Summary

  • At RIPE 93 on 26 October, RIPE NCC's Ties de Kock proposed opening RIS to more BGP peers, measuring their contribution, and deciding on permanent archiving separately.
  • The current RIPE NCC peering page promises that received BGP traffic is recorded, stored and publicly available; the proposal leaves trial-feed access and archive criteria for community discussion.

The guest list and the family album have always been close together at RIPE RIS. Let a BGP peer connect, and its messages join an archive that researchers can revisit years later. Ties de Kock's RIPE 93 talk asks whether those decisions should be uncoupled. Invite more guests to speak. Decide later which recordings go into the permanent album. Suddenly the interesting person in the room is whoever holds the delete key.

De Kock's original slides describe a network begun in 1999, now operating 23 route collectors. RIS publishes BGP snapshots every eight hours and update files every five minutes, a cadence also specified in its MRT documentation. Those files make yesterday's view inspectable tomorrow. Their volume also grows whenever more peers join and more changes arrive.

Slide 12 makes a useful correction to the instinct to blame a noisy peer. A feed with many updates may be relaying churn from elsewhere; the turbulence can be genuine visibility. Slide 15 says country is a weak stand-in for topological novelty, while remote and multihop sessions cost more to operate. For the new strategy, slide 18 prefers full tables, distinct views and new upstream or customer visibility. Multiple sessions from one AS can make sense when the locations really differ. A peer directly at a collector would face a lower acceptance threshold than a multihop session.

Then comes the consequential line on slide 19: collect first, decide on permanent archival later. Volume and information gain could be judged from what a feed actually shows. Permanent retention becomes its own policy decision. De Kock asks what tools should display data that never reaches the long-term archive. The current RIPE NCC peering policy tells participants that all BGP traffic will be recorded, stored and made available to the community. A future policy will need to explain any new distinction clearly to the peers supplying the data and the people reading it.

The hard part is a record of the choice. Slide 20 seeks views on the evaluation period, whether trial data should be public, what counts as meaningful information gain and how transparent the inclusion rules should be. Those are live questions. An excluded feed might still have shown an unusual route that later matters; without a durable trace of the retention decision, a future reader cannot tell whether silence means no event, no observation or a discarded observation. That is the governance story hiding inside the storage bill.

Original sources