"Giampaolo, people using only 3.7 should probably use asyncio. Fixing
asyncore is more important to those that can use only 2.7 (e.g.Centos/RHEL) or have to support both python 3 and 2."

IMHO starting to use closing in asyncore *is* a backward incompatible change. We don't know how applications use this existing attribute. Maybe it's not set to a boolean. Maybe it uses a different policy.

I would be ok to start using closing in Python 3.7, but I'm not confortable with backporting such change.

I would be if would be add a new private attribute, as the proposed "_closed" name.

Maybe nobody uses closing. Maybe people using closing have a similar usage. The thing is that we don't know, and according to what you wrote Nir, closing *is* used.

What I dislike in asyncore is that subclassing is not prohibed, it seems wanted by design. Ok, but what if a subclass overrides completely a method and doesn't set "closing/_closed" anymore? That's another reason why I dislike the idea of making any change in asyncore *especially* in Python 2.7 which is now considered as "very stable".