APNIC

APNIC says a maintenance step intended for one load-balancer node was run against the other, causing an outage across eight listed services. The notice calls it 15 minutes, although its displayed start and end times are 17 minutes apart.

The current call asks applicants to describe their own operational work in their own words. It also gives a closing date: 12 October at 23:59 Hong Kong time.

APNIC’s published service-quality line reads 94%, 95%, then 93%. The numbers remain high, but the 2026 endpoint brings a broader respondent mix into a measure previously reported for Members. A Member-only bridge would preserve the trend without erasing the wider community’s view.

The smallest bar in an APNIC 62 chart is precise to one decimal place. The footnote does more work than the bar: 200,000 geolocation rows were sampled, publishers were reduced to distinct netnames, and two of the five RIRs were left out. The result is a useful signal about one observed success set. It is not yet the percentage of APNIC networks that adopted geofeeds.

Four service names appear in APNIC's notice for 29 September, with no downtime expected. One of them is not merely a page to reach. The AS0 ROA Service makes a security assertion about address space that APNIC has not delegated. If an address moves out of that category, the useful question is when the assertion moves with it—not whether an endpoint still answers.

APNIC 62 showed a useful cross-server mismatch signal. Its meaning begins one step earlier: which daemons, views and zone configurations entered the cache, when they were fetched, and what the fetch could not observe.

Presto’s pseudo-segment solves a hard pipeline problem by turning a backward state correction into another forward pass. That synthetic record can be authoritative inside the transport without being user traffic, a completed socket operation or proof that an application received anything.

APNIC Labs added a second dual-stack authoritative nameserver and watched average queries per test fall from 3.43 to 2.57. The experiment did not reveal a cause. It revealed something more operationally awkward: a public recursive-resolver address is often a routing handle, not a faithful identity for the machinery making each decision.

The 2026 APNIC Survey does not say technical training has failed. It says something harder to measure: respondents value skills development, yet want the journey from a lab to a working network to remain connected, observable and locally supported.

A route-origin filter can leave an operator’s table clean while a downstream neighbor forwards toward the rejected destination. An APNIC 62 presentation made that uncomfortable distinction concrete. The useful response is to preserve separate evidence of received routes, local decisions and actual paths—not to promote a monitoring feed into a packet witness.

The certificate was on the revocation list. That was not the end of the story. Geoff Huston's APNIC 62 demonstration put the same adverse record in front of three browser policies and obtained different results—a useful warning about the distance between publishing a security decision and making it operative for a user.

Two MyAPNIC incidents ended after exactly 28 minutes within roughly 49 hours. APNIC gave each a bounded explanation; it has not publicly said whether their dependency paths were compared.

Presto’s POSIX layer can preserve the application’s familiar socket boundary. The published prototype nevertheless asks operators to condition the NIC and every peer that meets it on the wire.

An account should not inherit yesterday’s lock answer when today’s IRT contact state says something else. APNIC’s 2026 roadmap proposes replacing reliance on a static account lock flag with a login-time check of continuously tracked IRT email validity. That is a small sentence with a large control consequence: the source of an access decision moves closer to the evidence it is meant to enforce.

The band could be lawful, the access point could contain the right radio, and a device could still never see the network. APNIC 62 turned that gap between authorization and usable capacity into an operational record—but not yet a reproducible one.

APNIC measured fewer authoritative DNS queries in a later run using two nameservers than in an earlier run using one. The observation is real in the published aggregates; the cause is not settled. A reproducible decision needs the two treatments to share a clock, cohort and topology receipt.

APNIC's new survey puts RDAP at 20% among respondents asked about core registry services. The number is useful evidence about reported use; it is not a census of deployed systems, a traffic share or a verdict on whether RDAP can replace WHOIS.

On 14 September, two official APNIC pages described prop-170 as having reached consensus at APNIC 62. A third official page still labelled it “For Discussion at APNIC 62 OPM”. The operative policy supplied a more important answer: APNIC-127 still contained the old IPv6 allocation rule and no occurrence of the word “nibble”. None of those surfaces needs to be false. They record different moments in a governance chain—and only one of those moments changes what an applicant can actually receive.

A shorter opening application can widen access to Internet-development grants. It does not make the entire selection sequence an independent committee decision—and the 2026 rules do not claim that it does.

BIRD can stop a leaked route before ordinary filters have an object to retain. The September warning asks for more than softer rejection: an explanation that survives without giving the route permission to travel.

APNIC’s leadership consultation identifies a barrier that an election cannot remove: somebody must make the working day available. Published EC guidance distinguishes unpaid service from reimbursable expenses. The harder question is how to support participation without turning the supporter into the person being represented.

APNIC plans to show current and historical RPKI VRP information for an individual Internet number resource in REx. That could turn a specialist dataset into useful public memory. But a dated line is not self-explanatory: unless it names the observation, derivation and scope behind the date, readers may mistake REx’s view for the object a holder signed, the route a network received or the policy a router applied.

IRINN’s APNIC 62 update offers an unusually useful test of capacity-building evidence: thirteen visible webinar counts reconcile exactly to 3,567, while a footer on the same page reports more than 5,000. Both may describe real activity, but the public record does not say whether they count the same thing.

An API call can finish successfully without settling the more important question: which registry rule did it execute? APNIC’s public records show an ASN-capable NIR endpoint, a direct-delegation project that reached final QA, and a roadmap card with no release history. The missing link is small, but it separates usable software from auditable authority.

APNIC found a majority willing to consider paying extra for premium support—but only inside a preselected group, and without naming the price or the service bought. That makes 53% a reason to ask the next question, not an answer to it.

An APNIC 62 case study exposed a small but consequential accountability gap: the registry data could be right, several geolocation databases could agree, and one downstream service could still send a customer to the wrong country.

APNIC asked its community to divide exactly 100 points among eight strategic activities. The resulting chart looks unusually complete: every point is accounted for and no priority overwhelms the rest. The publication trail is less settled. Three official artifacts attach three different response bases to the same question, while the averages disclose a centre without showing how any individual respondent built a budget.