Summary

  • RouteViews' January 2025 peering policy keeps accepting routes through exchange route servers, while generally reserving direct bilateral sessions for networks willing to send a full table.
  • At RIPE 92, Nina Bargisen tested what peers contribute through unique prefixes, origins, AS-path edges and paths. The question is what a session reveals beyond the feeds already on disk.

An archive can collect the same road map twice and still boast of having two maps. Nina Bargisen, RouteViews' peering coordinator, has made that arithmetic harder to sell. The project once operated an open peering policy. Its published replacement says yes to everyone’s route-server routes, but a direct exchange session normally calls for a full routing table. A multihop peer must show its full view; two sessions provide ordinary redundancy, with additional ones when their views differ enough. The data remain publicly available.

Bargisen gave the selection a measurement problem at RIPE 92. She compared what individual feeds add to the rest of RouteViews: new prefixes, origins, neighbouring AS pairs, complete paths and shorter paths. In her examples, full feeds from content networks at multiple locations exposed local interconnections that a standard customer-cone feed did little to reveal. A traditional backbone contributed some shorter paths but comparatively few new paths in her illustrated sample. The useful ranking is of the additional information in each feed.

The cost side is concrete. In her January 2026 account, Bargisen reported 883 full-route sessions from 277 unique ASNs at that point. RouteViews' storage for RIBs and updates had grown from 11.1 to 67 TB during 2025, driven by more peers, larger tables and noisy update streams. More than 50 existing peers began sending full tables. A selective policy lets the team pursue that kind of added view while declining sessions that repeat what a route server already supplies or destabilise a collector.

The sharp edge of the policy is discretion: it explicitly leaves exceptions to the peering coordinator and allows policy changes without notice. RouteViews has not published a per-request decision ledger or a marginal-value score for each admitted feed; those records would show how consistently the rule finds new regional views. For now the public test is simpler: does another session reveal a connection the existing archive could not see?