Summary

  • APNIC's public member directory lists The Hong Kong Internet Exchange Limited as a MEDIUM member with country code HK and links it to a website. Both are accurate administrative facts about identity, and both are silent about throughput.
  • No traffic volume, participant count, port speed or site list for this exchange is established in our own record, and the operator's two statistics headings produced no readable figure at either the 17 October 2026 or the 11 November 2026 review.

An internet exchange is a room full of networks that have agreed to stop shouting past each other. Its public worth is measured in what crosses that room. Its registry presence is measured in whether somebody typed its name into a membership list correctly. Those are not the same measurement, and the space between them is where a great deal of confident nonsense lives.

Start with the row. APNIC's public member directory lists The Hong Kong Internet Exchange Limited as a MEDIUM member with country code HK, and links the member to its website (APNIC public member directory record). That is a real fact with a real use. It tells you an organisation exists, that an RIR recognises it as a member, that it is administered from Hong Kong, and that it claims an address on the web. Anyone doing a first sweep of who operates in a market needs exactly this, and gets it cheaply.

It is also, on its own, a terrible basis for judging an exchange.

Our own directory entry for the entity is unusually candid about where it stops (directory record). Three limits are written straight into it. The legal registration record, public website and leadership contacts have not been verified against official or company-registry sources. The entity's ASN and IP resources have not been mapped from RDAP, WHOIS, IRR, BGP or RIR delegation data. And the entity should not be classified as an ISP, IP transit provider, cloud provider, registry or managed-network provider without a current public source. The record does offer a legal type — private company, Hong Kong jurisdiction — and in the same breath flags that legal profile as unverified against any company register.

Read those caveats as weakness and you have misread the instrument. A record that marks its own uncharted territory is doing its job. The alternative is a card that fills the blanks with plausible guesses, and a reader who then cannot tell which line came from a registry and which came from someone's intuition. Accurate incompleteness beats false precision every time.

What the card cannot do is answer the only question anyone asks about an exchange: is it big, is it busy, is it redundant, is it still switching? A tier label is not a traffic count. A country code is not a port inventory. A hyperlink to a website is not evidence of what that website says — the APNIC entry points at a member's own address, and a link is a pointer, not a quotation.

The operator's own publication is where measurement should exist. On 17 October 2026, prior coverage of this subject found the site presenting "Aggregate Switching Statistics" and "Number of Route Prefixes Announced" as two separate headings, and dividing participants into four labelled classes — R&E Participants, FTNS Providers, Resellers, and Satellite and Reseller Sites. No numeric value under either heading came back readable at that review. On 11 November 2026 the same pages again yielded no readable numeric text. The operator's website is, for the record, the same address the APNIC entry links to (the linked website).

Two separate headings is a taxonomy decision, and it is a defensible one: traffic switched and prefixes announced measure different things, and a single blended number would hide more than it revealed. Four participant classes is likewise a classification choice with consequences — an academic network, a fixed-carrier licensee, a reseller and a satellite site are not the same kind of member, and collapsing them into one headline count would flatter whoever has the most of the cheapest kind.

None of that tells us what the numbers are. And the reason they were not read matters less than what people do in their absence. The cause was not determined: client-side rendering, access control, a chart delivered as an image, or a genuinely static page are all consistent with what came back, and none of them may be converted into a claim that the counters read zero. An access gap is an access gap.

The temptation is to fill it with the directory. If the exchange is a MEDIUM member of APNIC, the reasoning goes, it must be a mid-sized operator, and the fabric must be mid-sized too. That inference fails at the first joint. A membership tier describes a relationship with a registry — a category of participation, not a volume of traffic. Nothing in the record defines what MEDIUM means operationally, and a label created for one purpose does not silently acquire a second purpose because a reader needs it to.

The costs of that confusion are asymmetric. A registry entry is cheap to publish and cheap to read. A traffic series with stated definitions, aggregation windows and update timestamps is expensive to produce and easy to get wrong. When the second is unavailable and the first is right there, the cheap artifact wins the argument by default, and due diligence quietly shifts onto peers, customers, journalists and anyone else trying to size an interconnection point from outside. The party controlling the published surfaces keeps the interpretive advantage; the party reading a tier label mistakes status for capacity.

What would close the gap is unglamorous and specific. A readable time series, timestamped, with definitions for what each counter includes and excludes. An ASN and prefix mapping that attaches the entity to routing that can actually be observed. Independent route-collector evidence that distinguishes fabric announcements from unrelated traffic. Each of those closes one gap and none of them substitutes for the other two: a prefix list proves participation, not throughput; a traffic series proves volume, not redundancy.

Until something like that lands, the honest reading of this exchange is narrow and useful. It is a registered member of an RIR, administered from Hong Kong, with a public site that poses the right measurement questions and did not answer them for this review. That is a fact about the record. It is not a fact about the fabric, and the two should never be spent as though they were the same currency.