Re: Patch (2): Implement failover on libpq connect level.

On 10/24/15 7:55 AM, Robert Haas wrote:
> On Fri, Oct 23, 2015 at 4:02 PM, Korry Douglas
> <korry.douglas@enterprisedb.com> wrote:
>> When you call pg_is_in_recovery(), you should schema-qualify the function
>> name, just in case some other version of that function exists in the
>> search_path.
> 
> I wonder whether it's really a good idea to put this kind of logic
> into libpq at all.  I think there was some previous votes against
> doing so, and I tend to agree with that viewpoint.  Shouldn't probing
> for the state of the connection be the caller's job, not libpq's?  If
> somebody wants to write a wrapper function around this that runs this
> query after connecting - or any other query - they can do so.

Well, earlier there was a debate whether any of this should be in libpq.

I think asking for a read-only or read-write connection would be quite
useful, and there isn't really a lot of room for interpretation that
would lead to a lot of application-specific implementations.

That said, it might still be best to omit this from the first round of
the patch, to simplify the discussion.