Re: Atomic rename feature for Windows.

On Fri, 8 Apr 2022 at 11:30, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
>
> On Fri, Apr 8, 2022 at 10:12 AM Greg Stark <stark(at)mit(dot)edu> wrote:
> > On Thu, 9 Dec 2021 at 23:36, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> > > I'm not for dropping support for some platform just because it's old.
> >
> > I guess I'll have to spin up the Vax again :)
>
> This is a pretty good summary of what's wrong with our current
> deprecation policy.

I didn't intend it that way but, ok.

> Like Tom, I kind of hate removing support for old
> systems. But I've also come to realize that we often end up supporting
> systems because there's one PostgreSQL developer who has access and
> sets up a buildfarm member ... which tends to mean that we support all
> the stuff that lots of people are using, plus a pretty random subset
> of older systems that do funny things and most people can't access to
> debug any problems that may occur. And that's kind of annoying.

Generally I think supporting older systems that do funny things is
helpful in avoiding problems that either 1) Can happen on newer
systems but rarely 2) Can happen on other systems that people are
using but we don't know about and aren't testing and 3) Can happen on
future systems or future compilers and we might not even find out
about.

But that's useful for some things and not for others. Like, it's
useful to be sure we don't have odd dependencies on timing quirks of
the specific machines that are currently common, or depend on gcc/llvm
compiler behaviour that isn't guaranteed. But less so for supporting
some quirky filesystem behaviour on Windows 8 that newer Windows
doesn't have and Unix guarantees not to have. (Or supporting non-IEEE
Vax FP now that we've decided we just don't any more).

--
greg