Re: INSERT ... ON CONFLICT IGNORE (and UPDATE) 3.0

Dean,

* Dean Rasheed (dean(dot)a(dot)rasheed(at)gmail(dot)com) wrote:
> In all of this, I think we should try to keep things as simple as
> possible, to give the end user a chance to understand it --- although
> I'm not sure I've achieved that with my explanation :-)

Thanks a lot for this. It goes along with my thinking also and matches,
I believe, what I had explained to Peter on our call.

Peter, please let me know if you agree.

Dean, I've been working through your patches over the past couple of
days (apologies for the lack of updates, just been busy) and hope to
push them very shortly (ie: by the end of the weekend).

One thing that I was hoping to discuss a bit is that I've gone ahead and
added another set of hooks, so we can have both "permissive" and
"restrictive" policies be provided from the hook. It's a bit late to
make the grammar and other changes which would be required to add a
"restrictive" policy option to the built-in RLS, but adding the hooks is
relatively low-impact.

I'm also going to be including a test_rls_hooks module into
src/test/modules which will test those hooks and provide an example of
how to use them.

As for the discussion- there was some concern raised about extensions
being able to "override" built-in policies by using the hooks a certain
way. I don't entirely follow the logic behind that concern as an
extension has the ability to read the files on disk directly from C
code, should it be written to do so, and so not providing a hook to add
"permissive" policies is denying useful functionality for very question
gain, in my view at least.

Thoughts?

Thanks!

Stephen