> Java memory management is done via an interface. Hence you can implement
> any memory management scheme you like in Java.

I'm not too bothered about what *can* be done in Java. There are surely decent
enough real-time garbage collection schemes in some implementations of Java, for
example, and that would appease at least a few people, but what worries me more
is this sequence of events:

1. The suggestion that Java should bear the principal implementation of Python.
   One of the reasons given is Java's "superior" garbage collection mechanisms.
2. The expression of doubts that such garbage collection mechanisms are "good
   enough". These doubts are founded on "real world" experiences.
3. The assertion that good enough schemes exist which would not necessarily go
   against people's Python expectations, and that Java will get better over
   time (as if we haven't heard that enough in the last 4-5 years).
4. The repeated suggestion that there are "real world" problems with Java.
5. The repeated assertion that Java doesn't have to be done the way it
   frequently is.

Now we end up stuck in steps 4 and 5.

I really don't care what "hosts" Python in the future as long as most people are
sufficiently happy with the implementation, are not caught out by weird
goings-on which go against their expectations, and can at least run it on their
preferred platform. The C implementation might have some shortcomings, although
I am not that affected by them in practice, and it is right to draw attention to
them - I have seen users of certain computing platforms bury their heads in the
sand over some major problems in the systems that they use, with the result of
such overly-defensive behaviour being the severe decline in the usage of and
development for such platforms. However, given the less-than-rapturous reception
of previous "true GC in Python" work, I don't think that I am out of line in
saying that it isn't necessarily as high a priority as some contributors to this
group/list insist.

Paul




More information about the Python-list mailing list