Thomas Munro <thomas.munro@gmail.com> writes:
> Playing the devil's advocate here: why shouldn't we routinely drop
> support for anything that'll be EOL'd when a given PostgreSQL major
> release ships?
I don't like the word "routinely" here. Your next bit is a better
argument:
> Something EOL'd over a year ago that has a bunch of features we've
> really always wanted, like Unix domain sockets and Unix link
> semantics, seems like a reasonable choice to me...
My general approach to platform compatibility is that when we
break compatibility with old versions of something, we should do so
because it will bring concrete benefits. If we can plausibly
drop support for Windows versions that don't have POSIX rename
semantics, I'm 100% for that. I'm not for dropping support for
some platform just because it's old.
regards, tom lane