Re: postgresql embedded mode

On Mon, May 23, 2016 at 9:17 AM, Albe Laurenz <laurenz(dot)albe(at)wien(dot)gv(dot)at>
wrote:

> David G. Johnston wrote:
> > On Mon, May 23, 2016 at 6:54 AM, aluka raju <alukaraju2894(at)gmail(dot)com>
> wrote:
> >> As given in the FAQ's that postgresql cannot be embedded
> >> https://wiki.postgresql.org/wiki/FAQ#Can_PostgreSQL_be_embedded.3F .
> >>
> >> Is their any possibility to make it embedded. Till now postgresql has
> not done this embedded
> >> mode and i want to work on how it can be embedded and contribute.
> please help me how to start or
> >> suggest the idea-.
> >
> > ​Step 1 - Fork the project...
> >
> > ​Changing the fundamental architecture of the system is not something I
> would expect a project with
> > this long of history to attempt to do directly. While the people
> working on the core product are
> > welcome to spend their time however they like I don't imagine any of
> them would be willing to commit
> > code to core pertaining to this capability. Most, if not all, of them
> likely don't believe it is even
> > a good idea generally.
>
> There was this patch by Tom Lane in 2012:
> http://www.postgresql.org/message-id/12511.1346631791@sss.pgh.pa.us
>
> This can be used as something like an embedded database.
> Nothing more happened with this patch, though.
>

​Good read.

​The patch as written was really geared toward making PostgreSQL's
single-user mode accessible from psql - which itself is an administrative
mode.

​The main concern is that if you give people an inch here they will take
that capability and place in front of users ​and then complain that our
product sucks when the database isn't being vacuumed and checkpoints aren't
happening and there is no way to connect to the database externally while
the application is running (or from programs that rely on JDBC for that
matter).

There is a non-trivial gap between what that patch proposed to accomplish
and what an embedded configuration would look like.

In reading the thread I wish Tom hadn't brought up the concept of embedded
- and flatly rejected any attempt to allow it into the discussion (aside
from discussion of potential real world use cases where people abuse the
capability beyond what we explicitly document it to be supported for).

If someone were to present a fully functioning, well designed and
documented, ideally in-the-wild, patch/fork of PostgreSQL that operates in
embedded mode I'm sure the community and core would be willing to review it
and render their multitude of opinions. Beyond, and because of, that we
are a FOSS community and the people involved are welcome to purse whatever
endeavors they wish. The aforementioned thread is a good start, though I
suspect there is more material available, if someone were to consider going
down this path.

David J.