Need help finding a past discussion

[Stefan Monnier <monnier@iro.umontreal.ca>]

A year or two ago, I remember a discussion here about something like
a hook where we wanted to make an incompatible change which required
somehow distinguishing those functions that expected the old behavior
from those that expected the new one.

Maybe the change was about making the hook function return another
function, so they could be called in two steps and passed arguments at
two separate times.

I vaguely also remember that we mentioned relying on symbol properties
(when the function was represented as a symbol) to distinguish the kind
of function at hand.

Does that ring a bell for someone?


        Stefan