From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from waste.org ([66.93.16.53]:38062 "EHLO waste.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751453AbXFQEr6 (ORCPT ); Sun, 17 Jun 2007 00:47:58 -0400 Date: Sat, 16 Jun 2007 23:46:36 -0500 From: Matt Mackall To: "John W. Linville" Cc: Andrew Morton , Linus Torvalds , Christoph Hellwig , johannes@sipsolutions.net, linux-wireless@vger.kernel.org, dcbw@redhat.com, hs4233@mail.mn-solutions.de, linux-kernel@vger.kernel.org, marcelo@kvack.org Subject: Re: libertas (private) ioctls vs. nl80211 Message-ID: <20070617044636.GL11166@waste.org> References: <1181757483.29767.99.camel@johannes.berg> <20070614170901.GA13384@infradead.org> <20070614103836.23b07a72.akpm@linux-foundation.org> <20070614181307.GA16856@infradead.org> <20070614190936.GA28917@tuxdriver.com> <20070614125618.a7338ad8.akpm@linux-foundation.org> <20070614200844.GC28917@tuxdriver.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <20070614200844.GC28917@tuxdriver.com> Sender: linux-wireless-owner@vger.kernel.org List-ID: On Thu, Jun 14, 2007 at 04:08:44PM -0400, John W. Linville wrote: > On Thu, Jun 14, 2007 at 12:56:18PM -0700, Andrew Morton wrote: > > On Thu, 14 Jun 2007 15:09:36 -0400 > > "John W. Linville" wrote: > > > > > It does not make sense to me to rip this out purely for aesthetic > > > reasons. > > > > Aesthetics are good, but that's not the main issue. > > > > What is most worrying is that there appears to be a risk that these > > newly-added interfaces will later become obsoleted by another interface. > > This means that we'll need to maintain, test and support _both_ interfaces > > for a very long time. This is the sort of foot-shooting we should avoid. > > True enough. However, I hope you will agree that we should not > confuse foresight with speculation. > > At present there is no sign on the horizon of either any mac80211 > mesh code or any other full-MAC wireless driver supporting mesh. > Without either of those, it would seem imprudent to rush toward > a gneric configuration interface even if nl80211 was prepared to > sprout one. Making it generic may be premature optimization. But at the very least, we should deal with the three other problems Christoph has pointed out: subfunctions, pointer indirections, and 32on64 cleanliness. That last one may come back to bite you in a couple years' time, if not other people. It's a major pain in the ass. -- Mathematics is the supreme nostalgia of our time.