Content-Type: text/plain
Content-Disposition: inline
Content-Transfer-Encoding: 7bit
MIME-Version: 1.0
From: "Peter Haworth" <pmh@edison.ioppublishing.com>
To: David Wheeler <david@wheeler.net>, Tom Lane <tgl@sss.pgh.pa.us>
Cc: Tim Bunce <Tim.Bunce@pobox.com>, dbi-dev@perl.org,
	pgsql-interfaces@postgresql.org
Subject: Re: DBD::PostgreSQL
Date: Tue, 19 Nov 2002 12:19:48 +0000
Message-Id: <PGM.20021119121948.8441.2705@edison.ioppublishing.com>
References: <8F8B367F-FB19-11D6-93B3-0003931A964A@wheeler.net>

On Mon, 18 Nov 2002 09:17:05 -0800, David Wheeler wrote:
> On Monday, November 18, 2002, at 09:12 AM, Tom Lane wrote:
> > As of 7.3 that's not necessarily so anymore; you can "SET autocommit TO
> > off" and get the behavior where any statement starts a transaction block
> > (and so an explicit COMMIT is required to commit its effects). Not sure
> > if this helps you or not.
>
> PostgreSQL gets better and better. Yay. However, although I might be able
> to use compile-time macros to determine the PostgreSQL version, I have to
> support a minimum version of PostsgreSQL in the driver.

You'd also need a runtime check that the server you connected to was of a
sufficiently high version. I realise it would make things more complicated,
but would it be possible to keep the manual transaction starting behaviour,
as well as adding the 7.3 behaviour, and deciding which to do based on the
server version?

Also, I'd like booleans to be returned as 't'/'f', rather than 1/0, to match
the behaviour of other drivers. Or at least have a driver-specific flag to
control which values get used.

-- 
	Peter Haworth	pmh@edison.ioppublishing.com
"To vacillate or not to vacillate, that is the question ... or is it?"