Excerpts from Robert Haas's message of Thu Nov 24 16:02:38 +0200 2011:
>
> > So, in that light, do we still think that letting the user specify a
> > service name in the URI makes sense? (My personal opinion is yes).
>
> service is just a connection parameter, so if we choose a URL format
> that allows any connection parameter to be specified, this falls out
> naturally, without any additional work. And if we don't choose such a
> URL format, we are, in my humble opinion, crazy.
The patch draft I have uses that format, yes: so any keyword libqp recognizes can be given in form of param=value URI
queryparameter.
> e.g. if we used the format suggested in my previous email, this would
> just boil down to:
>
> postgresql:///?service=foo
Oh, well, that would make sense. It also appeared to me that we should deny overriding host, port and dbname by the
queryparameters to prevent confusion, e.g:
postgresql://host:port/dbname?host=otherhost&port=otherport&dbname=otherdb
--
Alex