Joseph Koshakow <koshy44(at)gmail(dot)com> writes:
> Is this an intentional decision to not support a binary encoding for
> aclitem types? Or is it just a lack of a feature?
I think it's at least somewhat intentional, to have a core type
that has no binary I/O so that that case can be tested. In any
event, a binary representation probably wouldn't be terribly
useful, as it'd contain role OIDs that wouldn't reliably transport
from one system to the next, and also privilege bitmasks that
we sometimes redefine.
regards, tom lane