Re: extended operator classes vs. type interfaces

Robert Haas <robertmhaas@gmail.com> wrote:
> I dowanna rework the type system.  I'm not even 100% sure I want
> to implement what I actually proposed.  I do want to find out if
> people think the framework makes sense and whether it's the right
> way forward for those projects that need these features.
What you proposed sounds like it would be cleaner and less work than
further perverting the index system as a source of information about
types or hard-coding knowledge anywhere else.
> What you're proposing here sounds suspiciously like something that
> should be handled by creating domains
Not really.  Unless I've missed something domains are a single-level
layer over a data type.  I find them very useful and use them
heavily, but the standard implementation is rather limited.  Perhaps
that would be the area to add the functionality I suggested, though.
I'm totally at the hand-waving stage on it, with no concrete ideas.
I just thought that if you were adding more type information,
oriented aournd the types themselves rather than index AMs, some form
of inheritence might fit in gracefully.
> in any case it's almost entirely unrelated to what I was talking
> about.
OK
-Kevin