Arguably I've got the dwimminess backwards on this. It seems like
%outer «+=» %inner
ought to dwim that I want the union of keys. In that view the standard
non-dwimmy behavior is to ignore keys that are not in the other hash,
and dwimmery consists of using keys even if they aren't in the other hash.
On the other hand, it seems odd that the "big" end of a hyper would imply
fewer keys. Obviously it doesn't make much difference whether we write
%foo «+=» 1
or
%foo »+=» 1
since the key set has to be considered fixed in any event.
Well, as usual, I can argue it both ways, and will doubtless flipflop
on the issue a couple of times yet... :)
Larry
