[From nobody Sat Mar 24 09:36:40 2007
Return-Path: <itojun@itojun.org>
X-Original-To: erno@erno.iki.fi
Delivered-To: erno@erno.iki.fi
Received: from ainavaan.iki.fi (ainavaan.iki.fi [212.16.98.51])
	by erno.iki.fi (Postfix) with ESMTP id BEBD91889AF
	for <erno@erno.iki.fi>; Fri, 23 Mar 2007 11:36:35 +0200 (EET)
Received: from coconut.itojun.org (coconut.itojun.org [221.249.121.227])
	by ainavaan.iki.fi (8.13.8/8.13.8) with ESMTP id l2N9abMS022078
	for <erno@iki.fi>; Fri, 23 Mar 2007 11:36:38 +0200 (EET)
Received: by coconut.itojun.org (Postfix, from userid 501)
	id BC6A61C04E; Fri, 23 Mar 2007 18:36:28 +0900 (JST)
To: erno@iki.fi
Cc: itojun@itojun.org
Subject: Re: (fwd) Re: [Python-3000] IPv6 address tuple
In-Reply-To: Your message of "Fri, 23 Mar 2007 08:56:36 +0200"
	<20070323065636.GN14803@erno.iki.fi>
References: <20070323065636.GN14803@erno.iki.fi>
X-Mailer: Cue version 0.8 (061214-1013/itojun)
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Message-Id: <20070323093628.BC6A61C04E@coconut.itojun.org>
Date: Fri, 23 Mar 2007 18:36:28 +0900 (JST)
From: itojun@itojun.org (Jun-ichiro itojun Hagino)
X-Spambayes-Classification: ham; 0.11

> Hello,
> 
> Enclosed is a message exchange I had on the python-3000 list (in
> which I mistyped Python 2.0 instead of Python 3.0).
> 
> I'd like to see the IPv6 address tuple in Python 3000 aka Python 3.0
> changed to be (adddress, port) so that socket code written for
> IPv4 would work with it. The 4-tuple with additional flowinfo and scopeid
> elements in my opinion isn't worth it compared to the hardship caused.
> Another approach could be to allow both 2-tuple and 4-tuple, and
> always generate a 2-tuple for global addresses without flowinfo...
> What is your take on this proposal?

	i understand your concern.  current 4-tuple is direct mapping from
	struct sockaddr_in6 (which has those two additional members).

	scopeid element is very important and we cannot just drop it.
	IPv6 includes "scoped" addresses, i.e. you may see the same address
	(fe80::1) on two of your ethernets, and they are different entity,
	therefore you need to disambiguate with scopeid (usually interface
	name).
	luckily there is a way to handle address+scopeid as single string -
	see RFC4007 section 11.  this way, we can remove scopeid field.

	noone uses flowinfo at this point of time (it is for QoS - diffserv,
	intserv, and that kind of stuff) but i'm not sure about future use.
	please discuss it with python guys (cc: to me if needed).

	as for actual code change, i'm willing to help but i got my right hand
	injured so i can't work on anything seriously for a couple of weeks.
	please keep me posted.

itojun
]