Bruce Momjian <bruce@momjian.us> writes:
> Tom Lane wrote:
>> My proposal would be to continue to accept the option but just ignore it
>> (ie, error out on version mismatch whether or not -i is given). This
>> way we wouldn't break any scripts that use the option, but things would
>> still be safe.
> A larger question is why the option was added in the first place.
It probably seemed like the conservative choice at the time: allow the
user to be smarter than pg_dump when necessary. What we couldn't have
foreseen was the way the option has been abused by tools that are not as
bright as they think they are. With the current situation where -i is
used by default, without the user's knowledge (and without showing him
the warning messages, which is why your patch isn't going to improve
matters), it just seems too dangerous to continue to accept the switch.
(I wonder whether some of the complaints we've seen about broken
dump/restore are courtesy of pgAdmin forcing the dump to be taken with
a too-old copy of pg_dump.)
One point after looking back at the previous discussion is that the
current version test is too strict: it will complain if your server is
8.2.7 and pg_dump is 8.2.6. We probably should not make a newer minor
number a hard error, since 99.99% of the time it would be fine. So
while I think newer major should be a hard error regardless of -i,
we could consider several responses to newer minor:* silently allow it always* print warning and proceed always* allow
-ito control error vs warning for this case only.
regards, tom lane