On Fri, May 20, 2016 at 2:25 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> "David G. Johnston" <david.g.johnston@gmail.com> writes:
> > On Fri, May 20, 2016 at 2:02 PM, John R Pierce <pierce@hogranch.com>
> wrote:
> >> pg_dump and pg_dumpall are supposed to be able to dump all supported
> prior
> >> versions, thats the documented procedure for updating... I contend
> this
> >> *is* a bug. quote-all-identifiers is a workaround, for sure.
>
> > If indeed the prescribed procedure is quote-all-identifiers I'd say we
> > should at least update the docs.
>
> Yeah, agreed. I'll put something in.
>
>
=E2=80=8BI'd hit up the "Notes" section's last paragraph in addition to the
command-line option. The former is going to be much more visible. The
cross-version dynamic possibly warrants it own section header, IMO, given
that it is a primary use case for the tool.=E2=80=8B Consider placing it
subsequent to the Description section - maybe calling it "Upgrades".
Alternatively, we could automatically turn on quote-all-identifiers when
> dumping from a server of a different major version; but I imagine we'd
> get complaints about that, too, since it's so seldom an issue.
>
>
True, the volume of people using even potentially reserved words as
identifiers is not all that high. Given fail-fast behavior, and a
reasonable pointer to overcoming it, catering to this minority doesn't seem
worth disrupting the status quo; even in this limited fashion.
=E2=80=8BDavid J.=E2=80=8B