On 2/25/19 11:59 PM, Christophe Pettus wrote:
On Feb 25, 2019, at 13:38, Andres Freund <and...@anarazel.de> wrote:
I think you might be right about this specific issue. But to me it
sounds like you also don't appreciate that development resources are
really constrained too, and providing endless backward compatibility for
everything is going to use both resources directly, and indirectly by
making the whole system more complex.
One of the things I've tried not to do in this discussion is turn it into a
policy debate about backwards compatibility in general, rather than this very
specific feature. Of course, there's a cost to keeping around features, and I
fully appreciate that. In this discussion, the code complexity didn't seem to
be the fundamental driver behind removing the feature; the relative safety of
it was, along with maintaining the documentation.
Code complexity is very much an issue. Since the two methods are merged
with a ton of conditionals, working on one without affecting the other
is a real chore, especially since exclusive backups have *zero* tests.
Personally I'm not interested in writing those tests so I try not to
touch the code either.
I jumped in because there seemed to be an argument going on that all of the
cool kids will have moved off the old interface and there was essentially no
cost to removing it in v12 or v13, and that didn't correspond to my experience.
I don't think anyone believes there will be zero cost, but the issues
with the exclusive method seem to outweigh the benefits, at least in my
experience.
--
-David
da...@pgmasters.net
