[manet] Re: MANET Internetworking

Don, an important update - AERO/OMNI have been updated to use the Segment Routing Header (SRH) per RFC8754 which enables Internetworking for MANETs that may include internal clustered partitions. This allows for navigation of clusters nested to arbitrary depths within the local and remote MANETs yet still be able to use an Internetwork as transit.

This is clustering without aggregation; each cluster can operate based on /128 "host" routes as long as at least one cluster head node publishes its address to the next-higher cluster on the path to the Internetwork. Therefore, each cluster need only discover the host routes of other cluster members which may be far fewer than the number of host routes present in the entire MANET. This will especially be helpful for proactive routing protocols like OSPF-MDR and OLSRv2 that do best when there are fewer numbers of routes.

During the IETF122 session meeting, there was a preliminary discussion of wg charter updates. Can we have continued charter update discussions to consider MANET Internetworking solutions?

Thank you - Fred Templin

From: Templin (US), Fred L <[email protected]>
Sent: Thursday, March 20, 2025 9:17 AM
To: Don Fedyk <[email protected]>; [email protected]
Subject: [manet] Re: MANET Internetworking

Hi Don,

>> Have you done a comparison with LISP? There seems to be a lot of similarity when you look at IPv6 LISP carrying private >> IPv6 Addresses.  A comparison of the approaches might illustrate where you have value versus LISP.

It is a very good question. What has eventually become AERO/OMNI as we now know
it got its start in Y2K with the publication of:

https://www.ietf.org/archive/id/draft-templin-ngtrans-v6v4compat-00.txt

This was an early examination of automatic tunneling, with concepts still observed by
AERO/OMNI today. At that time, the MANET working group was focused on developing
efficient proactive and reactive routing protocols for mobile ad-hoc networking with
intense competition between OLSR / TBRPF and AODV / DSR. As a contributor to
TBRPF, I brought the automatic tunneling concept to the attention of MANET but the
Internetworking problem space was set to the side to maintain focus on the routing
protocols. The addressing question got kicked to a side group called Autoconf that
produced a single publication in RFC5889 before the group closed and declared
success. My efforts on MANET Internetworking continued, however, and have
evolved to the current day to the point that I was able to give the presentation you
heard on Tuesday.

About LISP, development of AERO/OMNI and LISP have progressed in parallel for
decades but with very little synchronization between the two efforts. LISP has
enjoyed an active following whereas AERO/OMNI has largely been all based on
my own efforts with occasional "assists" from anonymous supporters - it is
something of a David vs. Goliath story. But, I have carried the MANET principles
with me since the beginning of this millennium, and I can tell you that they have
guided the development of AERO/OMNI to the point that I believe MANET
Internetworking is now a solved problem.

I can't speak authoritatively about LISP, but what I can say about AERO/OMNI is
that they address the "6M's" of modern Global Mobile Internetworking including:


  1.  Multilink
  2.  Multihop
  3.  Multinet
  4.  Mobility
  5.  Multicast
  6.  (Performance) Maximization

The way AERO/OMNI is able to do this is by establishing a virtual interface
abstraction based on IPv6 encapsulation to create a new layer in the architecture
known as the adaptation layer. LISP does not do that. Addressing within a local
MANET is therefore based on adaptation layer addresses, and the OMNI
virtual interface supports global Internetworking for MANET nodes by
encapsulating IP packets with global addresses over the MANET adaptation
layer routing service. This allows for multihopping within the local MANET as
well as Internetwork traversal for contacting correspondent nodes on external
networks. All of the other "6 M's" follow naturally from the adaptation layer
abstraction.

So, I would ask the group to consider this not as a random topic hoping to
find a working group home but rather as the culmination of multiple decades
of R&D that surveyed the entire mobile Internetworking problem space and
has finally returned to its point of origin.

Thank you for your consideration,

Fred

From: Don Fedyk <[email protected]<mailto:[email protected]>>
Sent: Wednesday, March 19, 2025 7:59 PM
To: Templin (US), Fred L <[email protected]<mailto:[email protected]>>; [email protected]<mailto:[email protected]>
Subject: Re: MANET Internetworking

Thanks Fred

Have you done a comparison with LISP? There seems to be a lot of similarity when you look at IPv6 LISP carrying private IPv6 Addresses.  A comparison of the approaches might illustrate where you have value versus LISP.

Thanks
Don




________________________________
From: Templin (US), Fred L
Sent: Wednesday, March 19, 2025 1:23 PM
To: [email protected]<mailto:[email protected]>
Subject: [manet] MANET Internetworking


Hi, during my talk at the Tuesday MANET session there were questions about my charts where I showed two disjoint MANETs using a wide area Internetwork (e.g., the global Internet) as transit (charts here):



https://datatracker.ietf.org/meeting/122/materials/slides-122-manet-manet-internetworking-01



One question asked why not just connect up the two MANETs with a BGP or OSPF router and have MANET1 <-> MANET2 communications go directly across the connecting router without having to transit the wide aera Internetwork?



The answer is that MANETs typically operate under de-aggregated address spaces where most/all routing information exchanged is for IPv6 /128 or IPv4 /32 prefixes ("host routes") with no fixed topology reference points. So, there would be no way for a connecting BGP/OSPF router to represent MANET1 as a small set of aggregated IP prefixes to MANET2 and vice-versa without redistributing all host routes between the two MANETs. And, redistributing host routes would cause the MANET routing protocols (e.g., OSPF-MDR, OLSRv2, etc.) to have to carry excessive routing protocol information over potentially performance-limited radio frequency links.



Instead, MANET aggregation is through intermittent or continuous connectivity through Internetwork Proxy/Servers that represent each MANET as an aggregated IP prefix to the rest of the Internetwork. MANET routers can further register end user networks with Internetwork Proxy/Servers based on Provider-Independent IP prefixes delegated to them by a mobility service that supports full Distributed Mobility Management (DMM).



Another comment was made about past efforts in the Autoconf intarea working group. We have already been there and declared success through the publication of RFC5889 where the "host route" IP addressing model for Ad Hoc Networks is articulated. This publication already provides sufficient guidance that we do not need to go back and re-open that group.



Thank you - Fred



https://datatracker.ietf.org/doc/draft-templin-6man-mla/

https://datatracker.ietf.org/doc/draft-templin-6man-aero3/

https://datatracker.ietf.org/doc/draft-templin-6man-omni3/

_______________________________________________
manet mailing list -- [email protected]
To unsubscribe send an email to [email protected]