Summary

  • Tomoya Yoshida helped write APNIC prop-061 after network tutorials ran out of harmless AS numbers to put in examples. A copied configuration can carry an example into a live router.
  • The APNIC Policy SIG agreed in August 2008 to reserve sample numbers on both sides of the 16-bit/32-bit divide. The work became RFC 5398, authored by Geoff Huston, that December.
  • IANA now marks 64496–64511 and 65536–65551 for documentation and sample code. These are two sets of 16 numbers with a job: stay out of real peering.

A training slide needs a cast of characters

There is a particular sort of embarrassment in discovering that the “made-up” network in your training slide belongs to somebody. Internet routing examples need AS numbers: an operator here, a neighbour there, a path between them. Pick a memorable real number and the diagram looks tidy. Paste the lesson into a router and the tidy diagram gets ideas above its station.

Tomoya Yoshida knew the machinery behind that risk. APNIC's 2008 speaker biography describes his work on the OCN backbone, peering and Internet Routing Registry services at NTT and JPNIC. With Philip Smith and Gaurab Raj Upadhaya he co-authored prop-061: a small, deliberately boring supply of ASNs that writers could use without borrowing anybody's identity.

The final APNIC proposal reads like a catalogue of the shortcuts available before that supply existed. An author might use an employer's ASN, a private-use number, a convenient public number or an unassigned value. The last option carried a time bomb: today's empty slot can become tomorrow's assignment. The proposers warned that examples copied onto lab equipment connected to production networks could produce confusion and disruption. The paste command alone makes the risk easy to picture.

Two small blocks, one very useful boundary

At APNIC 26 in Christchurch on 28 August 2008, the Policy SIG reached consensus with a practical amendment: examples needed numbers in both the old 16-bit and newer 32-bit ASN space. Version 2 followed the next day. APNIC's history records the eight-week comment period and the October decision to await an RFC. The answer arrived in December as RFC 5398. Geoff Huston wrote the RFC and acknowledged Yoshida and his two fellow proposers. Keeping those roles straight matters: a regional policy idea became a globally legible documentation convention.

The chosen ranges are compact: 64496–64511 and 65536–65551. IANA's registry still labels both for documentation and sample code. The first block ends immediately before the private-use range begins. The second lives just beyond the 16-bit ceiling, so a tutorial about four-byte ASNs can use four-byte examples. RFC 5398 says operational BGP configurations should avoid peering with neighbours using these reserved numbers. The warning belongs beside the examples, because somebody may still paste them.

Yoshida's contribution sits in a modest corner of Internet governance. Thirty-two numbers make a lesson safer to copy, a manual less likely to impersonate an operator, and a boundary easier to teach. Sometimes a good network policy is a stage prop with a clearly marked exit.

Original records