On Tue, Nov 3, 2015 at 2:57 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Paul Ramsey <pramsey@cleverelephant.ca> writes:
>> [ 20151006b_postgres_fdw_extensions.patch ]
>
> Starting to look through this now. I'm dubious of the decision to have
> ExtractExtensionList throw errors if there are un-installed extensions
> mentioned in the FDW options. Wouldn't it be a lot more convenient if
> such extension names were silently ignored? You cannot guarantee that the
> list is always up to date anyway; consider creating a server, setting some
> extension options, and then dropping some of those extensions. Moreover,
> the current semantics create a hazard for pg_dump, which can't reasonably
> be expected to know that it must restore extensions X and Y before it can
> create foreign server Z.
>
> There might be a case for raising a WARNING during
> postgres_fdw_validator(), but no more than that, IMO. Certainly ERROR
> during regular use of the server is right out.
Agreed. I don't know whether it's better to emit a WARNING or some
lower-level message (INFO, DEBUG), but I think an ERROR will suck due
to the pg_dump issues you mention.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
