Summary

  • The RIPEstat AS-overview response for AS24385, read at 08:00 UTC on 12 November 2026, carries a holder string reading "Smartlinx3 - Wellington Regional Internet Exchange", an APNIC-assigned block of 23552–24575 and announced=false — and no allocation date, service list or legal form.
  • The only Wellington peering product with a published price belongs to a different operator, NZIX: 10G at NZD 350 per month and 100G at NZD 1,250 per month, which re-derives to NZD 0.035 and NZD 0.0125 per Mbps per month. Neither surface tells a buyer how many networks actually connect.

One JSON response is the entire documentary basis for an internet exchange that appears in a published list. The RIPEstat as-overview for AS24385 returns four things worth reading: resource 24385; a holder string, SMARTLINX3-AS-AP - Smartlinx3 - Wellington Regional Internet Exchange; a block, 23552–24575, marked "Assigned by APNIC"; and announced: false.

That last field is the one an engineer notices. For that single query window, RIPEstat showed no route origination against the number. One instant is a photograph, not a trend — it does not establish that the holder runs an exchange fabric, sells ports, or has stopped existing.

And the response stops there. No allocation date. No last-change date. No service description, no legal form, no address. A registry coordinates uniqueness: one number, one holder, no collision. It is not a service catalogue, and this one does precisely the job it was built for.

What the card adds that the record does not

The directory entry behind this briefing goes further. It lists two service roles — managed_network and peering_ixp — a legal type of private_company, and a country field reading "Global", on a confidence score of 0.7 with exactly one evidence basis: that same RIPEstat response. Its own data-gaps field then asks the reader to verify the legal entity country, current services, visible assets, parental links and leadership contacts.

Read those two fields side by side and the card is admirably candid about itself. Read only the service roles and it becomes something else: an inference dressed as an attribute. "Wellington Regional Internet Exchange" is a fragment of a holder string. Lifting "peering IXP" and "managed network" out of it is a reading of English, not a measurement of a port.

That is the ordinary arithmetic of enrichment. Sparse registry data looks thin; a profile with service tags looks authoritative; the confidence score and the data-gaps note are the fine print. The fine print here is correct. The headline is not.

The organisation behind the holder string

A mirrored copy of APNIC-derived data goes one level deeper and separates the number from an organisation. It shows AS24385 SMARTLINX3-AS-AP with the description "Smartlinx3 - Wellington Regional Internet Exchange" and an aut-num last modified 2023-04-14; it attributes the resource to ORG-CL4-AP, Vital Data Ltd, org-type LIR, last modified 2025-08-06.

That answers a different question. Not who the exchange is, but which organisation holds the number — and the two are not the same claim. The mirror is also the only retained document that labels an allocation date, 2005-03-29, and the only one carrying an exchange-policy reference. That reference points at nzix.net, the legacy domain, while the operator describing a Wellington exchange today publishes from ix.nz. A policy address inside an aut-num record is not a current contract, and nothing here establishes what that document says.

The allocation date needs its own line. A second index displays 30.03.2005 for the same number. One day apart, unreconciled. The date belongs in this briefing as a mirror label with a conflict attached, not as a registry fact.

The only Wellington product with a price

Now the thing a network can actually buy. The New Zealand Internet Exchange describes itself as established in 2016, not-for-profit, carrier-neutral, owned and operated by its members. Its peering page states that it runs exchanges in Auckland, Christchurch and Wellington, and hangs a price ladder on Wellington: 10G at NZD 350 per month, 100G at NZD 1,250 per month, present at Xtreme DC and Spark Wellington Central Exchange.

Divide and the unit prices fall out. NZD 350 ÷ 10,000 Mbps gives NZD 0.035 per Mbps per month; NZD 1,250 ÷ 100,000 Mbps gives NZD 0.0125 per Mbps per month. Multiply each back and the monthly totals return exactly. The 10G unit price is 2.8 times the 100G unit price. Ten times the capacity costs 1,250 ÷ 350 = 3.57 times the money.

That is a structure you can argue with, because it has denominators. The peering services schedule tightens the boundary: Layer 3 BGP peering between autonomous systems; the customer supplies its own access facilities; peering is not internet access. What is sold is a port, a protocol and a peer set.

And there the trail ends. On the peering page as read on 12 November 2026 there is no Wellington participant count, no port inventory, no traffic volume and no service-level commitment. The price sheet is public. The measurement layer is not.

What this does to a decision

Consider the person this is written for: an engineer or a commercial lead at a New Zealand network choosing where the next port goes. The registry name is no help — an identity, not an offer. The card is no help either, because its service tags were derived from the same string and its own verification note concedes the gap.

The rate card answers exactly one question well. It says the unit price falls by a factor of 2.8 when a network moves from 10G to 100G, which is the most useful number in this briefing. It also says the operator publishes no count of who is already connected. A buyer signing at NZD 1,250 a month still cannot tell from these pages whether the peer set justifies the port — and the peer set is the product.

Two Wellingtons, then. One is a name in a registry, free to assert, impossible to audit. The other is a service with a price, a protocol and two street addresses, still missing its denominator.

The exchange worth publishing

BTW's position, stated plainly: a registry record earns its authority by coordinating identity and nothing beyond it. The moment a name is upgraded into service roles, legal form and market presence without a document behind it, the record stops informing and starts persuading. The fix is not better inference. It is a published, measurable operating surface — capacity tiers, unit prices, sites, obligations and, above all, the size and shape of what is actually running.

Smartlinx3's Wellington Regional Internet Exchange may be an operating fabric, a dormant identifier or something between the two. The documents retained here establish none of those. NZIX's WLG-IX, by contrast, is buyable at a stated price and still leaves a buyer guessing at the figure that would settle the argument.

Wellington is small enough that this matters more than it sounds. A regional exchange proves itself by letting a network inside the region reach its neighbours without leaving the country. That proof needs a participant list and a traffic curve — not a holder string. Publish the denominator and the exchange becomes a market. Keep publishing the name, and the market stays a rumour.