[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ipng
Subject:    Re: [Q] Routing header
From:       Thierry Ernst <thierry.ernst () inrialpes ! fr>
Date:       2001-05-31 17:56:04
[Download RAW message or body]

Michael Thomas wrote:

>  > I agree that a MN behind the MR is supposed to send a BU, but I disagree
>  > that the MR "is supposed to" do it as well.
> 
>    If this is true, what is the purpose of putting a subnet prefix
>    into the BU at all? How does the MR (ignoring MN for the moment)
>    eliminate triangular routing through its HA? Surely the purpose
>    of adding the prefix wasn't just to say that MN was reachable
>    at a range of IP addresses.
> 
>  > First, the MR is not the
>  > final destination of the packet.  Second, if you think about a naive
>  > implementation, MR would send a BU to the sender that appears in the
>  > inner header in the encapsulated packet, i.e. the CN of MN "behind"
>  > MR.   And what would contain the BU ?  A binding between the MR's home
>  > address and its COA.   This does not give much usefuk information to the
>  > CN.
> 
>    Remember it's the MR's home _prefix_, not address. That's
>    what my reference to the -13 draft was about below.

Mickael,

You might probably not read the mipv6-13 draft correctly.   The purpose
of the home_prefix in the MIPv6 spec. is for the Home Agent to defend
ALL home addresses of the MN on the home link (link local, site local,
global), not for support of Mobile Routers.    This has ab-so-lu-te-ly
nothing to do with the Mobile Router case.   If it has, may be I should
rather going to grew vegetables or follow English lessons (don't know
which one would be the best though :-)

The prefix you are refering to is therefore included by the MN to inform
its HA (not its CNs) [remember the 'H' bit that stands for 'Home
Registration" ?] about the size of the prefix, not to manage mobile
routers and networks.   In addition, whatever the mobile (host or
router), a binding is ALWAYS established between the 128-bits Home
Address of the mobile and it's 128-bits Careof Address.   The only place
were is mentioned a binding between a prefix and a 128-bits address is
my draft.  Otherwise, the contribution of my draft would be pointless. 
Why should I introduce a new sub option if there is already all we need
in the MIPv6-13 draft ? 
 
>  > Then, the issue your are describing (including two addresses in the
>  > routing header) would not happen with a naive implementation of MRs.
>  > If any proposal would advocate such use of the routing header, I think
>  > it should also describe how this should be done.
> 
>    I think the problem here is that the addition of prefix
>    to the BU is not sufficient to support mobile routers,
>    and as such it's open to interpretation what the
>    implementation should do.

IMHO, there is nothing to miss-interpret since the addition of prefix in
the BU is not in the MIPv6-13 draft.  What may be miss-interpreted by
implementation is redirection at the HA of packets intended to nodes
behind the MR (problem described in my draft - sorry, I always refer to
my draft, but I do not want to explain everything in each email and this
is clearly described there)
 
On the other hand,  I advocate that my draft may be miss-intrepreted,
but this is only version-1 and I clearly stated that my proposal does
not consider MN behind the MR (BTW, I did state this exaclty for the
reason you were describing at the beginning of this thread - what would
the CN do with 2 BUs, one from MN, and one from MR)


>    While I generally like the addition of subnet prefix
>    to BU's (mainly because it treats BU's like the routing
>    updates that they are), it may actually be better to
>    get the MIP WG to take the whole subject of MR's up
>    as working group item 

This is exactly what many people already want (to list just a few of
them Hesham Soliman, Francis Dupont, Claude Castelluccia, Glenn Morrow,
and may others).   May be I should ask for it more explictly.

> and delete it till then because
>    it may unfairly prejudice the final outcome.

And delete it ?  Sorry, I do not understand.   (may be I should really
take English lessons then :-)

There is a problem with mobile routers because this special case has
never been considered seriously (or not at all) in any of the existing
MIP WG specs.   My interpretation is that implementers should only
implement the MIP WG drafts for hosts.  Another set of specs. is
required for mobile routers.

Thierry.
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to majordomo@sunroof.eng.sun.com
--------------------------------------------------------------------

[prev in list] [next in list] [prev in thread] [next in thread]