From: craig at 2ndquadrant.com (Craig Ringer)
Date: Mon, 20 Jul 2015 10:10:02 +0800
Subject: [Pljava-dev] The pljava Project is dead (LONG but I hope
	interesting)
In-Reply-To: 
 <CALBNtw4w+rVHWyR+38vBtwDYfgtS=ihp3Kmx41GQVJxG7vsROQ@mail.gmail.com>
References: <54DCEC2D.5050909@cgf.upv.es> <54DCF04D.9080805@manifest.co.uk>
 <54DD08B7.5070902@tada.se> <55AC048F.30401@anastigmatix.net>
 <CAMsr+YGJ6MPec5OPgp-Rh2i8mVQn8A6ToeEMsPa04nJYxaE7-g@mail.gmail.com>
 <CALBNtw4w+rVHWyR+38vBtwDYfgtS=ihp3Kmx41GQVJxG7vsROQ@mail.gmail.com>
Message-ID: 
 <CAMsr+YE5YyU96ho9=qqxa3nhAnpaVgRg2DKRSaLnEfoV5e-uhQ@mail.gmail.com>

On 20 July 2015 at 08:42, Bear Giles <bgiles at coyotesong.com> wrote:
> It's a big jump in scope but don't some databases have a native type for
> java classes and a way to execute them? (It's not hard to imagine a UDT for
> jar files as well.) It's not always possible to load jar files from the
> operating system.

Personally that fills me with horror, but each to their own.

You should be able to do this with PL/Java already if you really want
to. Just implement a custom classloader.

Anyway, that's a totally separate topic to how to make Pl/Java easier
to maintain, more attractive to use/work on, and more lively looking.

Personally I think the underlying impedence mismatch between
PostgreSQL shared-nothing-by-default multiprocessing architecture and
Java's shared-everything-by-default multithreading model limits the
appeal and utility of PL/Java as much as anything. Especially since
the JVM doesn't like being fork()ed and is designed around a
heavyweight startup long-running process approach. That said, Pl/Java
could expose the new dynamic shared memory features in PostgreSQL to
applications, which would permit apps to pass serialized objects
around via shared memory, and that'd meet most inter-process
communication needs.

-- 
 Craig Ringer                   http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services