The Boeing Company

P.O. Box 3707 Seattle WA 98124 USA fltemplin@acm.org

National Security Institute, Virginia Tech

2202 Kraft Dr. Blacksburg VA 24060 USA djj@vt.edu

I-D Internet-Draft defines a MANET as "an autonomous system of mobile nodes. The system may operate in isolation, or may have gateways to and interface with a fixed network" (such as the global public Internet). This document presents a MANET Internetworking problem statement and gap analysis.

Mobile Ad-hoc Networks (MANETs) often include mobile nodes with limited range wireless transmission media interfaces that establish links via a dynamically changing set of neighbors within operational range. Each mobile node engages a MANET routing protocol to discover links to first hop neighbors as well as multihop paths to reach other nodes beyond. MANET routers represent multihop paths as "host routes" established through either proactive or on-demand discovery. These host routes name singleton destinations based on either L2 addressing and forwarding in the case of L2-MANET or IP addressing and forwarding in the case of L3-MANET. Individual MANETs typically include modest numbers of mobile nodes (e.g., O(10**1), O(10**2), etc.) which naturally limits the number of host routes needed in the local routing system. MANETs can merge to form larger MANETs and/or partition into smaller MANETs according to dynamic network conditions such as mobility. MANETs may also have internal clusters with IP addressed cluster heads that limit the extent over which host routes propagate to reduce control message overhead. Finally, MANETs often operate autonomously until they encounter Internetwork access points of opportunity. A MANET should be regarded as a mobile network that is either disconnected from the global Internetwork or only intermittently connected to the global Internetwork with the ability to move rapidly between different Internetwork attachment points. Due to the dynamic multilink nature of MANETs, MANET routers must configure a unique Multilink Local Address (MLA) and often also use prefix delegation to obtain a unique Mobile Network Prefix (MNP) to connect nodes on their downstream-attached interfaces. Data communications between two nodes within the same MANET local routing region follow host routes using MANET-internal links. When a MANET border router establishes an Internetwork link, it can provide "Internet connection-sharing" access to the rest of the MANET as a connected "stub" network. Per , "stub networks carry traffic originating at and/or destined for internal nodes, but do not permit exogenous traffic to "transit" through the stub network". Practical applications however suggest that MANETs can act as either true stub networks (e.g., a cellphone providing an Internet connection sharing peer for a local multihop Wi-Fi region) or as "not-so-stubby" networks (e.g., Intelligent Transportation Systems where the 5G/6G "SideLink" service supports vehicle-to-vehicle (V2V) multihopping). In the former case, the cellphone acts as an IP router for a stub Wi-Fi MANET behind it and the individual Wi-Fi nodes act as dependent nodes. In the latter case, individual 5G/6G SideLink nodes can connect the stub MANETs they aggregate across not-so-stubby V2V multihop forwarding paths. MANET Internetworking must therefore be capable of accommodating all such scenarios. A widely-accepted axiom at the time of this writing suggests that there are more cellphones than people on the planet . According to Wikipedia, the world population reached 8 billion in 2022 and is expected to reach 10 billion by 2056 . Each mobile node that connects to the global public Internet can in some sense be regarded as a singleton "MANET" with the potential to connect still larger MANETs. A common concern cited for regarding a cellphone as a MANET router is battery consumption, but this not an issue when the cellphone is connected to a power source such as when used inside of a vehicle. Even in dismounted scenarios, however, having a (limited-duration) multi-hop communications capability may be deemed worthwhile and could provide a safety-of-life capability such as for emergency response, disaster relief, search and rescue, etc. Mapping services, streaming video and many other applications drain cellphone batteries, but people use them all the time. An endless variety of cellphone apps already exists, with many more released every day. For these reasons, it is expected that MANET apps for cellphones will eventually emerge and proliferate. With such a large population of candidate nodes, MANET Internetworking therefore regards the global Internet as a "network of (mobile ad-hoc) networks" with unrestricted dynamic relationships between distinct MANET local routing regions joined by a Non-Broadcast Multiple Access (NBMA) virtual overlay link manifested through encapsulation. illustrates an example of 2 distinct MANET local routing regions connected via the NBMA overlay using the Internet as transit:

While the figure depicts just 2 MANET local routing regions, many others worldwide will also want to connect to the virtual link. Since a sustained increase in both the world population and number of mobile wireless devices is certain, MANET Internetworking must therefore accommodate populations on the order of O(10**10) or more. This includes address duplication avoidance through operational assurance since statistical properties alone may be insufficient to avoid duplication in such large populations. Since each MANET itself is mobile, its border routers may continuously encounter different Internetwork access points as they roam while potentially configuring a different IP address at each access point. MANET border routers therefore must not continuously inject and withdraw mobile network prefixes into the Internetwork BGP routing system as they move since this would result in unacceptable churn. MANET routers instead maintain stable addresses in a Mobility Service Provider (MSP) overlay over the Internetwork such that the overlay addresses remain stable even if the border router's underlay address changes frequently. The overlay network BGP routing system must similarly be protected from mobility churn by maintaining address to mobility anchor point mappings in a scalable directory service such as the domain name system (DNS). Efficient path traversal across the overlay must be supported through traffic flow diversity per with differentiated services applied within each flow. This is especially important for the common case of multilink MANET nodes (such as cellphones with both 5G and Wi-Fi interfaces) where overlay flow bindings are necessary for multilink coordination. BGP scaling in the modern Internet backbone accommodates O(10**6) prefixes that rarely change. Frequent advertisement and withdrawal of MLAs/MNPs can cause BGP churn resulting in instability in the global routing system such that BGP alone is not well suited to supporting mobile Internetworking by itself. With an overlay, the BGP routing service only needs to cover the addresses of Mobility Anchor Points (MAPs) which are stable nodes not subject to mobility. Within the overlay, the MAPs keep track of MLA/MNP to MAP associations which may change dynamically. This is accommodated through a combination of a hierarchical overlay BGP arrangement of the MAPs in combination with dynamic updates to the (reverse) DNS resource records that associate MLAs/MNPs with MAPs. Finally, the architecture is not limited to only one gigantic global overlay; there could be many smaller overlays operated by MSPs. "IPv6 Wireless Access in Vehicular Environments (IPWAVE): Problem Statement and Use Cases" provides a complimentary analysis but does not explore MANET-specific idiosyncrasies. This document presents a MANET Internetworking problem statement and gap analysis.

The following terms are defined within the scope of this document: a MANET router that connects a mobile network to a multilink Internetworking service via Proxy/Servers in a Non-Broadcast, Multiple Access (NBMA) overlay. an overlay multilink network service node that runs an interdomain routing protocol (e.g., BGP) and directory service (e.g., DNS) to track Client-to-MAP associations. Gateways furthermore join multiple Internetworking segments in an overlay multilink virtual bridging service to form larger Internetworks. a more stable and wide-area terrestrial, non-terrestrial or hybrid network that can serve as transit to interconnect disjoint (or partitioned) MANET local routing regions. The global public Internet is an example, as are private operator service networks either individually or in concatenations with other service networks. an IEEE EUI-48 or EUI-64 address, or an L2 address format specific to a custom data link specification. the same as defined in ; often includes mobile nodes with limited range wireless transmission media interfaces that establish links via a dynamically changing set of neighbors within operational range and engage in a MANET local routing protocol. The MANET local routing protocol can be configured to operate as either an L3 (IP-based) or L2 (MAC address-based) service according to operational requirements. In both cases, each MANET can dynamically partition into multiple smaller MANETs or merge with others to form larger MANETs. This may present a challenge to the classical IP "on-link" subnet model. a MANET router that also has a continuous or intermittent interface connection to a transit Internetwork. a MANET router that joins multiple smaller MANET local routing regions to form a single larger local routing region. Each smaller region is seen as a cluster within the larger region. a node's (typically wireless) limited range transmission media interface with indeterminant connectivity properties. a node that runs a routing protocol over one or more MANET interfaces to establish multihop forwarding paths within a local routing region. Forwarding nodes in both L2- and L3-MANETs are considered "routers" even when forwarding is based on MAC addresses. All MANET routers are also considered MANET nodes, but not all MANET nodes necessarily act as MANET routers. a Proxy/Server that also provides mobility, address/prefix autoconfiguration and address resolution services to Clients. The MAP also runs an interdomain routing protocol (e.g., BGP) to announce its Client associations to Gateways. All (reasonably) stable Proxy/Servers are eligible to serve as MAPs as part of a Distributed Mobility Management (DMM) service. a longer IP prefix delegated from an MSP (e.g., 2001:db8:1000:2000::/56, 2002:192.0.2.8::/46, etc.) and assigned to a Client. Clients receive MNPs from MAP Proxy/Servers and sub-delegate them to routers in downstream networks. an aggregated IP GUA prefix (e.g., 2001:db8::/32, 2002:192.0.2.0::/40, etc.) assigned to a mobility service provider's overlay and from which more-specific MNPs are delegated. an IPv6 address based on a cryptographic hash of the node's public key regarded as its full identify. The MLA is regarded as a unique identity tag that, when assigned to an interface, becomes a valid IPv6 address with multilink-local scope. The MLA must be assured globally unique and routable at least within the overlay limited domain to support MANET Internetworking. A candidate MLA type appears in . an overlay multilink network service node in an Internetwork that provides proxy forwarding services to MANET border routers and other MANET router Clients.

MANETs have an important role in emergency response communications, disaster relief situations, communications in remote and rural areas, military operations, vehicular and swarm communications, and low-powered Internet of things (IoT) applications. MANETs provide the ability to establish and maintain communications when infrastructure- based networks, such as 5G cellular communication systems, are not accessible. As described above, MANETs may also provide Internet connectivity to internal nodes, for example, as a "stub" network via MANET routers which possess an Internetworking capability and an external connection to a radio access network. Example use cases of such MANETs include the following: Disaster Relief: Disaster situations may compromise network infrastructure, such as through the loss of base stations in a cellular radio access network (RAN). In this scenario, MANET networks can play a role in closing coverage gaps through multi-hop routing to nodes within the coverage area of uncompromised base stations. This use case is broadly applicable to any situation in which nodes are operating outside or at the periphery of RAN coverage. Tracking and Monitoring: Another example use case is the tracking and monitoring of data from low-cost low-power IoT devices ("tags") which may be placed on packages during shipment or storage. Such devices may transition in and out of coverage of infrastructure-based networks, often being located in environments that are not conducive to RF propagation (e.g., shipping container, warehouse, etc.). The ability to discover and connect to neighboring MANET-enabled devices and to establish Internet connectivity through such MANETs, enables real-time logistics and inventory data to be collected opportunistically. UAV Swarms: local communications within swarms for coordination and cooperation is a good use case for MANET networks due to the highly mobile dynamic nature of such networks. Yet swarms may also benefit from connectivity to the Internet, or other external networks. And in large swarm-based MANETs, routing of traffic through infrastructure networks to MANET endpoints, rather than traversing the entire MANET can improve communications throughput and reliability.

Under mobility conditions, distinct UAV swarms defined by MANET local routing regions will encounter situations where the local regions enter into communications range of each other. In this case, it is desirable to establish cluster heads between these regions and to propagation host routes over them as new interconnections are available and discovered. Moreover, UAV swarms for which internetworking will be persistent should be able to perform local region merger. In this case, internetworking protocols must support seamless merger of the MANET local routing regions into a larger region. Conversely, nodes or collections of nodes which leave coverage of the local region should be capable of establishing and operating an independent local region at a future time.

MANET routing for nodes operating in a common local routing region can be configured to operate as either a layer-3 (L3-MANET) or layer-2 (L2-MANET) service. In L3-MANET, routing and addressing occurs at the IP layer and each MANET forwarding hop decrements the IP TTL/Hop-Limit. IP address uniqueness assurance is essential to avoid conflicts, while MAC address collisions may not pose significant operational issues. The MANET interface observes a partially-connected multi-access link with some nodes as single-hop IP neighbors and others reachable only over multi-hop paths. IP multicast transmissions only reach single-hop neighbors which must engage an IP layer service such as Simplified Multicast Forwarding (SMF) to flood multicast messages to all MANET nodes. IP header compression state must be maintained at each MANET forwarding hop and may become invalidated due to path changes. In L2-MANET, routing and addressing occurs at the MAC layer and the IP TTL/Hop-Limit is not decremented. Both IP address and MAC address uniqueness assurance are essential to avoid conflicts. The MANET interface observes a fully-connected multi-access link with all nodes appearing as single-hop IP neighbors even though some may be separated by multiple L2 hops. IP multicast transmissions are propagated to all MANET nodes even though the L2 service may be required to forward at the layer below IP. IP header compression state is only maintained at the MANET ingress and egress nodes and is therefore not invalidated due to interior MANET path changes. Instead, the L2 header must carry additional overhead (including a TTL, packet identification and ancillary addresses) to support forwarding at a layer below IP. Since L3-MANETs propagate IP host routes, multicast-based address resolution services are unnecessary thereby reducing dependence on MANET multicast. By default, L2-MANETs require multicast-based address resolution services which can be minimized through proactive distribution of address resolution information. In both the L2- and L3-MANET models, MANETs can partition into smaller separate MANETs or merge to form larger MANETs. This presents a challenge to the classical IP subnet model where all nodes within the subnet are presumed to be reachable as single-hop neighbors connected to the same (shared) link. All MANET routers should therefore only configure the MLA and MSP prefixes as on-link prefixes on an overlay interface. This causes the node to invoke address resolution for both MANET-local nodes and target peers in other networks reached via the overlay. MANET Internetworking concepts apply equally to both the L3-MANET and L2-MANET models as discussed throughout the document.

MANETs present a unique set of Internetworking challenges not addressed by earlier works . The following sub-sections provide MANET Internetworking problem statements for aspects not (fully) satisfied by existing services that must be addressed by any solution proposals.

MANET Internetworking observes the IP addressing model in ad hoc networks . Each MANET router requires a unique IP address for MANET local communications and a unique router ID for participation in the local routing protocol. The address is termed the Multilink Local Address (MLA) configured from a shared IPv6 prefix from which each node within the MANET internetworking overlay must configure a unique MLA. For MANETs that are only intermittently connected to an Internetwork, the MLA must be generated from a prefix of scope greater than link-local but not associated with any infrastructure aggregation points. For all MANET types, each MLA and router ID must be locally-unique within the (limited) MANET local routing region. For not-so-stubby MANETs, each MLA must also be globally-unique among all MANET local routing regions worldwide. The locally-unique property ensures that no two nodes that participate in the routing protocol within the same MANET local routing region configure the same MLA and/or router ID. The globally-unique property for MLAs may seem moot until one considers that a first MANET can merge with other MANETs, and nodes from a first MANET can freely move to other MANETs. This may allow a node from a first MANET where there are no duplicates to interact with other MANETs where a duplicate address may be encountered resulting in unpredictable behavior and/or communication failures. Although the node population for each MANET local routing region is likely to be modest, the total population of MANET routers that may join the MANET Internetworking overlay may be on the order of the number of worldwide mobile connections (see: ). Assuming O(10**10) wireless connections, if MANET routers assigned random addresses from a 64-bit space, the probability of one or more collisions within the total world population (i.e., when multiple nodes independently configure the same address) exceeds 98% . With such a high likelihood of duplication in the worldwide population, unresolvable collisions could disrupt communications. When MANET Internetworking is applied to connect routers in different not-so-stubby MANETs, independent local routing regions are dynamically joined by an overlay that spans any underlying Internetworks as a normal course of operational data communications. When multiple MANET local routing regions merge in this way, the MLAs present in all MANETs must be mutually exclusive. In the limiting case, all worldwide MANET local routing regions may be considered to be persistently merged over the MANET Internetworking overlay at all times. Statistical uniqueness properties of random assignments from even very large populations may therefore be insufficient to ensure collision freedom since MANET Internetworking exposes the full world population of MLAs as potential duplicates. Nodes in not-so-stubby MANETs should therefore configure MLAs managed for global uniqueness even if they first self-generate the MLAs (e.g., based on a secure hash of a public key) before enrolling them in a registration service. The node assigns its MLA to an overlay multilink network interface or another virtual interface such as a loopback. Routers in all MANETs also configure unique router IDs that may be derived from the MLA or randomly generated with sufficient statistical uniqueness properties within the local region. In addition to MLAs, L2-MANET routers must also configure L2 MAC addresses on their MANET interfaces that are assured unique at least within the MANET local routing region. MAC addresses can be configured either through random generation with universal/local (U/L) bit set to "local" or administratively assigned according to an organizationally unique identifier (OUI) with U/L set to "universal". If multiple routers within the same MANET local routing region somehow configure the same MAC address they must detect and deconflict the duplication to ensure routing system integrity. In addition to MLAs, an important use case for IPv6 Link-Local Addresses (LLAs) remains. MANET routers assign LLAs to their MANET interfaces by embedding their interface MAC addresses within the LLA Interface Identifier (IID) . This provides a next hop address for routes discovered by the MANET routing protocol and supports stateless forwarding based on the LLA's embedded MAC address without requiring address resolution messaging. For L3-MANETs, if each router assigns a MAC address independently, the possibility for duplication exists but this does not present a problem if the MANET router sets its router ID to an assured-unique value instead of an LLA. Any packets forwarded according to a duplicate next hop LLA would simply be deconflicted by the IP layers of the multiple next hop routers.

When a MANET comes in contact with a fixed Internetwork such as the global public Internet, nodes in the MANET that engage global mobile Internetworking services require some means of autoconfiguring global-scoped IP addresses or prefixes that are properly routable by network elements accessible from the current point of attachment. These network elements are typically proxies or gateways of some variety that connect to the mobile routing system. Nodes in L3-MANET local routing regions that are multiple IP hops away from a MANET border router with an Internetwork connection cannot use unmodified standard autoconfiguration services including IPv6 Neighbor Discovery (IPv6ND) or DHCPv6 over a MANET interface since these services are link-scoped in nature. (The DHCPv6 architecture includes a "relay" function, but the dynamic nature of links in (multi-link) MANET local routing regions may interfere with straightforward application of DHCPv6 relays.) Nodes in L2-MANET local routing regions see the MANET border routers as single-hop IP neighbors, but require a means of selecting specific border routers when multiple are present. Two methods of supporting generalized autoconfiguration for nodes within a MANET have been suggested. In a first method (conducted directly over MANET interfaces) first-hop neighboring nodes within the MANET collectively participate to repeat link-scoped autoconfiguration discovery requests to other neighbors that are topologically closer to a MANET border router. This hop-by-hop process continues between neighbors until the request arrives at a MANET border router that can then contact an Internetwork element capable of delegating an Internet Service Provider (ISP) Provider-Aggregated (PA) IP address or prefix. The Internetwork element then returns the delegated IP address/prefix in a reply that traverses the reverse path to the original requesting node. Each MANET router then configures a route to this IP address/prefix within the MANET local routing protocol, i.e., the MANET local routing protocol becomes aware of the delegation. In a second autoconfiguration method, the requesting node configures a (virtual) overlay multilink network interface over its (physical) MANET interface(s) and issues standard link-scoped IPv6ND and/or DHCPv6 requests over the virtual interface. The virtual interface applies encapsulation to provide the appearance of a single NBMA link spanning the entire (multilink) MANET. This virtual link supports standard link-scoped autoconfiguration services coordinated with an Internetwork element capable of delegating an address. For stub MANETs, the MANET border router itself delegates a public or private IP address. For not-so-stubby MANETs, an overlay Internetwork Mobility Anchor Point (MAP) delegates a MNP as an IP prefix maintained by the overlay independently of the Internetwork attachment point. The MAP then returns the delegated IP prefix in a link-scoped reply over the virtual interface that traverses the reverse path to the original requesting node. In one alternative, MANET routers located one or more hops from a MANET border router can regard the MANET border router as a Network Address Translator (NAT) to convert MANET local addresses into MNP addresses with no autoconfiguration requirements; they can also request address delegations directly from the border router's MNP(s) and use them to support communications with Internetwork peers according to the stub model. MANET routers can instead (or in addition) request their own MNPs and register their MLAs with a MAP while using the MANET border router as a transit intermediate system according to the not-so-stubby model.

Two nodes located within the same MANET local routing region should be able to communicate (across multiple hops if necessary) using MLA addressing with no external Internetwork infrastructure reference points. As long as the MLAs configured by communicating peers are unique, the MANET local routing system maintains continuous multihop forwarding services to ensure session continuity. Nodes within the MANET local routing region can discover the MLAs of peers using services like Multicast DNS (mDNS) . Peer-to-peer communications can then be coordinated in multihop fashion using encapsulation and header compression via an overlay virtual link spanning any MANET intermediate hops in the path. Header compression represents a fundamental requirement for maximizing MANET-local communications efficiency, since the size of the compressed encapsulation and upper layer protocol headers is significantly smaller than the size of the uncompressed upper layer headers when no encapsulation is applied. However, header compression state must be synchronized between the endpoints of each IP hop. For L2-MANET, this means that only the MANET ingress and egress nodes (as well as any intermediate cluster heads) need to synchronize IP state since all other intermediate hops are based on MAC layer addressing and forwarding. For L3-MANET, all MANET hops between the ingress and egress nodes must also synchronize state meaning that resynchronization is required following a path change. A tension therefore exists between L2-MANET which includes larger MAC headers and L3-MANET which may be challenged to maintain state synchronization in dynamic networks. For both L2- and L3-MANETs, the MANET can partition into multiple smaller MANETs with some border routers remaining in a first partition and others relocated to other partitions. This means that it is not possible for a MANET to maintain a classic global IP shared subnet where all nodes are presumed to be connected to the same (shared )IP link. The only shared IP subnet that should appear on the link is the MLA prefix, since the overlay ensures reachability for MLA addressed nodes whether they are located in the local MANET or in a different MANET reached via a border router connected to the overlay.

When an originating peer (or its stub MANET border router) within a not-so-stubby MANET needs to communicate with correspondents connected elsewhere in an external Internetwork, the peer consults the global DNS which returns a (stable) globally-routable IP address for the correspondent. The peer can then use one of its MNP-based IP addresses obtained through autoconfiguration and the global IP address of the Internetwork correspondent as the source and destination addresses for packet exchanges. To initiate communications with an Internetwork correspondent, the MANET peer first establishes per-flow on-demand virtual circuits in the overlay to an Internetwork relay beyond the MANET border. MANET local multihop routing will then convey the peer's original packets to the MANET border which then forwards them via the overlay to an Internetwork relay which directs the packets to the correspondent node. In the reverse path, the correspondent uses the MNP-based IP address of the peer obtained from the source address of initiating packets as the destination address for reply packets. Standard Internetwork routing will direct the packets back to the relay which then forwards them via per-flow overlay virtual circuits to the originating peer's MANET border. MANET local routing and forwarding will then convey the packets over one or more MANET local hops until they ultimately reach the peer. In this case, the originating peer's IP address need not appear in the global DNS since the correspondent discovers the address by examining the source of received packets.

When an Internetwork correspondent needs to communicate with a target peer within a MANET local routing region, the correspondent consults the global DNS to determine an IP address for the peer. The correspondent then forwards packets via standard Internet routing until they arrive at a relay. The relay then establishes per-flow virtual circuits in the overlay to the MANET peer while forwarding packets via the virtual circuit until they reach the destination. Reverse path forwarding from the MANET peer to the Internetwork correspondent is then conducted in the same manner described in . IP addresses covered by delegated prefixes remain stable even across MANET-wide mobility events to the point that continuous dynamic updates to the DNS are not required to maintain uninterruptable communications. While it is possible that mobility events may cause minor temporary disruptions, transport protocol retransmissions will maintain continuity for any ongoing sessions.

When two prospective peer nodes are located in different MANET local routing regions separated by one or more transit Internetwork segments, both peers should include their IP addresses in global DNS resource records for the same reasons cited in . The peers then establish per-flow virtual circuits in the overlay to support peer-to-peer packet forwarding. The peers may use either an MNP address or their MLAs, which are routable within the overlay limited domain. The overlay therefore exhibits the outward appearance of a MANET-of-MANETs, where overlay interior nodes engage in an interdomain global routing service bridging many MANET local routing regions. A certain degree of coordination between peer nodes and their MAPs is then required to maintain address mappings. The overlay ensures that each peer remains reachable at its stable IP address/prefix through distributed mobility management. Note that a common use case includes joining multiple partitions of a formerly unified MANET local routing region. The partitions may arise from pre-planned or un-planned node mobility patterns, but as long as each partition includes a MANET border router with an Internetwork connection nodes within different partitions can continue to communicate using their MLAs.

When a MANET border router connects a stub MANET to an Internetwork, it can either delegate global-scoped IP addresses to stub MANET routers or apply Network Address Translation (NAT) to non-global scoped addresses (e.g., IPv6 Unique Local Addresses) to support external communications. In the public case, all manners of peer-to-peer communications are made possible due to the globally routable nature of the addresses. In the NAT case, only communications initiated by a stub network peer are supported since the reverse path terminates at the NAT. The stub MANET itself may configure a local overlay that regards the (multihop) MANET as a single unified link. In that case, the stub network overlay link is distinct from the overlay link that spans the global public Internet and the two links are joined by the MANET border router acting as an IPv6 router. In the not-so-stubby case, a single overlay link extends across both any transit Internetworks and the source and target MANETs themselves. All peer-to-peer communications are therefore conveyed across a globally-extended MANET Internetworking overlay.

Many current and past solution proposals address most aspects of the problems articulated in the previous section, but often have functional gaps that do not fully satisfy the requirements. The following gaps must be addressed by candidate solutions:

In anticipated practices, nodes will often have multiple and heterogeneous MANET interfaces connected to the same MANET local routing region. Most notably, the vast majority of cellphones have both Wi-Fi and 5G cellular interfaces which will soon support a "SideLink" mode of operation for device-to-device communications outside of the context of cellular infrastructure. With their Wi-Fi interfaces engaged in "ad-hoc" mode and 5G interfaces engaged in "SideLink" mode, nodes within a MANET local routing region can exchange control information and forward packets across multiple hops if necessary even if some hops are carried over Wi-Fi and others over 5G SideLink. A unique addressing scheme is also necessary across the heterogeneous combined MANET region to maintain consistent and compatible routing information. Most importantly, the addressing schemes between the heterogeneous media must be mutually compatible and address assignments must be assured unique so that no duplication is possible. Aside from cellphones and the like (which represent an enormous use case space), multiple heterogeneous MANET interfaces often occur on mobile platforms such as tactical military and/or unmanned air systems. These same considerations also apply within any such domains.

In conjunction with Gap 1, heterogeneous MANET interfaces often also support diverse maximum packet sizes also known as the Maximum Transmission Unit (MTU). For example, the initial and final hops of a multihop path may support MTUs of 1500 octets or larger while some intermediate hops support MTUs as small as 1280 octets. Communications must therefore remain continuous and without loss of large packets due to a path MTU "black hole". Any solution proposal must therefore address the need for maximum packet size diversity given the clear need to support heterogeneous environments.

MANET data plane packets often include sizeable upper layer headers including any encapsulations, the IP header and upper layer protocol headers. Especially for smaller packet sizes, the overhead for carrying significant header payloads can waste precious wireless transmission bandwidth. L2- and L3-MANET solutions have different approaches to header compression that may be better suited for some operational scenarios than others. L2-MANET includes larger L2 headers with every packet such that IP header compression state is required at the MANET ingress, egress and IP-addressed cluster head nodes, while L3-MANET uses minimal L2 headers but requires IP header compression state also in MANET intermediate nodes. Any solution proposal must therefore address the need for header compression for multi-hop packet flows within the context of intended use cases.

Ordinary MANET routers must be able to discover and effectively engage the best MANET border router(s) especially when multiple alternatives may be present. For example, a MANET with 100 ordinary nodes may include 4 MANET border routers that also have longer-range Internetworking links such as via LEO satellites. Each MANET router must discover border routers that may be multiple MANET local hops away and somehow cause their packets to flow through the MANET to a border router that in turn invokes MANET Internetworking. Any solution proposal must therefore address these needs.

The fact that a node is admitted into a MANET local routing region does not ensure that the node is authorized to establish flow state in the MANET and/or invoke MANET Internetworking services. For that reason, a gap exists that requires a solution for securely admitting MANET clients into the MANET Internetworking service. Any solution proposal must therefore address the security gap.

The following sections provide an analysis of solution proposals with respect to the problem statements and gap analysis. Any solution proposal must explain how it addresses the problems while also showing how it either does or does not satisfy the technological gaps.

This document is an informational problem statement and does not in itself request any IANA actions. IANA considerations can be found in solution space documents.

MANET Internetworking assumes a multi-layer security architecture. Physical and data link layer security is assumed within each MANET local routing region and with proper network layer authentication for admitting nodes into the MANET Internetworking overlay service. Transport and higher layer security should then be applied for end-to-end confidentiality, integrity and authorization.

Discussions on the MANET working group mailing list helped shape concepts exposed in this document. The following are acknowledged for their helpful comments: Abdussalam Baryun, Lou Berger, Stuart Card, Juliusz Chroboczek, Christopher Dearlove, Donald Eastlake, Joel Halpern and others who provided valuable input on the list or through private discussions.

<< RFC Editor - remove prior to publication >> Differences from -03 to -04: Introduced distinction between L2/L3 MANETs. Supporting text added in response to list discussions. Differences from -01 to -02: Expanded gap analysis. Differences from -00 to -01: Updated based on list comments and private communications between April-June 2026. Differences from earlier versions: First draft publication.