Suppose a DNA host is in a NETLMM network. It seems to me
like it would
remove the old information and use the new router if it did
not receive a
response from its old default router, right?
jak
>
>OK, that's what happens *now*. I'm asking what
*should*
>happen, given that we want to reduce handover latency in
order
>to support real time services.
>DNA is supposed to define what *should* happen for
reducing
>handover latency.
>
Ah! i did not forget NETLMM in my last response 
With DNA, you would remove the old information (Access
router, prefix
etc.)
as soon as you determine that you have moved links (RS/RA
exchange).
With NETLMM, you won't remove the old information and hence
the issues that we discussed.
-mohan
> jak
>
>----- Original Message -----
>From: <Mohan.Parthasarathy nokia.com>
>To: <kempfdocomolabs-usa.com>; <vidyanqualcomm.com>;
><gerardo.giarettatelecomitalia.it>;
<chvogttm.uka.de>
>Cc: <netlmmngnet.it>
>Sent: Tuesday, March 14, 2006 12:28 PM
>Subject: RE: Default Router in NETLMM (was RE: [netlmm]
New
>I-D: Security Threats to NETLMM)
>
>
>
>
>>Mohan,
>>
>>Forget about NETLMM for a minute.
>>
>>Suppose you've got a DNA host. It moves to a new
AP, which triggers
>>multicast RS to All_Routers_Multicast. The host
receives an RA back
>>from a router that is not its current default
router, but the subnet
>>information is the same. It receives no RA from its
current default
>>router.
>>Should the
>>host:
>>
>Receiving no RA from its current default router will not
trigger
>(a) and (b). Receving RA from a different router will
make it
>add a new router in its list of routers that it knows.
>
>>a) Start NUD using NS/NA according to RFC
2641/2641bis (expected
>>latency: up to seconds)
>
>This will happen because it is not receiving any
response for
>the packets it is sending via the old access router.
>VRRP for IPv6 claims that it will take upto 38 seconds
before
>it switches to another default router.
>
>>b) Change the default router to be the router that
sent the RA
>>(expected
>>latency: whatever it takes the OS to update the
host's routing table,
>>perhaps less than a ms)
>>
>The new default router will also be added. NUD will
>eventually remove the NCE for the old router. But the
old
>router itself is not removed till the router lifetime
expires.
>The default router selection in 2461 says that you can
select
>a router as long as its NCE is not in INCOMPLETE state.
After
>NUD has removed the NCE, the next packet could
potentially
>still select the old router but subsequent packets
won't
>because it will be in INCOMPLETE state.
>
>This is my understanding of 2461 that could be rusty.
>
>-mohan
>
>> jak
>>
>>----- Original Message -----
>>From: <Mohan.Parthasarathynokia.com>
>>To: <kempfdocomolabs-usa.com>; <vidyanqualcomm.com>;
>><gerardo.giarettatelecomitalia.it>;
<chvogttm.uka.de>
>>Cc: <netlmmngnet.it>
>>Sent: Tuesday, March 14, 2006 10:04 AM
>>Subject: RE: Default Router in NETLMM (was RE:
[netlmm] New
>>I-D: Security Threats to NETLMM)
>>
>>
>>
>>This is not what i read from this version..
>>
>>http://www.ctie.monash.edu.au/dna-dt/draft
-ietf-dna-protocol-pre00.txt
>>
>>From section 5.2.4, you can see that the RS is sent
to the
>all routers
>>multicast address. In section 5.2.5, it talks about
how you determine
>>that you have not moved links. In section 5.2.6.2,
you can see what
>>happens if you have not moved links.
>>
>>The DNA protocol document only talks about unicast
RAs from
>the router
>>AFAIK. It does not talk about unicast RS.
>>
>>-mohan
>>
>>>-----Original Message-----
>>>From: ext James Kempf [mailto:kempfdocomolabs-usa.com]
>>>Sent: Tuesday, March 14, 2006 9:46 AM
>>>To: Parthasarathy Mohan (Nokia-NET/MtView);
vidyanqualcomm.com;
>>>gerardo.giarettatelecomitalia.it;
>>>chvogttm.uka.de
>>>Cc: netlmmngnet.it
>>>Subject: Re: Default Router in NETLMM (was RE:
[netlmm] New
>>>I-D: Security Threats to NETLMM)
>>>
>>>The RS/RA confirms reachability with the router
that send the
>>RA. Even
>>>if NETLMM is not in use, there is no guarantee
that you have
>>>reachability with the old default router (it is
possible, for
>>example,
>>>that it may have been taken offline) in the
absence of an RA
>>response.
>>>So the MN is better off changing to the new
router if it wants to
>>>ensure reachability regardless of whether NETLMM
is being used.
>>>
>>> jak
>>>
>>>----- Original Message -----
>>>From: <Mohan.Parthasarathynokia.com>
>>>To: <kempfdocomolabs-usa.com>;
<vidyanqualcomm.com>;
>>><gerardo.giarettatelecomitalia.it>;
<chvogttm.uka.de>
>>>Cc: <netlmmngnet.it>
>>>Sent: Tuesday, March 14, 2006 9:24 AM
>>>Subject: RE: Default Router in NETLMM (was RE:
[netlmm] New
>>>I-D: Security Threats to NETLMM)
>>>
>>>
>>>
>>>>
>>>>No, that is exactly right.
>>>>
>>>>The only issue is the details. NUD is
performed using NS/NA and
>>>>typically has a latency of seconds,
according to RFC 2164.
>>>>DNA is performed using RS/RA, triggered by
an L2 event, and
>>>the latency
>>>>is typically only as much as the RTT between
the MN and AR, plus a
>>>>minimal amount of router processing time
(except in cases of
>>>>congestion).
>>>>
>>>Even if you do RS/RA exchange, why would the old
router
>>information be
>>>removed ?
>>>
>>>-mohan
>>>
>>>> jak
>>>>
>>>>----- Original Message -----
>>>>From: <Mohan.Parthasarathynokia.com>
>>>>To: <kempfdocomolabs-usa.com>;
<vidyanqualcomm.com>;
>>>><gerardo.giarettatelecomitalia.it>;
<chvogttm.uka.de>
>>>>Cc: <netlmmngnet.it>
>>>>Sent: Tuesday, March 14, 2006 9:17 AM
>>>>Subject: RE: Default Router in NETLMM (was
RE: [netlmm] New
>>>>I-D: Security Threats to NETLMM)
>>>>
>>>>
>>>>Ok, i will read it.
>>>>
>>>>In NETLMM, when you switch access routers,
DNA will conclude
>>that you
>>>>have not switched links and all the
information that you
>gathered on
>>>>the old link will be left as it is i.e the
old information
>>will still
>>>>be kept on the hosts. So, i must be missing
something..
>>>>
>>>>-mohan
>>>>
>>>>
>>>>>-----Original Message-----
>>>>>From: ext James Kempf [mailto:kempfdocomolabs-usa.com]
>>>>>Sent: Tuesday, March 14, 2006 8:48 AM
>>>>>To: Parthasarathy Mohan
(Nokia-NET/MtView); vidyanqualcomm.com;
>>>>>gerardo.giarettatelecomitalia.it;
>>>>>chvogttm.uka.de
>>>>>Cc: netlmmngnet.it
>>>>>Subject: Re: Default Router in NETLMM
(was RE: [netlmm] New
>>>>>I-D: Security Threats to NETLMM)
>>>>>
>>>>>Please take a look at Julian and
Sathya's draft on the MN-AR
>>>>interface.
>>>>>They describe how DNA can be used.
Neither NUD nor fixing the MAC
>>>>>address across routers is necessary if
the node uses DNA and
>>>switches
>>>>>default routers when it receives an RA
having a new MAC
>>>address in it.
>>>>>
>>>>> jak
>>>>>
>>>>>----- Original Message -----
>>>>>From: <Mohan.Parthasarathynokia.com>
>>>>>To: <vidyanqualcomm.com>;
<gerardo.giarettatelecomitalia.it>;
>>>>><chvogttm.uka.de>
>>>>>Cc: <netlmmngnet.it>; <kempfdocomolabs-usa.com>
>>>>>Sent: Monday, March 13, 2006 3:25 PM
>>>>>Subject: RE: Default Router in NETLMM
(was RE: [netlmm] New
>>>>>I-D: Security Threats to NETLMM)
>>>>>
>>>>>
>>>>>Vidya,
>>>>>
>>>>>When you handover and change the default
router, first NUD
>>>>has to kick
>>>>>in detect that the old default router is
not reachable
>>>>anymore. But the
>>>>>old router itself will not be removed
until you stop hearing RAs
>>>>>(depends on router lifetime) from the
router. So, till you
>>>completely
>>>>>remove the old AR from the system, it
can be disruptive. By just
>>>>>maintaining one virtual default router
(IP and MAC address)
>>>>across all
>>>>>links, you don't have to worry about
this part. But then if
>>>>you have to
>>>>>support multiple MAPs, then there are
some additional
>complexities..
>>>>>
>>>>>-mohan
>>>>>
>>>>>
>>>>>
>>>>>________________________________
>>>>>
>>>>>From: netlmm-adminngnet.it
[mailto:netlmm-adminngnet.it] On
>>>>Behalf Of
>>>>>ext Narayanan, Vidya
>>>>>Sent: Monday, March 13, 2006 1:57 PM
>>>>>To: Giaretta Gerardo; Christian Vogt
>>>>>Cc: netlmmngnet.it; James Kempf
>>>>>Subject: Default Router in NETLMM (was
RE: [netlmm] New I-D:
>>>>>Security Threats to NETLMM)
>>>>>
>>>>>
>>>>>Hi Gerardo,
>>>>>
>>>>>This is something we are currenlty
discussing in the Design
>>>>Team, even
>>>>>though it is part of the MN-AR
discussion. Since in a NETLMM
>>>>domain the
>>>>>MN does not change the IP address, the
NETLMM domain may be
>>>seen as a
>>>>>single subnet and therefore the default
router seen by the
>>>MN may not
>>>>>change. But, it really depends on the
MN-AR interface, imo: one
>>>>>possibility, as you mention, is that all
ARs in a NETLMM
>>>domain mimic
>>>>>the MAP. I still don't have a clear
opinion on this, though.
>>>>>What do you
>>>>>think?
>>>>>
>>>>>--Gerardo
>>>>>
>>>>><VN> It seems simpler and more
scalable to let the AR be
>>the default
>>>>>gateway. The important aspect is to
prevent the IP address
>>of the MN
>>>>>from changing - I see the default
gateway change as not so
>>>disruptive.
>>>>>
>>>>>Is there a reason why it would make
sense for the MAP to be
>>>>the default
>>>>>gateway?
>>>>>
>>>>>-Vidya
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>>
>>
>>
>>
>
>
>
_______________________________________________
netlmm mailing list
netlmmngnet.it
https://vesuvio.ipv6.cselt.it/mailman/listinfo/netlmm
|