Topic
Network-resource Evidence
Within the Topic facet, Network-resource Evidence topic intelligence connects articles that share a specific subject, signal focus, or monitoring theme. The page gives readers a richer path through related reporting, source evidence, market actors, and infrastructure implications, with enough context to understand why the topic matters across company movements, governance decisions, regional exposure, and operational risk. Readers can compare recurring signals, affected organisations, public evidence, market context, service continuity, procurement, competition, compliance, and strategic planning questions behind the subject instead of stopping at a thin list of matching articles. It explains what the topic covers, which infrastructure actors or policies are involved, what evidence supports the coverage, and why the subject may matter for operators, customers, investors, and policy readers.
CASE FILE
Yesterday’s Path Lent the New Flow Its Window. It Did Not Lend Its Capacity: RFC 9959 and the Authority to Reuse Congestion State
A service remembers that yesterday’s connection filled a long, fast path, so today’s connection starts with more confidence than an ordinary new flow. The gain can be real. So can the queue it creates when the same address now reaches another server, route or bottleneck. RFC 9959…
Story
APNIC Called RPKI Mirroring Marginal. Its NNIX Pilot Tests a Different Benefit
In October 2025, APNIC assessed a repository mirror as a short-lived aid during a long outage and called the benefit very marginal. Seven months later, its agreement with NNIX proposed a Hangzhou mirror pilot around latency, validation speed and data availability. The…
Story
ARIN's NET Permissions Change With the Lookup Path
One child NET, three ways to identify it, and three different questions about authority. An ARIN community suggestion reported that a direct registrant could reclaim a Detailed Reassignment and retrieve it through an exact address range, but could not use the same API key to GET…
IETF
Kireeti Kompella and the Echo Reply That Did Not Prove the Service
An MPLS Echo Reply can show that one carefully constructed probe reached a router able to account for a particular forwarding equivalence class. That is a precise and valuable result. It does not show that every equal-cost path worked, that a dormant backup was usable, that the…
History
The Host Table Said TCP. The Socket Had to Answer: How RFC 832 Measured Running Code
In December 1982, the Internet already had a list of machines said to support TCP. David Smallberg did not treat the list as deployment. He used it as a set of claims, tried the service ports, named the different ways a connection could fail, retried negative observations, and…
History
The Client Became the Service: How RFC 818 Put User Telnet Behind Port 107
A client normally knocks; a server normally listens. RFC 818 made that picture deliberately unreliable. On a small terminal concentrator, the useful application was the Telnet program that opened connections outward. By placing that User Telnet capability behind a well-known…
Story
APNIC Counted 2,974 IDNIC Allocations. Its Three Completed NIR Reviews Have No Public Denominator
APNIC can show, with unusual precision, how one preliminary review began: 2,974 IDNIC allocations, a fixed date window and 1,216 records requiring more investigation. It can also say that three other NIR reviews are complete. What the public record cannot do is connect those…
Story
AFRINIC's Current RPKI CPS Says It Manages Asia-Pacific
The Certification Practice Statement linked from AFRINIC's current resource-certification page is version 3.0 from May 2020. In its CA-termination clause, the registry assigns itself the Asia-Pacific region. AFRINIC and IANA assign that region to APNIC and place AFRINIC in Africa…
History
The Name Service Was Almost a Negotiator: How RFC 830 Separated Domains from Capabilities
A file-transfer program asks for NIFTP. The destination has no NIFTP implementation, but it does offer FTP. In the architecture proposed by RFC 830, that mismatch did not have to end the attempt. One part of the name service found the destination domain; another part asked what…
CASE FILE
The Collection Held Every Evidence Item. It Had Not Bound Them to One Machine: RFC 9999 and the Composite Attestation Boundary
A verifier can receive a neat package labelled CPU, SmartNIC and GPU, decode every member and validate several signatures. None of those facts alone proves that the three reports describe the machine requesting access now. RFC 9999 makes remote-attestation messages portable; it…
CASE FILE
The Envelope Named the Evidence. It Did Not Bind the Device: RFC 9999 and the Authority Inside an Attestation Collection
A server can present signed reports from its CPU, SmartNIC and accelerator in one neat package. The package may be well typed, easy to route and validly encoded while still failing the question that matters: do all three reports describe this device, in this appraisal, under one…
Story
AFRINIC Has 3,573 Address Objects Pointing to Geofeeds. Its Schema Still Has No Geofeed Field
AFRINIC’s public database already carries geofeed pointers at material scale, but it carries them as prose. The fallback is valid and useful. The missing question is whether a convention used by thousands of address records should remain a string that the schema cannot identify…
History
The Number Did Not Make It a Standard: How RFC 825 Made Intent Part of the Record
A procurement clause says a product is “RFC 825 compliant.” The number looks precise, the link is public and the document is real. Yet none of those facts tells the buyer whether the cited memo was a standard, a discussion, a status report or merely information. In 1982, RFC 825…
Story
RIPE NCC Reports a 467-Day IPv4 Wait. The Queue Needs a Flow Ledger
RIPE NCC’s August member update says 757 LIRs were waiting for IPv4 space and the first LIR had been in line for 467 days. Its public chart records the queue’s size and its oldest wait, but those two stocks cannot show whether movement came from allocations, new requests…
History
The Layer Was Not the Module: How RFC 817 Cut Across the Stack
One typed character could make three answers leave a server: a TCP acknowledgement, a window update and a Telnet echo. The protocol layers were each behaving correctly. The implementation was still wasting packets. RFC 817 used that small inefficiency to make a larger…
Story
RIPE NCC Needed 394 Supporters Before a Member Resolution Could Reach the Agenda
The number 394 was not a vote result. It was the price of admission to the vote: a cycle-specific calculation deciding whether a member-authored subject and its verbatim resolution would enter the General Meeting agenda. The rule has a sound institutional purpose. The public…
Story
LACNIC Published a 98% Outreach Score. The Response Count Is Missing
LACNIC says more than 330 organizations attended its 2025 member gatherings and that 98% of the feedback it received fell in the highest satisfaction levels. Both statements may be accurate. The public post does not disclose the number of responses, so the percentage cannot be…
CASE FILE
The Address Was Unique. The Wire Still Saw a Collision: RFC 10019 and the Authority to Reconcile Zeroconf Multicast
Two isolated network segments can each make a locally correct multicast allocation. Join them, and both histories may remain truthful while the merged wire can no longer honour both claims. RFC 10019 turns that edge case into an operating principle: an address is not unique until…
History
The Error Message Was Advice, Not a Verdict: How RFC 816 Layered Failure Decisions
A gateway that has crashed cannot send the error message explaining why packets disappear into it. RFC 816 began with that silence and built a layered answer: let routing repair distant failures, make the host replace a dead first hop, let TCP expose retransmission and timeout…
Story
ARIN Put 17,000 Legacy Networks in a Blog Post. The Status Table Is Missing
ARIN has published an unusually concrete warning about the contact records attached to legacy IPv4 space: roughly 17,000 networks, an `about half` risk description, 11% currently validated, another 20% incomplete, and a 76% comparison for resources under agreement. The figures…