Summary

  • APNIC's current resource policy gives a new Internet exchange point a /26 IPv4 delegation by default. Its IXP range runs from /26 to /22, with larger requests tied to projected peer numbers, actual use and the return of older space.
  • When an IXP receives a replacement block under the growth rules, it must return the previous one within three months. IXP-policy IPv4 is for peering, is non-transferable and must stay out of the global routing table apart from mistakes; APNIC's text says it will revoke resources used otherwise.
  • Implemented proposal prop-154 moved the default from /23 to /26. The change made scarcity management an operating condition for an exchange, not merely a smaller first envelope.

An IXP operator has a simple problem. The fabric needs numbered interfaces so members can exchange traffic. Those addresses live in the middle of the interconnection business, where a change of address can force members to update interfaces, sessions, filters and records. APNIC's policy supplies the numbers, but writes the renewal and exit terms on the same page.

The first envelope is /26: 64 IPv4 addresses in a block, before reservations and actual interface assignments. APNIC's section 6.2.4 sets that as the default for a new IXP and the floor for this type of delegation. A forecast of more than 60 peers on the peering LAN within 12 months can support up to /25. More than 100 can support up to /23 or the current highest delegation size. An exchange holding less than /24 can ask for as much as /23 after using 60% of its original block. The ceiling is /22.

These are different tests. Forecast peers may justify a larger initial block; demonstrated use can justify a replacement. The policy counts the peering fabric, not the exchange's press releases, municipal ambitions or the size of the country on its map. Its larger-block rules require an operator to turn membership growth into a number APNIC will accept.

The sharp edge sits after approval. The existing delegation must be returned within three months of the new one. A large legacy IXP that used 60% of a prior /23 may seek a contiguous /22, if available, on the same return condition. An IXP opening new points of presence can seek additional /26 or /25 space, within the /22 total. The exchange can grow; it cannot warehouse every earlier peering LAN block as permanent spare inventory.

That three-month clock lands on the people running the exchange and its member networks. They must plan the cutover, readdress the peering fabric where the old block was used, and coordinate the records and sessions that refer to it. APNIC's policy states the return deadline. It supplies no case-by-case transition cost in the public text. A specific exchange's migration plan would change the cost estimate.

The purpose restriction is blunter. The same section reserves these IPv4 addresses for IXP peering, says they will not be announced in the global routing table, exempts mistakes, and says APNIC will revoke resources used otherwise. It also makes every resource delegated under this IXP policy non-transferable. APNIC thus controls more than the size of the first allocation: the text follows the block into use and closes the ordinary sale route.

One adjacent sentence leaves global routability of delegations outside this policy to the IXP and its participants. The word “outside” matters. It does not turn an IXP-policy /26 into a general-purpose globally announced block. APNIC's separate IPv6 IXP rule offers a minimum /48 and expressly leaves global routability to participant discretion. Operators reading the two address families as one undifferentiated promise will miss the IPv4 restriction.