From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S265215AbTLaRkp (ORCPT ); Wed, 31 Dec 2003 12:40:45 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S265216AbTLaRko (ORCPT ); Wed, 31 Dec 2003 12:40:44 -0500 Received: from 80-169-17-66.mesanetworks.net ([66.17.169.80]:28387 "EHLO mail.bounceswoosh.org") by vger.kernel.org with ESMTP id S265215AbTLaRkn (ORCPT ); Wed, 31 Dec 2003 12:40:43 -0500 Date: Wed, 31 Dec 2003 10:42:06 -0700 From: "Eric D. Mudama" To: linux-kernel@vger.kernel.org Subject: Re: [PATCH] 2.6.0 - Watchdog patches (BK consistency checks) Message-ID: <20031231174206.GA14338@bounceswoosh.org> Mail-Followup-To: linux-kernel@vger.kernel.org References: <20030906125136.A9266@infomag.infomag.iguana.be> <200312300836.16559.edt@aei.ca> <20031230131350.A32120@hexapodia.org> <200312311001.48154.edt@aei.ca> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Disposition: inline In-Reply-To: <200312311001.48154.edt@aei.ca> User-Agent: Mutt/1.5.4i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Dec 31 at 10:01, Ed Tomlinson wrote: >I am not saying I do not want do have consistency checks done. I do want >to control _when_ and how often they run I don't see how it would be valid to not do a consistency check after every network operation, which is what it does now... Every time you are modifying your archive from a remote source, the consistency check makes sure you don't have corruption in the transfer that was previously undetected, or corruption in the underlying archive which is about to accept the changes. Since you can only do network updates at a time when your own archive is in a "checked-in" state, if the network update fails or detects corruption, you can clone what you'd already checked in to a new location and recover trivially. The most work you can lose is the delta between your last changeset and the current one, assuming you keep some sort of backups around. What exactly would you do if you'd been working "offline" for 3 weeks, went to sync, and it said that your archive was corrupted? Or should it try to merge into a corrupt archive anyway? Should that archive that was corrupt, and then had possibly good changes layered on top of it, be valid for attempting a clone? I think that limiting the consistency checking could be like opening a *very big* can of worms. -- Eric D. Mudama edmudama@mail.bounceswoosh.org