> On Aug 28, 2018, at 12:01 PM, Justin Clift <justin@postgresql.org> wrote:
>
> On 2018-08-28 18:32, Bruce Momjian wrote:
>> On Tue, Aug 28, 2018 at 01:19:28PM -0400, Jonathan Katz wrote:
>>> On Aug 28, 2018, at 1:06 PM, Magnus Hagander <magnus@hagander.net> wrote:
>>> Whether it's the EDB one or another one is a separate discussion, but we've
>>> been recommending the EDB one for a long time and none of our users have
>>> (AFAIK) complained about it (except when download servers have been down.
>>> Personally I'd really like to see the binaries hosted on the postgresql.org
>>> servers instead, but again that's an at least partially separate
>>> discussion).
>>> Partially separate discussion, but probably a separate discussion. However,
>>> since you opened it up, I would say +1 for it, but I understand there are
>>> additional considerations the infrastructure needs to make around hosting
>>> that, in addition to “which installer is chosen.”
>> Well, if we have another installer provider who has similar or better
>> performance for things we care about, e.g. reliable updates, support,
>> _and_ is willing to host the binaries where we want them, we can switch
>> to them. While it is kind of disloyal to change preferred providers, we
>> also didn't agree to have them be preferred forever, especially if
>> someone better shows up. Of course, switching might be disruptive for
>> our installers uses, so we have to consider that too.
>
> An alternative approach would be for us ("the project") to build official
> installers ourselves, with those being the recommended ones.
That would make sense if the PGDG installers were the "best" for
some definition of such. That seems like it'd be the place to start if
we wanted to make them the preferred ones.
In the macOS world there are a wider variety of PostgreSQL installers,
with different focuses (good for development, good for production, similar
to a unix install) and they're all useful.
> Whether it be 100% volunteer based, someone we hire (several people have
> pointed out we're not short of funds), or whatever. Just from the "neutral
> third party" (not vendor owned) perspective. ;)
If "we" wanted to put together a Windows installer it might be nice to start
with a smaller project to see what installer toolchain (e.g. wix) we wanted to
use, and what features we wanted to support.
A fairly regular request from Windows users is an installer for just the client
binaries - psql, libpq etc.. That would be a useful product in itself, as well
as a good proof of concept.
That's probably diving way off topic for -www though.
Cheers,
Steve