On Sep 25, 2017 1:39 PM, "Joshua D. Drake" <j...@commandprompt.com> wrote:
> On 09/25/2017 10:32 AM, Petr Jelinek wrote: > >> On 25/09/17 19:26, Tom Lane wrote: >> >>> Alvaro Hernandez <a...@ongres.com> writes: >>> >> > >> There is already about 3 million output plugins out there so I think we >> did reasonable job there. The fact that vast majority of that are >> various json ones gives reasonable hint that we should have that one in >> core though. >> > > And I am sure that 2ndQuadrant would be happy to add it to their version > of Postgres and maintain it themselves. > > That's not the case to be thinking about... Of course, 2Q can support some extensions for their customers that use their binaries, that is not a contraindication for having more in core. I'd rather think based on questions like... "Will that be supportable in an ordinary Amazon instance?" Or "... In a Heroku instance?" Those (Amazon and Heroku) are places that Slony won't run because it needs a bit too much in the way of database superuser capabilities. It's a very useful exercise to know which bits of this are easy to resolve versus difficult versus irreconcilable. Of course, we can't force everything into core, but given what IS in core, it can look mighty dumb if... a) we have neat and crucial features in core that are (validly!) trumpeted in release notes, but b) those features aren't usable without substantial out-of-core extensions that many users cannot use. Perhaps it's valid for logical replication to be considered out-of-scope for generic Amazon/Heroku instances, but I'd prefer that to be the result of a reasoned intent.
