From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrick McHardy Subject: Re: [PATCH 4/4] netfilter: xtables: schedule xt_state for removal Date: Thu, 25 Mar 2010 11:06:23 +0100 Message-ID: <4BAB359F.6030308@trash.net> References: <1269377101-13875-1-git-send-email-jengelh@medozas.de> <1269377101-13875-5-git-send-email-jengelh@medozas.de> <4BAA2969.8010106@trash.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 7bit Cc: netfilter-devel@vger.kernel.org To: Jan Engelhardt Return-path: Received: from stinky.trash.net ([213.144.137.162]:48131 "EHLO stinky.trash.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753246Ab0CYKGX (ORCPT ); Thu, 25 Mar 2010 06:06:23 -0400 In-Reply-To: Sender: netfilter-devel-owner@vger.kernel.org List-ID: Jan Engelhardt wrote: > On Wednesday 2010-03-24 16:02, Patrick McHardy wrote: >> Jan Engelhardt wrote: >>> xt_conntrack has been provided since v2.5.32. >>> >> I'm fine with the removal of old revisions, but how are you planning on >> informing users about removal of this module? Most people don't read >> feature-removal-schedule, and distributions are unable to help with >> user written scripts. > > I would suggest to do the same as we did with disallowing DROP in the > nat table: > > - a message printed by iptables whenever -m state is used > > - a kernel message whenever whenever a rule with xt_state is created > > We did not actually do the kernel side with nat-prohibit-DROP, but I > regard it as very useful, as the community was very much able to help > itself if only they got the word - and it turned out that dmesg is > _the_ place people look in especially when they don't supervise > iptables output directly, as with, for example, boot splash where > messages are hidden, or server/router devices that one tends to > forget about. Yes, a kernel message sounds fine and less annoying than an iptables message since we can limit it to print only once. I'm not really convinced of removing state though, I has never caused any maintenance overhead, it requires a lot less memory than xt_conntrack and it seems more intuitive to write "-m state" than "-m conntrack --ctstate" to me.