> I am curious. I read this whole current thread. What is "wrong" with the
> patch? As I understand it it does not increase complexity. It appears to
> only expose (or perhaps abstract?) existing functionality into a usable
> API that is not dependent on something being in core.
>
> Imagine if at some point to develop new index types or perhaps single
> purpose modified index types all you needed was knowhow, pgxs and too
> much time.
>
> Unless there is something "wrong" with this patch I say we need to stop
> arguing semantics and apply it.
+1. One of the points Heikki made in his original post about
rejecting the patch was that the two examples Simon provided were not
very useful: one was a one-database-only rmgr that didn't actually
work and the other was an unimpressive logging plugin. So maybe there
is an argument that Simon should, say, throw away the logging plugin
and finish the other one so that it actually works. But other than
that... if no one uses it, who cares? It's not a lot of code. If
someone does use it, and they mix two incompatible versions and trash
their database... well, there are lots of ways to trash your database
by loading arbitrary C code into your database engine. So, uh, don't
do that unless you know what you're doing.
We allow extensibility and hooks in other parts of the database where
the use case is pretty thin and tenuous. I betcha there aren't many
people who try writing their own eqjoinsel() either.
...Robert