Calling registerProducer() multiple times without calling unregisterProducer() is confusing

[exarkun]

registerProducer() will happily replace an existing one with
the new producer, leading to behavior where an old producer
is never resumed again.  This seems like it would only ever
be a bug.  Would an assert or a log warning make sense in
registerProducer() to ensure no one ever does this without
realizing it?

[itamarst]

alternatively, the old producer should get stopProducing
called. But we do need a specific behaviour to be documented
and implemented.

[spiv]

I'd lean towards "explict is better than implicit", i.e.
don't magically call stopProducing.  I think a
warning.warn("registerProducer called, but a producer is
already registered") would be my preference, perhaps to
become an exception in a later release.

[itamarst]

seeing that doing this is a bug (since the old producer will
continue to produce if it's streaming) I suggest it be an
exception from the start. Anyone disagree?

[exarkun]

Exception sounds good to me.

[spiv]

Ok, that makes sense.  I'll look at making it raise an
exception, then (seeing as someone kindly, uh, "volunteered"
me :)

[glyph]

On Tuesday, August 26, 2003, at 12:50 AM, Andrew Bennetts wrote:

> perhaps to become an exception in a later release.

I'd say make it an exception now.  AFAICT this behavior is undefined as 
it is.

[moshez]

/cvs/Twisted/twisted/internet/abstract.py,v  <--  abstract.py
new revision: 1.40; previous revision: 1.39
/cvs/Twisted/twisted/test/test_internet.py,v  <--  test_internet.py
new revision: 1.35; previous revision: 1.34

Fixed, resolving.

[moshez]

woops really resolving