Summary

  • RFC 2009 put the multicast address of the smallest available partition containing a target polygon in the packet header, while preserving the exact polygon in the body for last-mile filtering.
  • Local propagation of small-area state and designated routers reduced routing-table growth, but accepted non-optimal paths, redundant copies, false positives and a final location decision that the network could not prove.

A flood warning addressed to a hand-drawn shape creates a problem that an ordinary destination address avoids. The shape may cross cells, fail to align with administrative boundaries and contain no stable list of receivers. RFC 2009’s answer was not to make every router understand every coordinate. It split the job.

The paper called the smallest geographically addressable areas atoms. Larger areas containing atoms were partitions. A sender could name a place or draw a polygon. If the polygon did not correspond to an addressable atom, the first leg used the smallest available partition that contained it. Its multicast address went into the network-layer destination. The original, more precise polygon remained in every packet body.

That duplication was the architecture’s central honesty. The header described where the network could route economically; the body described what the sender actually meant. A base station whose service area intersected the partition received the packet, then the second leg—the paper’s “last mile”—decided whether a receiver’s reported position fell inside the polygon.

A smaller table bought a coarser first answer

Mapping every possible small area into a complete multicast tree would have defeated the scaling goal. RFC 2009 therefore proposed that small-partition membership travel only locally. County information could travel farther; state or coast-sized aggregates farther still. A remote router did not need to know which atom contained a distant base station. It could fall back from exact S/C/x state to county S/C/0 and then state S/0/0.

Simple aggregation still risked presenting many links as routes toward one large area. The paper selected one or a few designated base-station routers per partition whose aggregate memberships could propagate. One designated router could leave a remote table with one aggregate link. The RFC explicitly acknowledged the price: the path might not be optimal.

Nor did designation eliminate duplication. Alternative routes and overlapping base-station ranges could deliver the same packet more than once. The receiver side still needed a duplicate-suppression mechanism. Coarse partitions could also touch cells outside the exact polygon, creating false alarms that disappeared only after later filtering.

The last mile did not possess a location oracle

The simplest edge method let the base station multicast the message and asked each receiver to compare a GPS reading with the polygon. A second method advertised a temporary multicast group and polygon; clients that judged themselves inside joined before the long message arrived. This spent setup time and group state to reduce irrelevant traffic.

Both methods depend on a position assertion. Outdoors it might come from a GPS card. Indoors, where the RFC admitted GPS could fail, it proposed using the configured position of an indoor base station. That locates an attachment point, not necessarily a person. A successful polygon comparison therefore proves only that a calculation accepted supplied coordinates.

The evidence chain remains longer: routing reached a coarse region; a device supplied or inherited a position; software calculated containment; the device joined or received; the intended person actually possessed it; the message was understood; and the service produced the intended result. No one link proves the next.

An Experimental address was not an allocated Internet

The address diagrams were proposals. RFC 2009 noted that its geographic multicast subspace and use of 240.0.0.0 were not registered. Its status was Experimental, and RFC 2026 placed such specifications in research and development rather than the standards track. Publication records an idea; it does not record adoption.

The security section also blocks a modern success story. Location could be used to restrict information or to track a user’s activities. The paper called for safeguards but did not specify a complete system for authentication, authorization, confidentiality or protection against false coordinates. IP multicast membership, grounded in RFC 1112’s host-group model, is not proof of audience identity. A DNS LOC record under RFC 1876 can express coordinates and precision; it cannot attest physical presence.

Lu Heng’s distinction between a coordination artifact and voluntary adoption keeps the historical claim narrow. His running-code test asks what operators actually implemented, while his reality-layer distinction separates a symbolic polygon from executable forwarding and observed service. RFC 2009 is valuable precisely because it exposed those layers: approximate routing could scale without pretending to answer the final question.

Sources