[manet] Re: MANET WG Charter and MANET Internetworking

RFC 5444, as used by OLSRv2, saves some of that overhead with its address compression that only includes any common part of the addresses usually twice (one address block, plus the router’s own address that is not compressed).

Some, not all of course.

> On 21 Jul 2025, at 18:50, Templin (US), Fred L <[email protected]> wrote:
> 
> Hi Jim,
>  
> Yes, it is generic to all “proactive” routing protocols already specified by the IETF (e.g., OLSR, OSPF-MDR, Babel, etc.) and I think also applies for “reactive” protocols (e.g., AODV, DSR, etc.) but have not thought as much about them. The proactive routing protocols have Router IDs (RIDs) based either on a statistically or administratively-assigned identifier or IP address so the length of the RID has implications for routing protocol overhead in the control plane. By inserting an Adaptation Layer mid-layer in the architecture, the service becomes generic to any radio type operating at Layer 2 and also works with multiple heterogeneous radio types combined.
>  
> So, AFAICT, there may be some extra control message overhead in carrying longer IP host routes and/or RIDs but where the Adaptation layer makes up for it is in header compression in the data plane. The generic application to all radio types is also important for multi-access systems (e.g., 5G and WiFi in commercial devices, tactical radio waveforms in tactical applications, etc.) which we think will become more and more ubiquitous for Global Mobile Internetworking.
>  
> Thank you - Fred
>  
> From: Stevens, Jim <[email protected]>
> Sent: Monday, July 21, 2025 10:20 AM
> To: Templin (US), Fred L <[email protected]>; [email protected]
> Subject: [EXTERNAL] RE: MANET WG Charter and MANET Internetworking
>  
> Fred, what you are proposing is, I believe, generic for a wide range of (if not all?) combinations of MANET wireless network use cases.
>  
> I can’t compare against the AERO/OMNI solution since I haven’t analyzed/modeled its convergence, but the reason we go with the layer 2/layer 3 separation is that we got orders of magnitude lower over-the-rf-air overhead over the range of our customer/user desired topology dynamics when we compared to running standard routing protocols like RIP & OSPF and a few MANET routing protocols like OLSR in our design study a decade ago. 
>  
> This was critical to us supporting hundreds of nodes per wireless network with unicast & multicast fast convergence with low overhead across large interconnected networks of wireless networks with scores of networks (so thousands of user IP subnets) in low speed wireless networks of 10s of Kbps to 100s of Kbps to 1s of Mbps burst rates (shared across all radios in the wireless network).
>  
> Our approach will not have similar overhead reductions for all wireless layer 2/1 networks (for example, it won’t have as large an overhead reduction with standard 802.11), but it does for the different broadcast layer 2/1 MANET wireless networks we work with.
>  
> Jim
>  
> From: Templin (US), Fred L <[email protected] <mailto:[email protected]>>
> Sent: Monday, July 21, 2025 11:44 AM
> To: Templin (US), Fred L <[email protected] <mailto:[email protected]>>; Stevens, Jim (USA) <[email protected] <mailto:[email protected]>>;[email protected] <mailto:[email protected]>
> Subject: RE: MANET WG Charter and MANET Internetworking
>  
> Important addition to what I already said: “MANET routing and addressing can and should operate as an IPv6 Adaptation Layer function that supports either IPv6 or IPv4 at Layer 3 and with multiple heterogeneous radio types at Layer 2. ” Thank
>  
> Important addition to what I already said:
>  
> “MANET routing and addressing can and should operate as an IPv6 Adaptation Layer
> function that supports either IPv6 or IPv4 at Layer 3 and with multiple heterogeneous
> radio types at Layer 2.”
>  
> Thank you - Fred
>  
> From: Templin (US), Fred L <[email protected] <mailto:[email protected]>>
> Sent: Monday, July 21, 2025 9:32 AM
> To: Stevens, Jim <[email protected] <mailto:[email protected]>>; [email protected] <mailto:[email protected]>
> Subject: [manet] Re: MANET WG Charter and MANET Internetworking
>  
> Thank you, Jim, but you seem fixed on a strict Layer2 / Layer 3 separation while the
> AERO/OMNI solution provides is an “Adaptation Layer” as a middle layer in between.
> MANET routing and addressing can and should operate as an IPv6 Adaptation Layer
> function that supports either IPv6 or IPv4 at Layer 3. Over the air reductions are
> through header compression, and IETF RFC4903 and 5889 rules are observed.
>  
> Fred
>  
> From: Stevens, Jim <[email protected] <mailto:[email protected]>>
> Sent: Monday, July 21, 2025 9:20 AM
> To: Templin (US), Fred L <[email protected] <mailto:[email protected]>>; [email protected] <mailto:[email protected]>
> Subject: RE: MANET WG Charter and MANET Internetworking
>  
> Hi Fred,
>  
> Thanks for that response.  I had forgotten about RFC 5889. 
>  
> We have some customers/users that want to maintain logical IP subnets with prefixes shorter than /32 for IPv4 and /128 for IPv6 which is different from RFC 5889.
>  
> If the MANET WG wants to maintain this limitation, then I will stop bringing up the RFC4903 issue.
> (the only caveat is that if we ever decide to propose to release our currently proprietary solution to the RFC 4903 multi-link issues as an RFC, then I reserve the right to bring up that solution at that time.) 
>  
> And, if the MANET WG wants to only support MANET routing at layer 3, then I will stop bringing up MANET routing below layer 2 also.
> (again, the only caveat is that if we ever decide to propose to release our currently proprietary solution for combining layer 2 routing with overlay layer 3 routing for significant over-the-air overhead reductions as an RFC, then I reserve the right to bring up that solution at that time.)
>  
> This will be my last email on these topics unless someone in the MANET WG brings up these topics (or the caveats above).
>  
> Jim
>  
> From: Templin (US), Fred L <[email protected] <mailto:[email protected]>>
> Sent: Monday, July 21, 2025 10:54 AM
> To: Stevens, Jim (USA) <[email protected] <mailto:[email protected]>>; [email protected] <mailto:[email protected]>
> Subject: RE: MANET WG Charter and MANET Internetworking
> ZjQcmQRYFpfptBannerEnd
> Hi Jim,
>  
> You seem to be still concerned about RFC4903 Multilink Subnet issues, but that only
> matters if you are trying to get the MANET to maintain a logical IP subnet with a prefix
> shorter than /32 for IPv4 or /128 for IPv6. But, Section 6 of RFC5889 has an Addressing
> Model for that has declared that no on-link subnet prefixes should be configured on
> a MANET interface. I am not saying that maintaining a logical IP subnet in a MANET
> can’t be done – only that that is not the way the IETF has decided to do it.
>  
> Thank you - Fred  
>  
> From: Stevens, Jim <[email protected] <mailto:[email protected]>>
> Sent: Monday, July 21, 2025 8:02 AM
> To: [email protected] <mailto:[email protected]>
> Subject: Re: MANET WG Charter and MANET Internetworking
>  
> Date: Fri, 18 Jul 2025 08:11:33 +0200
> From: Henning Rogge <[email protected] <mailto:[email protected]>>
> Subject: [manet] Re: MANET WG Charter and MANET Internetworking
> To: "Stevens, Jim" <[email protected] <mailto:[email protected]>>
>  
> Hello,
>  
> On Fri, Jul 18, 2025 at 12:12 AM Stevens, Jim
> <[email protected] <mailto:[email protected]>> wrote:
> > 
> > Some of the recent emails presume that MANET radios are IP routers and radios *only* route using IP, so that questions arise with whether you decrement TTL for each radio hop.  Similar questions arise with overlay/underlay networks and MANET wireless networking getting partitioned.  And, also multihop MANET networks that get partitioned lead to the issues raised in RFC 4903 about Multi-Link Subnet Issues.
>  
> MANET normally use only host routes (or routes to prefixes that are
> reachable through a single router), so the subnet split is not that
> important for them. It is also a good idea to NOT put a subnet on your
> MANET interface but put a /32 address on it (or just a linklocal one).
>  
> There can be different MANET use cases.  Most of the different MANET networking waveforms (with their specific MANET protocols) that I work on have one or more IP stub subnets with host computers attached to the MANET radios and multiple MANET radios that also gateway to the rest of the internet.  So the MANET wireless networks can be transit wireless networks.  (Note that these are MANET networks other than standard 802.11, LoRaWAN, etc.)
>  
> I agree with the idea of either putting linklocal IP addresses on the wireless side of the MANET radios and/or using overlay/underlay IP addressing on the baseband interface(s) of the MANET radios (independent of whether the MANET routing is done at layer 3 or layer 2).
>  
> > So the MANET IETF working group also needs to consider MANET wireless networks where the wireless MANET routing is below IP – similar to how Ethernet switches route below IP.   This eliminates the TTL issues.  And overlay/underlay networks (whether MANET radios route using IP or below IP) can allow MANET radios in one IP subnet to transmit to another radio in that MANET wireless network that is in a different IP subnet.  And overlay/underlay can be adapted to solve Multi-Link Subnet issues as well
>  
> MANET can work with layer-2 forwarding below, but using two MANET
> layers independently on top of each other can lead to all kinds of bad
> side effects.
>  
> MANET protocols can also be working on layer-2 itself if you can
> change the used address type to MAC addresses (and ignore the parts
> being able to use a prefix).
>  
> Henning Rogge
>  
> Agree, we usually run our MANET protocols on the wireless layer-2 itself and standard IP routing to hosts/external routers to the MANET radios.   
>  
> Also agree that having two MANET layers running independent of each other can lead to bad side effects.  This is reason for my comment about some wireless networks run MANET routing below IP and thus that it can be bad to include MANET functions in the IP routing above the layer-2 MANET layer, such as doing DLEP to ask for radio link information when the MANET radio radio can often do a better job of adapting to changes in the rf environment (such as changing modulation, frequency, channel access time slot allocations, etc).
>  
> But there can still be some special cases that IP routing needs to handle, such as the multi-link issues addressed in RFC 4903 that arise when the wireless networks are partitioned, especially if users wanted to have the IP subnet equivalent of a VPN above the wireless network(s).
>  
> Jim
> _______________________________________________
> manet mailing list -- [email protected]
> To unsubscribe send an email to [email protected]

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