default routes in MANETs (was: RE: [Autoconf] Configuring 'default' in Multi-Subnet MANETs)

["Templin, Fred L" <[e-mail]>]

Joe,

In retrospect, it was more of a general question for
the MANET routing protocols than anything specific to
Multi-Subnet MANETs so I'm going to change the subject
line. I agree that with the proactive routing protocols
the MANET border routers can inject a default route into
the routing protocol and it propagates through the MANET
just fine. (In fact, I now recall that we did this with
TBRPF back in the 2001-2002 timeframe as I'm sure you
have done with any proactive routing protocols you may
be using in your studies.)

However, from reading the globalv6 proposal, it seems
that the landscape may be different for the reactive
protocols. In 'draft-wakikawa-manet-globalv6-05.txt',
sections 5.3 and 6 the authors assert that special
handling is needed in the reactive protocol case to
establish a default route and direct packets through
a default gateway.

In particular, since the reactive protocols do not
maintain complete topology information at each MANET
router, globalv6 says that nodes sending packets via a
particular MANET border router must insert a routing
header that names the final destination and insert
the address of the MANET border router in the IPv6
destination address. This seems like it would require
a kernel hack to make it work.

So its clear that this discussion is aside from the
question of Multi-Subnet MANETs, but it might be a
differentiator between reactive and proactive protocols
from the standpoint of Internet attachment.

Fred
[e-mail]  

-----Original Message-----
From: Joe Macker [mailto:[e-mail]] 
Sent: Friday, March 31, 2006 10:20 AM
To: Templin, Fred L; [e-mail]
Cc: [e-mail]
Subject: RE: [Autoconf] Configuring 'default' in Multi-Subnet MANETs

Fred/All:
I know this question was addressed at Dave, but for at least proactive
routing issues should addressed with some relationship to architecture.
e.g., stub areas, totally stub areas, NSSA type, transit. So I think you
have to first say what is the deployment scenario. 

For stub or totally stub the border router(s) should proactively inject
a
default route.  This is normal operation for things like MANET-OSPF and
OLSR
types of deployments.  Other cases are more complex. Are there reasons
why
this should be different than normal intra-domain types of approaches?
I
see draft-thaler-autoconf-multisubnet-manets as a valid subcase of what
we
can do. Is there some motivation why there is a difference here in
default
route injection approach?  I guess I am thinking in terms of typical
OSPF
default route handling/deployment use cases.
  
Did you have something in mind or just curious? I think it's a good
question, but I would like to see this discussion be more generally
addressed and am presently hoping we could follow standard best
practices
where it makes sense.  I'd  also like to hear if someone has a reason
why we
should do something different.

-Joe

>-----Original Message-----
>From: Templin, Fred L [mailto:[e-mail]] 
>Sent: Thursday, March 30, 2006 7:59 PM
>To: [e-mail]
>Cc: [e-mail]
>Subject: [Autoconf] Configuring 'default' in Multi-Subnet MANETs
>
>Dave,
>
>I notice that 'draft-thaler-autoconf-multisubnet-manets'
>is silent on the subject of how MANET routers can configure 
>and use a default route. Do you have any thoughts on this?
>
>Fred
>[e-mail] 
>
>_______________________________________________
>Autoconf mailing list
>[e-mail]
>https://www1.ietf.org/mailman/listinfo/autoconf
>



_______________________________________________
Autoconf mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/autoconf