Filter() allows None as a stand-in for either the still-nonexistent
identity() or the now existent bool().  Removing the now redundant
option would slightly simplify the language.  It is certainly a glitch
for newcomers.  Would this option be included if filter were being added
now?

Fixer: I should think filter(None,it) => filter(bool,it) should be
sufficient as I expect filter(f_f_exp, it), where f_f_exp is a filter
function expression that sometimes gives None, should be fairly rare.

Efficiency: filter() could (if it does not now) avoid calling bool twice
internally by checking for bool instead of None).  Even if removing None
is rejected, filter should also check for bool (if it does not now) to
not penalize those who write the obvious filter(bool,it).

If this item is rejected, turn it into a doc item for PEP3099.

FYI, filter() already special-cases "bool". I think the discussion
whether to remove the None argument has been had, but I could be
mistaken. Guido, do you remember more?

It was deferred to me and I decided to keep the None argument as this 
is the standard way of spelling a missing argument.  Also, the None 
argument provides a pretty good clue that a special fast path is being 
taken.

Okay, closing as won't fix.

Reopening for the alternate suggested fix of adding a note to 3099,
which I checked before submitting this, so this need not come up again.

I don't think this is important enough to get into 3099.