From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755845AbXH2Of6 (ORCPT ); Wed, 29 Aug 2007 10:35:58 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752740AbXH2Ofv (ORCPT ); Wed, 29 Aug 2007 10:35:51 -0400 Received: from wx-out-0506.google.com ([66.249.82.227]:28454 "EHLO wx-out-0506.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752525AbXH2Ofu (ORCPT ); Wed, 29 Aug 2007 10:35:50 -0400 DomainKey-Signature: a=rsa-sha1; c=nofws; d=googlemail.com; s=beta; h=received:from:to:subject:date:user-agent:cc:references:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:message-id; b=LdW4ubamhNDUQLPa1UPRU9nI6EttPCFmiQUXCv9QJ4kaNpxxneRGwu9iWCrKCg7dNnae6jq785DfJrmKP+4naynV0WwhS8g1xwjz79GNVKje7F94XbhxwbrCPcRH05HMVQqp+8E+5JEoA4laOww3S7nIUaHdGCaFXHrKS01Apdw= From: Denys Vlasenko To: "Robert P. J. Day" Subject: Re: ok to kill "ether=" kernel parm? Date: Wed, 29 Aug 2007 15:35:41 +0100 User-Agent: KMail/1.9.1 Cc: "H. Peter Anvin" , Linux Kernel Mailing List References: <46D45A02.6010306@zytor.com> In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200708291535.41594.vda.linux@googlemail.com> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Tuesday 28 August 2007 18:58, Robert P. J. Day wrote: > On Tue, 28 Aug 2007, H. Peter Anvin wrote: > > Robert P. J. Day wrote: > > > On Tue, 28 Aug 2007, H. Peter Anvin wrote: > > >> Robert P. J. Day wrote: > > >>> given that "ether=" has been officially obsolete since 2.6.18 > > >>> (replaced by "netdev="), is there any reason to keep it around? > > >>> or can it be blasted? > > >> > > >> That sounds like way too short of a timeline for breaking people's > > >> working boot setup. For a lot of people, 2.6.18->current is going > > >> to be a single step. > > > > > > actually, now that i look more closely at the code browser at > > > lxr.linux.no, "ether=" has been listed as "obsolete" since *at least* > > > 2.6.10. not to sound unsympathetic but anyone who tries to jump from > > > 2.6.10 to 2.6.24 in one step deserves what they get. :-) > > > > > > ok, that was cruel, but you see my point, right? > > > > Yes, and I think it's quite pointless. > > > > The thing is, people's boot setups have probably been that way since > > *long* before 2.6.9. They continue to work, as they should, so they > > aren't changed. This is why we very rarely break boot interfaces > > (and this is a user-visible interface you're talking about); we're > > still supporting interfaces that have been obsolete *SINCE BEFORE > > 1.0 WAS RELEASED.* > > > > What's the upside of changing? What's the downside? The upside is > > so infinitesimal that that leaving "ether=" in indefinitely seems > > like a good move to me. > > i've never found these "well, it's not hurting anything" arguments > terribly compelling. if that's the case, why remove *anything* from > the kernel? why obsolete *anything*? but that's not my actual point. > > why continue to support two different ways to do the same thing? in > situations like that, i can imagine the following (admittedly > hypothetical) conversation between old-timer and young geek: > > OT: "so, what the problem?" > YG: "i can't get my network module to work properly. i use modprobe > with netdev= and ..." > OT: "huh? netdev? why don't you use ether=?" > YG: "what's ether=?" > OT: "what's netdev=?" > > followed by a confused conversation as to whether they really > represent the same thing, or maybe not, or maybe mostly. > > if you want to keep the old way of doing it, that's cool. but it > would be nice if, in cases like that, a clear choice was made. if you > want to keep the old way, then *keep* it. make it clear that it's > official, and supported. > > or if you're going to delete it, then *delete* it. but let's not keep > doing this half-way, half-assed measure of tagging something as > obsolete, then just letting it hang out in the kernel forever. either > keep it, or delete it, and stop being so wishy-washy and doing things > halfway. > > and, finally, while "there's more than one way to do it" may be a > terrific perl philosophy, i don't think much of it as a kernel coding > standard. Add a printk("Deprecated, use netdev=xxx\n"); to the handler. After 1-2 years you can remove ether=xxx. -- vda