On Tue, Jul 15, 2003 at 04:19:34PM -0400, Andrew Sullivan wrote:
> I thought that now would be a good time to ask whether it should
> live as a separate project, or whether it should be in contrib. I
I have run into a number of arguments for putting the project on
gborg. Unless I hear strong arguments against that option, I think
I'll set up a project on gborg.
Here are the arguments I've got:
1. The code as it stands is not really a full replication
system, but more a toolkit. We at Liberty had a whole bunch of
bespoke stuff that we used with the eRServer code in order to make it
useful to us (various scripts that ran outside, for instance). That
can't be released, because (a) it doesn't help anyone else and (b) it
might reveal something about our internal databases (which would get
me in Big Heck).
2. Building the code is a pain, because it is real sensitive to
JDK versions, Ant installations, &c. As a result, the current CVS
tree actually has a specific version of ant in it; I'm sure I don't
have time to revisit the build scripts even if I knew what to do. So
we have to ship all this ancillary code around too, and that will
bloat the tarball and possibly cause a lot of additional support
requests.
3. The code as it stands works with everything back to 7.2, and
it'd be a shame to make people download a whole postgres tarball just
to get the replication code.
Arguments? None of those three address the obvious marketing benefit
of having replication shipping with the main tarball, I know.
I have most of a basic installation doc written, and a FAQ of gotchas
and "why doesn't this work" answers. I'll be delivering that as soon
as it's done to PostgresSQL, Inc; I hope in the next couple days.
A
--
----
Andrew Sullivan 204-4141 Yonge Street
Liberty RMS Toronto, Ontario Canada
<andrew@libertyrms.info> M2P 2A8 +1 416 646 3304
x110
