nemo@ietf.org
[Top] [All Lists]

Re: [nemo] About Test Specification in IPv6 Ready Logo

Subject: Re: [nemo] About Test Specification in IPv6 Ready Logo
From: "K.Kawaguchi"
Date: Tue, 28 Nov 2006 11:17:10 +0900
Hi,

""Pascal Thubert (pthubert)" <pthubert@xxxxxxxxx>" wrote:
> >> Your example is an extended Home Network case, and you have used a
> Home
> >> Address from the prefix on the Home Link. In that case, the HA
> expects
> >> that the MR is at Home when there is not binding, and it will deliver
> >> over the Home Link the packets routed via MR's HoA A:B:C:0::i.
> >
> >My example is an extended Home Network case. But, Home Address from
> >Mobile Network Prefix (5.3 Home Address from MNP in
> draft-ietf-nemo-home
> >-network-models-06).
> 
> [Pascal] OK. That was not obvious from the picture since you did not
> provide the HoA :)
> 
> As a reminder, we do not recommend the configuration that you are
> proposing.
> "
>    There are a number of issues with returning home when a mobile router
>    configures its home address from the MNP as described in Section 5.3.
>    Therefore we do not recommend this mechanism if the mobile routers
>    attach to the home network.
> "
> 
> If you still want to, you might for instance configure statically a
> connected route to the Home aggregation via the Home Link. At this point
> you are more or less back to the aggregated mode. But then again, why
> would you do that? Extended mode is meant to take the Home address from
> the /64 on the Home Link, thus my assumption on A:B:C:0::i.

OK. I find this method (Extend mode and Home address from MNP) not having
any advantage.


> 
> >
> >
> >>
> >> As Keiichi says there are 2 case.
> >>
> >> Implicit:
> >>
> >> HA knows A:B:C:i::/64 via A:B:C:0::i; if this is a static information
> >> (static or automatic route) then the HA keeps that route regardless
> of
> >> whether the MR is bound. The HA can share that information with other
> >> GWs on the Home Link using an IGP over the Home Link, but to keep it
> >> simple just assume that the HA is also the default GW in the Home
> Link.
> >>
> >> So if MR1 is at Home, the HA can still reach any LFN behind it
> because
> >> it has a static information for the route A:B:C:i::/64 via A:B:C:0::i
> >> and it expects A:B:C:0::i over the Home Link. If another MR at home
> >> needs to reach the LFN, packets will first reach the HA (default GW),
> >> and the HA will issue an ICMP redirect. MRs could also expose their
> >> prefix on the Home Link using RFC 4191 to save that flow.
> >>
> >> So MRs do not need to participate to the IGP on the Home Link, and
> that
> >> can be a benefit in a very large or very dynamic Home configuration
> >
> >I understand this case (thanks you).
> >
> >On the other hand, in the case Home Address from MNP.
> >Does MR need to join in IGP after configuring the address at the home?
> 
> [Pascal] As I said, you can use a static route to the Home aggregation
> over the Home interface. One more static route... Obviously you could
> mix and match static and dynamic routing, but if you are already using
> static routes, I'd go for one more static. But then again, I would not
> use HoA from MNP in extended mode if MRs need to go back home.

I understand.


> 
> >
> >>
> >>
> >> Explicit:
> >>
> >> The route in the HA is associated to the binding. When the MR comes
> back
> >> Home, the route is lost and the MR needs to participate to whatever
> IGP
> >> is run at Home. The choice of the IGP is a configuration issue, it
> can
> >> be any of the usual suspects (OSPF, RIP, EIGRP, ISIS, you name it).
> It
> >> could even be a MANET :) The choice of the IGP and how you deploy it
> >> will impact the capability for your Home Network to handle/survive a
> >> more or less high rate of changes (routers in/out)
> >>
> >> What NEMO adds: NEMO requires that the MR presents itself as a router
> >> and participates to the IGP only if it is at Home. So either you have
> a
> >> dedicated interface for going Home or you have some dynamics in the
> >> behavior of the roaming interface(s) that can reach Home to switch
> >> between at-Home and Roaming profiles.
> >
> >Ok, I understand.
> >This is the same also in the case from MNP, isn't it?
> 
> [Pascal] Yes... if I understand you well. More words to make sure:
> 
> The MR at Home autoconfigures a Link Local Address for the Home Link,
> and exposes the MNP route with the IGP using that LLA. This happens
> regardless of whether the HoA was derived from the prefix on the Home
> Link or from the MNP. In all cases, the HA obtains a route towards the
> MNP via the MR LLA.
> 
> Makes sense?

OK, I understand.
Hearty thanks for your e-mails.


Best regards
---
Kiyoaki KAWAGUCHI


> 
> Pascal
> 
> >
> >Best regards
> >---
> >Kiyoaki KAWAGUCHI
> >
> >
> >
> >>
> >> It can be expected that routing within a nested NEMO (MANEMO) will
> >> somewhat alleviate that restriction.
> >>
> >> Pascal
> >>
> >> >-----Original Message-----
> >> >From: K.Kawaguchi [mailto:kawaguti@xxxxxxxxxxxx]
> >> >Sent: Monday, November 27, 2006 2:49 AM
> >> >To: keiichi@xxxxxxxxxx; nemo@xxxxxxxx
> >> >Subject: Re: [nemo] About Test Specification in IPv6 Ready Logo
> >> >
> >> >Hi,
> >> >
> >> >"Keiichi SHIMA <keiichi@xxxxxxxxxx>" wrote:
> >> >> On 2006/11/25, at 13:29, Keiichi SHIMA wrote:
> >> >>
> >> >> >>> So, even in
> >> >> >>> the case 2, we can put a routing entry for the mobile network
> >> prefix
> >> >> >>> by not using any routing protocol.
> >> >> >>
> >> >> >> Please teach the method of not using routing protocol.
> >> >> >> Is there draft or RFC ?
> >> >> >
> >> >> > Since a mobile node knows its mobile network prefix, it can
> install
> >> >> > a routing entry for it after it receives a binding ack message.
> >> >> > The home agent of the mobile node will know the mobile network
> >> >> > prefix stored in a binding update message from the mobile node,
> it
> >> >> > can also install a routing entry when it receives the binding
> >> >> > update message.
> >> >>
> >> >> Some more minor additional notes...
> >> >>
> >> >> The above example is for the explicit mode.  And if we use
> implicit
> >> >> mode, then these two entities already know what to do when
> >> >> registration completes.  So either using a dynamic routing or not
> is
> >> >> just a configuration issue for route management and it has nothing
> to
> >> >> do with the network model.
> >> >>
> >> >> # if I'm not missing something.
> >> >
> >> >I still have my uncertain point.
> >> >Please look at the following figures.
> >> >
> >> >                   |
> >> >                   HA1
> >> >                   |
> >> >  -----+-----+-----+-----+----- Home Link: A:B:C:0::/64
> >> >       |     |           |
> >> >       |     |           | MR1-egress
> >> >       H     R(MR)       MR1
> >> >             |           | MR1-ingress (Home Address)
> >> >                         |
> >> >                        -+----- Mobile Network: A:B:C:i::/64
> >> >
> >> >
> >> >I agree as you say, HA and MR can install own routing table entry by
> >> >binding message. However, how do you tell it to other nodes on home
> >> >network?
> >> >How do you do when MR1 moves from the home link and the binding
> message
> >> >is completed? Also, how do you do when MR1 returns to the home link?
> >> >
> >> >
> >> >Best regards
> >> >---
> >> >Kiyoaki KAWAGUCHI
> >>
> >>
> 
> 


<Prev in Thread] Current Thread [Next in Thread>