Hi Adam

ND operation is a link operation and what it does varies across link
types. Simple as that.

That's hardly surprising. 

- RFC4861 is a form of reactive protocol that is optimized for a certain
environment, namely many nodes, each one being able to see another one
on a link, and a multicast capability. With a different constraint like
few nodes and stricter forwarding/routing plane separation you might
have ended up with a proactive protocol.

- For ND, there are small variations (like omitting address resolution
on certain P2P links) to larger ones (like inverse ND for FR/ATM).
You'll note that RFC 3122 addresses a problem that is closer to 6LoWPAN
because Frame Relay is non transitive. But we could hardly afford MARS
and the multicast infra that goes with it.

- Routing protocols are the same. They are optimized for certain
assumptions. But whereas ND operates on a link with link level
assumptions, routing protocols are optimized for network level
operations, and depending on what you are doing at the  network level at
which you are operating, you have to select OSPF or RIP or OLSR or even
BGP.

Ipv6 is not broken. It is adaptable and extensible. Routing Protocols
and ARP in IPv4 have similar causes and effects, though ND in its
variations is incredibly richer than ARP.

Pascal

>-----Original Message-----
>From: [email protected] [mailto:[email protected]] On
Behalf Of Adam Dunkels
>Sent: lundi 12 octobre 2009 16:41
>To: Carsten Bormann
>Cc: 6lowpan
>Subject: Re: [6lowpan] Fundamental concerns about 6lowpan ND
>
>Hi Carsten,
>
>these sounds like some serious architectural concerns with IPv6. Should
>these really be dealt with by an adaptation layer that defines how to
>transport IPv6 packets over a particular link layer? I am not too
>accustomed to IETF practices, but isn't there a wg specifically for the
>purpose of forwarding the IPv6 architecture (6man) where issues like
>these should be raised?
>
>Also, one specific question: how would an IPv6 host deal with an
>802.15.4 network interface if the IPv6 adaptation layer would require
>changes to the core of the IPv6 stack to function properly?
>
>Thanks,
>
>/adam
>
>Carsten Bormann wrote:
>> On Oct 12, 2009, at 14:47, Julien Abeille (jabeille) wrote:
>>
>>> the issues arised on lowpan networks as far as ND is concerned are
not
>>> huge
>>
>>
>> [WG member hat]
>>
>> Julien,
>>
>> I'd like to know more about that.
>>
>> As far as I can see, certain parts of 4861-ND just DO NOT WORK on
>> non-transitive networks.
>> It's really as simple as that.
>>
>> So you either make 6LoWPANs transitive at huge cost, or you need
>> something like 6LoWPAN-ND for those parts.
>> (Or, you simply ignore that they don't work, which you mostly can for
>> DAD; we didn't want to do that.)
>>
>> Please reread section 1, paragraph 2 of 4861 for its area of
applicability.
>>
>> Gruesse, Carsten
>>
>> PS.: re the charter:
>> Getting rid of 4861's address resolution by multicast is indeed just
an
>> optimization.
>> I happen to believe it is a good one.  We can (and should!) debate
that.
>> I would have argued to simply get rid of DAD as well, but there is
the
>> issue of counterfeiting.
>> So we made a limited extension to make DAD work, which it doesn't for
>> 4861-ND on non-transitive.
>> etc.
>>
>> _______________________________________________
>> 6lowpan mailing list
>> [email protected]
>> https://www.ietf.org/mailman/listinfo/6lowpan
>>
>
>--
>Adam Dunkels <[email protected]>, +46707731614
>http://twitter.com/adunk | http://www.sics.se/~adam/
>_______________________________________________
>6lowpan mailing list
>[email protected]
>https://www.ietf.org/mailman/listinfo/6lowpan
_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan

Reply via email to