Re: Primary Key on Inherited Table

"Robert B. Easter" <reaster(at)comptechnews(dot)com> writes:
> Inherit is a very general term. I have broken it down into
> CLONES, EXTENDS, and ASSIMILATES (please consider):
> [ much snipped ]

I think you've put your finger on a fundamental issue: "inheritance"
seems to mean different things to different people, depending on how
they are trying to use it. Rather than arguing about who is right,
we have to recognize that they're all right in their own contexts.
The trick is to figure out how to support all those variant usages.

I'm not sure that you've proposed quite the right conceptual breakdown.
My inclination would be to try to specify several independent
implementation properties that can be mixed & matched to create the
behaviors that different applications want. But clearly you are
thinking about the right problem.

I hope that Chris Bitmead will pop up and contribute to this thread;
awhile ago he was bugging us regularly about shortcomings in Postgres'
inheritance support, but I'm afraid he might've got discouraged and
gone away :-(. (If you have not done so already, I suggest you search
the pghackers archives for threads mentioning inheritance. Chris
pointed out a lot of problems and interesting application examples
a year or so ago.)

I do have to tell you that most of the core developers are not thinking
much about inheritance --- we are busy with SQL92 compatibility,
performance, reliability, and other low-level concerns. I'd love to see
someone step up to the plate and start working on inheritance as such.

> Well, anyhow, I hope some of this will be considered when the inheritance
> system in PostgreSQL is maybe redesigned.

Do I hear a volunteer? ;-)

regards, tom lane