Re: Extension Templates S03E11

* Dimitri Fontaine (dimitri(at)2ndQuadrant(dot)fr) wrote:
> Stephen Frost <sfrost(at)snowman(dot)net> writes:
> > What I've been trying to point out is that there's absolutely zero need
> > for the 'extension template' part of this to make a pg_restore work for
> > an entirely-in-the-catalog extension. I realize that's how you've done
> > it with this patch set but that doesn't make it necessary.
>
> If it's an extension, it's filtered out of pg_dump, so it's not part of
> your pg_restore. Full Stop. This point has been debated and there has
> been a very clear conclusion a year ago.
>
> What am I missing here?

I don't buy it. I understand that you heard that from Tom with regard
to extensions- but that was also when he was argueing that what you're
building here shouldn't be called extensions. I have a huge problem
with the idea that we can't dump out the definition of an
in-catalog-only extension, but it's okay to dump out some big blob of
text which is supposed to represent the same thing but in a much, much
more kludgey and inexact way. How are we going to handle new keywords
being added in new major versions? A pg_dump of the extension template
script is then going to be loaded into the new major version but will
not actually be able to be run because it'll error out...

I'm 100% confident that's not the only problem with this approach which
sticks script text blobs into the catalog as representations of
database objects.

Thanks,

Stephen