On Mon, Feb 25, 2019 at 11:20:14AM -0500, Tom Lane wrote:
> It appears to me that f831d4acc required a good deal more adult
> supervision than it actually got. That was alleged to be a small
> notational refactoring, not a redefinition of what gets put into
> dump files.
How much consistent do we need to be for custom dump archives
regarding backward and upward-compatibility? For dumps, we give no
guarantee that a dump taken with pg_dump on version N will be
compatible with a backend at version (N+1), so there is a effort for
backward compatibility, not really upward compatibility.
It seems to me that f831d4acc should have bumped at least
MAKE_ARCHIVE_VERSION as it changes the dump contents, still it seems
like a lot of for some refactoring? FWIW, I have gone through the
commit's thread and I actually agree that instead of a mix of empty
strings and NULL, using only NULL is cleaner.
--
Michael