Re: [Mipshop] Re: Mobile IPv6 Interop Testing Draft

Hi Keiichi,

Sorry it took me 11 months to reply. But thanks for your comment. Right 
now we are revising this draft, and the virtual home link is one topic 
to revisit. Of course, there have been various discussions about 
virtual home links since then.

In terms of configuring the home agent end to allow the right prefix 
and routing info, I would imagine a few couple possibilities that might 
work which I will try to describe.

1. The first scenario you described: a physical link and then multiple 
point-to-point tunnels. The physical link owns the prefix, but the 
local and remote addresses of the tunnel also come from that prefix. If 
necessary, a host route for the MN can be added with the tunnel 
interface as outgoing interface. rtadvd would also have to be signaled 
to send advertisements on each tunnel (haven't tried that personally; 
hopefully it's possible).

2. The physical link does, or does not exist. Either way, bridging is 
enabled between all of the tunnels and either a virtual or real link, 
which owns the prefix. The tunnel has a local and remote address on a 
different prefix (or no address, depending on tunneling used), but then 
the home address is configured from RAs that are echoed to the tunnel 
by the bridge.

3. ? There are probably more ways to do this.


On Oct 24, 2003, at 5:32 AM, Keiichi SHIMA / 島慶一 wrote:

> Hello Samita and T.J.,
>
> It is very interesting and good idea to establish a remote test
> environment. (and we are interested in participating the test, of
> course :-)
>
> By the way, I have a question (or concern) about the virtual home link
> described in Section 8 in the draft.
>
> Section 8 proposes the method for mobile node implementers to test
> 'returning to home' with other home agent vendor's implementation,
> even when the mobile node cannot access the real home link of the home
> agent.
>
> My question is, how is a home agent (and a home link) configured?
>
> I imagined the following network, for example.
>
>
>       ----+----------  home link P::/64
>           |
>           |P::1 (home agent address)
>        IF1|200.0.0.1
>        +--+-+
>        | HA |    IF3(200.0.0.1 - 100.0.0.1)
>        +--+-+
>        IF2|
>           |
>      ((Internet))
>           |
>           |
>        ---+---------+------------ visited link of MN tester Q::/64
>                     |
>                     |Q::100 (care-of address)
>                  IF1|100.0.0.1
>                  +--+-+
>                  | MN |   IF2(100.0.0.1-200.0.0.1)
>                  +----+
>
> HA has physically two interfaces, IF1 and IF2.  In this case, IF2 has
> nothing to do with this scenario, I ignore it.  HA also has one tunnel
> interface IF3 between 200.0.0.1 and 100.0.0.1.
>
> MN has physically one interface IF1, and one tunnel interface IF2
> between 100.0.0.1 and 200.0.0.1.
>
> The draft says, MN receives a router advertisement for home link from
> HA via the tunnel interface IF2.  However I think it is difficult,
> since HA must have two different interfaces and those interfaces must
> be assigned the same prefix (in this case, P::/64).  This means HA is
> required to support a kind of multi-link subnet technology.  (Or did I
> misread?)
>
> If there is no physical home link, I think it is possible to configure.
> For example,
>
>        +--+-+  IF1(200.0.0.1 - 100.0.0.1)
>        | HA |       P::1/64 (home address)
>        +--+-+
>        IF2|200.0.0.1
>           |
>      ((Internet))
>           |
>           |
>        ---+---------+------------ visited link of MN tester Q::/64
>                     |
>                     |Q::100 (care-of address)
>                  IF1|100.0.0.1
>                  +--+-+
>                  | MN |   IF2(100.0.0.1-200.0.0.1)
>                  +----+
>
> In this case, I see no problem.  However, this scenario requires as
> many home subnets as the number of mobile nodes, since HA must arrange
> different tunnel interface and home prefix for each mobile node.
>
> I don't think the former scenario is not appropriate.  It is of course
> more efficient than the latter, if HA can do that.  But I just think
> not so many home agent can do it.  I may be wrong...
>
> Best Regards,
> ---
> Keiichi SHIMA
> IIJ Research Laboratory <keiichi@xxxxxxxxx>
> KAME Project <keiichi@xxxxxxxx>
>
> _______________________________________________
> Mipshop mailing list
> Mipshop@xxxxxxxx
> https://www1.ietf.org/mailman/listinfo/mipshop