Christopher Browne <cbbrowne@gmail.com> writes:
> On Fri, Aug 23, 2013 at 3:10 AM, Fabien COELHO <coelho@cri.ensmp.fr> wrote:
>> Would it make sense for such identifiers be standard UUID
>> (http://en.wikipedia.org/wiki/UUID)?
> There is sense to this, sure.
That ship's already sailed, though. As was pointed out upthread, we don't
really want to change the way that pg_controldata prints the system ID,
and we don't want this SQL function printing something different either.
> I'd think that constructing a Type 5 (SHA-1) UUID based on some local
> information would make a lot of sense.
> In effect, based on constructing SHA-1 on a string looking like:
> "Database system identifier: 5651554613500795646
> Maximum data alignment: 8
> Database block size: 8192
> WAL block size: 8192
> Maximum length of identifiers: 64
> Date/time type storage: 64-bit integers
> Version: PostgreSQL 9.1.1 on x86_64-unknown-linux-gnu, compiled by
> gcc (Debian 4.6.1-4) 4.6.1, 64-bit"
> ==> SHA-1 of b1b012cc85149d2fe4bf0fc18c38dcf1218e95a5
Including the version string would be a seriously bad idea --- you
don't want the sys ID to change just because you did a minor version
upgrade, or even recompiled the same version with a newer compiler,
do you? There might be some point in factoring in those other values,
but I'm not terribly excited about them either.
regards, tom lane
