Re: [Autoconf] Closing summary on consensus-call for RFC5889modifications

Alexandru Petrescu <> Thu, 26 August 2010 20:08 UTC

Return-Path: <>
Received: from localhost (localhost []) by (Postfix) with ESMTP id 765B23A6AAE for <>; Thu, 26 Aug 2010 13:08:34 -0700 (PDT)
X-Virus-Scanned: amavisd-new at
X-Spam-Flag: NO
X-Spam-Score: -1.612
X-Spam-Status: No, score=-1.612 tagged_above=-999 required=5 tests=[AWL=0.637, BAYES_00=-2.599, HELO_EQ_FR=0.35]
Received: from ([]) by localhost ( []) (amavisd-new, port 10024) with ESMTP id YrqXA8wManeo for <>; Thu, 26 Aug 2010 13:08:33 -0700 (PDT)
Received: from ( []) by (Postfix) with ESMTP id E86463A697D for <>; Thu, 26 Aug 2010 13:08:31 -0700 (PDT)
Received: from [] (unknown []) by (Postfix) with ESMTP id 5166494016B; Thu, 26 Aug 2010 22:08:57 +0200 (CEST)
Message-ID: <>
Date: Thu, 26 Aug 2010 22:08:32 +0200
From: Alexandru Petrescu <>
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 6.0; fr; rv: Gecko/20100802 Thunderbird/3.1.2
MIME-Version: 1.0
To: Teco Boot <>
References: <><ABE739C5ADAC9A41ACCC72DF366B719D035CA5CE@GLKMS2100.GREENLNK.NET> <> <ABE739C5ADAC9A41ACCC72DF366B719D03609162@GLKMS2100.GREENLNK.NET> <> <> <> <> <> <>
In-Reply-To: <>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 8bit
X-Antivirus: avast! (VPS 100826-1, 26/08/2010), Outbound message
X-Antivirus-Status: Clean
Subject: Re: [Autoconf] Closing summary on consensus-call for RFC5889modifications
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: Ad-Hoc Network Autoconfiguration WG discussion list <>
List-Unsubscribe: <>, <>
List-Archive: <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>
X-List-Received-Date: Thu, 26 Aug 2010 20:08:34 -0000

Let's try to loop this Router-Host definition towards the original post
intention, not immediately, but let's at least try.

Le 26/08/2010 20:07, Teco Boot a écrit :
> Alex,
> In my environment, host don't use TCP very often. My apps use UDP.
> Internet core routers often use BGP, which run on TCP. Let's stick
> on more commonly used differentiators. Like forwarding packets,
> sending RA or sending routing protocol packets.

If the distinguisher for Routers is "forward packets whose dst is not
self, _instead of dropping_ " the distinguisher for Hosts could be "drop
packets whose dst is not self, instead of forwarding".

In this sense, sending or receiving routing protocol packets is actually
a Host characteristic, because dst is self at reception.

Remind the original topic problem?  How do the different suggestions
Host-Router we make could affect the use of "Router" vs the use of
"Host" in the title?

I could live with the RFC title saying this is about Routers, because
people on this list seem to understand a Router to be a Host too and

Nobody seems to be ready to actually list the things a computer does
making it a Router (and not a Host).  If so, why the struggle?

If you want to list the things, let me start:

An IPv6 Router (remark, not simply "a Router") does these things which
an IPv6 Host does not:
-MUST send RA (a Host MUST NOT).
-MUST forward packets whose dst address is not assigned to any of its
  interfaces (a Host MUST drop packets whose dst is not assigned to any
  of its interfaces, and must further process and still not forward
  packets whose dst is assigned to at least one address assigned to one
  of its interfaces, _or_ part of a multicast group to which this host
  has previously subscribed).
-MUST send ICMP Redirect (a Host MUST NOT) if the need is present.
-MAY join a multicast group, in the perspective of using that group
  subscription later as a Host.
-MUST decrement the Hop Limit (a Host MUST NOT).
-MUST NOT process the IPsec headers (a Host MUST).
-there are surely others - but are you willing to continue.

Remark the sending of routing protocol messages is not part of this
list, because it's actually a Host characteristic.

And, if you find exception to any of these rules - feel free to do so,
then please come up with some means to define them better, and continue
on it.

There is much lack of discussion synthesis here.


> Teco
> Op 26 aug 2010, om 15:57 heeft Alexandru Petrescu het volgende
> geschreven:
>> Le 26/08/2010 14:31, Teco Boot a écrit :
>>> Alex,
>>> You could take some time for research, on hosts having a routing
>>>  table. Take a start with host requirements (RFC 1122): |  As an
>>>  extra feature, a host IP layer MAY implement a table |  of
>>> "static routes".
>> Ah well spot.  In practice that is actually a MUST - every IP
>> stack has a table of routes, be it ran by Hosts or by Routers.
>> There's no IP stack without routing table.
>> This is one reason why I think it's difficult to identify a Host
>> which is not a Router, nor a Router which is not a Host.
>> Probably one could call a Host a Host if it does more TCP
>> instructions during 1 second, than routing instructions.
>> Alex
>>> Teco
>>> Op 25 aug 2010, om 18:03 heeft Alexandru Petrescu het volgende
>>> geschreven:
>>>> Le 25/08/2010 15:21, Teco Boot a écrit :
>>>>> Alex,
>>>>> Your statement is not accurate. You say: "A router with
>>>>> [whatever] is a router to. Would someone doubt on that?
>>>> Right, a router is a router - always valid.
>>>> A "machine" with static routes is a router too.
>>>>> If you intended to say:
>>>>>> A node with static routes (no routing protocol messages)
>>>>>> is a router too.
>>>>> This is definitely not true. Every host may have static
>>>>> routes.
>>>> Right.  That's why I tend to accept that there are no Hosts in
>>>> this world and they're all routers, because they all execute
>>>> longest prefix match searches in their routing tables, they
>>>> all have at least two interfaces (lo is one), they all have
>>>> entries in their routing tables.
>>>> They're all routers, Hosts don't exist.
>>>>> I call a node a router if it: - may forward packets; - may
>>>>> send routing protocol packets; - may send router
>>>>> advertisements.
>>>>> Reworded: a host - may not forward packets; - may not send
>>>>> routing protocol packets; - may not send router
>>>>> advertisements.
>>>> Ah "may" makes it impossible to really distinguish.
>>>>> I have device here on my desk. It is called a Wireless-N Home
>>>>> Router. I use it as WiFi AP, Ethernet switch and DHCP server.
>>>>> I don't use it for forwarding packets, because on the yellow
>>>>> marked port it does some nasty NAPT operations, which I can't
>>>>> use in my setup. Shall I bring it back to the shop, and ask
>>>>> for a Wireless-N Home Host?
>>>> HA haha!!  I doubt shop vendor understands "Host" because s/he
>>>>  never sells Hosts to anyone!  S/he could sell Routers,
>>>> Switches, Desktops, Servers ; or it could Host your website if
>>>> you wish. But never sell you a Host.  Who sells Hosts?
>>>>> It: - may forward packets, but I disabled it; - may send
>>>>> routing protocol packets, but I disabled it; - may send
>>>>> router advertisements, but I doubt if it supports IPv6.
>>>> But that Access Point does have routing table entries, does
>>>> execute the longest prefix match algorithm, hence it's a
>>>> Router.
>>>>> By the way, if I use packet forwarding, NAPT and MAC NAT, it
>>>>>  acts as a host on the Internet port.
>>>> In a sense.  What do you mean it "acts as a Host on the
>>>> Internet port"? What does NAPT does as algorithm, data
>>>> structures, which a Router does not, on the Internet port?
>>>>> Providers can't detect it is a router, it is all hidden.
>>>>> Powerful feature, for where providers don't allow routers
>>>>> connected to their networks.
>>>> Hmm...
>>>> I think also, as you say, that it is good to distinguish it
>>>> based on sending RA or NA: if it sends RA then it's a Router,
>>>> otherwise it's a Host; but disabling RAs on a Router doesn't
>>>> make it a Host :-) - it makes it an IPv4 Router (another
>>>> Router :-)
>>>> Alex
>>>>> Teco
>>>>> Op 25 aug 2010, om 12:05 heeft Alexandru Petrescu het
>>>>> volgende geschreven:
>>>>>> Le 25/08/2010 10:41, Dearlove, Christopher (UK) a écrit :
>>>>>>> It's running the routing protocol, and not just
>>>>>>> listening to it, but engaging actively in it - sending
>>>>>>> necessary routing protocol messages. It's a router.
>>>>>> And a router doesn't necessarily have to run a dynamic
>>>>>> routing protocol.  A router with static routes (no routing
>>>>>> protocol messages) is a router too.
>>>>>> Alex
>>>>>> _______________________________________________ Autoconf
>>>>>> mailing list