From owner-freebsd-chat Sat Nov 23 19:14:46 2002 Delivered-To: freebsd-chat@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 655CC37B417 for ; Sat, 23 Nov 2002 19:14:44 -0800 (PST) Received: from picard.skynet.be (picard.skynet.be [195.238.3.88]) by mx1.FreeBSD.org (Postfix) with ESMTP id 47C5443E4A for ; Sat, 23 Nov 2002 19:14:43 -0800 (PST) (envelope-from brad.knowles@skynet.be) Received: from [192.168.0.3] (ip-26.shub-internet.org [194.78.144.26] (may be forged)) by picard.skynet.be (8.11.6/8.11.6/Skynet-OUT-2.20) with ESMTP id gAO3Ed918276; Sun, 24 Nov 2002 04:14:39 +0100 (MET) (envelope-from ) Mime-Version: 1.0 X-Sender: bs663385@pop.skynet.be Message-Id: In-Reply-To: References: <3DE00781.F47E59A1@mindspring.com> X-Grok: +++ath X-WebTV-Stationery: Standard; BGColor=black; TextColor=black Reply-By: Wed, 1 Jan 1984 12:34:56 +0100 Date: Sun, 24 Nov 2002 04:12:50 +0100 To: drew-dated-1038529607.6fb7a6@poured.net From: Brad Knowles Subject: Re: mail queues on softdep (was Re: Sharing calendars?) Cc: freebsd-chat@freebsd.org, Drew Raines Content-Type: text/plain; charset="us-ascii" ; format="flowed" Sender: owner-freebsd-chat@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org At 6:43 PM -0600 2002/11/23, Drew Raines wrote: >> Why are soft updates bad for mail queues, in your opinions? > > My opinions aren't authoritative because I've never (knowingly) > experienced it, but apparently a message could be lost if it's > accepted by the MTA and there's a system failure before the > metadata gets written to the disk. Softupdates gives us the ability to optimize out certain types of very fast file creation/deletion pairs, such as happens with mail messages being in the queue for only the briefest moments of time between our accepting the message and our successfully delivering the message to the next hop. Yes, softupdates does slightly increase the probability of loss of individual messages, if there should be a crash at a critical time. However, for those MTAs which take care to make sure that they perform certain file operations in a certain order (e.g., sendmail & postfix), this is generally not considered to be a violation of RFC 2821, section 6.1: 6.1 Reliable Delivery and Replies by Email When the receiver-SMTP accepts a piece of mail (by sending a "250 OK" message in response to DATA), it is accepting responsibility for delivering or relaying the message. It must take this responsibility seriously. It MUST NOT lose the message for frivolous reasons, such as because the host later crashes or because of a predictable resource shortage. For those MTAs which do not take care to perform those file operations in a certain order (e.g., qmail & exim), this is not true. They are inherently incompatible with softupdates, and in order to comply with this section of the RFC, you must either not use them, or you must turn off softupdates. > This could happen with sync, too, I guess, but the latency is > increased with soft updates. There might be characteristics of > qmail's queue which causes it to be more susceptible than others, > but the logic seems to apply to all. This is a specific problem with qmail and exim. This is not a general problem with softupdates as a whole. -- Brad Knowles, "They that can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety." -Benjamin Franklin, Historical Review of Pennsylvania. GCS/IT d+(-) s:+(++)>: a C++(+++)$ UMBSHI++++$ P+>++ L+ !E W+++(--) N+ !w--- O- M++ V PS++(+++) PE- Y+(++) PGP>+++ t+(+++) 5++(+++) X++(+++) R+(+++) tv+(+++) b+(++++) DI+(++++) D+(++) G+(++++) e++>++++ h--- r---(+++)* z(+++) To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-chat" in the body of the message