Date: Wed, 9 Feb 2005 02:23:09 +0000
From: Andrew Suffield <asuffield@debian.org>
To: gnu-arch-dev@lists.seyza.com, gnu-arch-users@gnu.org
Subject: Re: [GNU-arch-dev] Re: [Gnu-arch-users] how to fix a bad log message?
Message-ID: <20050209022309.GA21357@suffields.me.uk>
Mail-Followup-To: gnu-arch-dev@lists.seyza.com,
	gnu-arch-users@gnu.org
References: <F0CD2E5262214F44959AD6AF34ADA66E01D12FD1@mail2.netezza.com>
Mime-Version: 1.0
In-Reply-To: <F0CD2E5262214F44959AD6AF34ADA66E01D12FD1@mail2.netezza.com>
User-Agent: Mutt/1.5.6+20040907i
Cc: 
Precedence: list
Content-Type: multipart/mixed; boundary="===============0225381828=="
Mime-version: 1.0
Sender: gnu-arch-dev-bounces@lists.seyza.com
Errors-To: gnu-arch-dev-bounces@lists.seyza.com
Status: RO
Content-Length: 2428
Lines: 73

--===============0225381828==
Content-Type: multipart/signed; micalg=pgp-sha1;
	protocol="application/pgp-signature"; boundary="jRHKVT23PllUwdXP"
Content-Disposition: inline


--jRHKVT23PllUwdXP
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Tue, Feb 08, 2005 at 07:30:46PM -0500, John Yates wrote:
> Here is a half-baked idea...
>=20
> It seems to me that arch has conflated two use cases:
>=20
> 1) committing work to an archive
> 2) publishing committed results to the rest of the universe
>=20
> Not being an arch developer I can wave my hands and imagine
> decomposing commit into two distinct operations.  Consistency
> of mirrors would be achieved by having the code pulling a
> mirror copy harvest only published changes.  (Ideally a user
> updating a mirror could query some volatile/ephemeral state to
> discover the existence committed-but-not-yet-published work.)
>=20
> This would provide a well defined notion of what changes can
> be undone and which only be reversed.

Only if you are willing to accept that nobody, not even *you*, can
access the 'unpublished' stuff. Which makes it pretty useless
really. In essence you would rename 'commit' to 'publish' and
introduce a no-op called 'commit'.

This stubborn wrong-headed notion that you ever want to modify a
revision once it's been created needs to stop. You don't. Whatever
you're trying to do can and should be done in some other way. Usually
by committing a new revision that changes the offending thing again.

--=20
  .''`.  ** Debian GNU/Linux ** | Andrew Suffield
 : :' :  http://www.debian.org/ |
 `. `'                          |
   `-             -><-          |

--jRHKVT23PllUwdXP
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: Digital signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)

iD8DBQFCCXQNlpK98RSteX8RAuhfAJ49NrHgjD5fa/XO/FDWuC0E9kGl7gCfUHJr
dnVWmzdHVdzAQMhAstPz8rY=
=DCNa
-----END PGP SIGNATURE-----

--jRHKVT23PllUwdXP--

--===============0225381828==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Gnu-arch-dev mailing list
Gnu-arch-dev@lists.seyza.com
http://lists.seyza.com/cgi-bin/mailman/listinfo/gnu-arch-dev

--===============0225381828==--