On Thu, May 13, 2021 at 11:04:54AM +0900, Michael Paquier wrote:
> On Wed, May 12, 2021 at 01:30:27PM -0700, Andres Freund wrote:
> > That said, I don't think it's a good practice to use the control file
> > version as an identifier for the major version. Who knows, it might be
> > necessary to add an optional new format in a minor version at some point
> > or such crazyness. And then there's the beta stuff you'd mentioned, etc.
>
> Yes, PG_VERSION, as you wrote upthread already, is already fine for
> the job, and FWIW, I have yet to see a case where being able to easily
> detect the minor version in a data folder matters.
Would it even make any sense? It could be "has version X.Y ever started the
cluster", or "was it the last version that started the cluster", or something
else?
> And, I am of the opinion to not change the control file version if
> there is no need to do so.
+1