Hello,
At Thu, 16 Apr 2015 14:43:33 -0700, David Fetter <david@fetter.org> wrote in <20150416214333.GA797@fetter.org>
> On Wed, Apr 15, 2015 at 09:35:05AM +0900, Kyotaro HORIGUCHI wrote:
> > Hi,
> >
> > Before suppressing the symptom, I doubt the necessity and/or
> > validity of giving foreign tables an ability to be a parent. Is
> > there any reasonable usage for the ability?
...
> I have a use case for having foreign tables as non-leaf nodes in a
> partitioning hierarchy, namely geographic.
Ah, I see. I understood the case of intermediate nodes. I agree
that it is quite natural.
> One might have a table at
> HQ called foo_world, then partitions under it called foo_jp, foo_us,
> etc., in one level, foo_us_ca, foo_us_pa, etc. in the next level, and
> on down, each in general in a separate data center.
>
> Is there something essential about having non-leaf nodes as foreign
> tables that's a problem here?
No. I'm convinced of the necessity. Sorry for the noise.
At Wed, 22 Apr 2015 17:00:10 -0400, Robert Haas <robertmhaas@gmail.com> wrote in
<CA+TgmobZVHp3D9wWCV8QJc+qGDu7=tEKNCbXOwijZKhjuCmRWg@mail.gmail.com>
> Gee, I don't see why that would be unreasonable or invalid
Hmm. Yes, as mentioned above, there's no reason to refuse
non-leaf foregin tables. I didn't understood the real cause of
the problem and thought that not allowing foreign *root* tables
seem better than tweaking elsewhere. But that thought found to be
totally a garbage :(
regards,
--
Kyotaro Horiguchi
NTT Open Source Software Center