Content Type
Research
Within the Content Type facet, Research intelligence gathers BTW.MEDIA articles that share the same editorial format, helping readers compare briefings, profiles, risk notes, market analysis, and event coverage without mixing different kinds of evidence. The page explains how this content type frames internet infrastructure events, company movements, governance decisions, operational signals, and public evidence across the site. Readers can compare which actors or infrastructure systems appear most often, how source quality changes interpretation, and whether the material is a durable profile, a time-sensitive event, a strategic market signal, or a governance development. The result is a useful search page for operators, investors, customers, analysts, and policy stakeholders who need to understand the consequence, timing, and evidence behind similar article formats.
IETF
RFC 9874: A Successful EPP Delete Can Break Another Client’s DNS
The dangerous EPP deletion is not necessarily the one that fails. It is the one that succeeds for the requesting client while changing name-server state relied on by domains sponsored by somebody else. RFC 9874 turns that hidden dependency into an operational question. The…
Leaders
Remco van Mook and the IPv4 Gateway That Has No IPv4 Neighbour
An IPv4 default gateway normally looks like an address that a host can find with ARP. Remco van Mook's current IETF proposal turns that expectation inside out: `192.0.0.11` is not a neighbour to discover, but an instruction to reuse a neighbour that IPv6 has already discovered.…
IETF
The Network Agreed the Root Was Down. It Did Not Prove a Crash: RFC 9866
RFC 9866 lets a low-power network reach a coordinated decision that its current routing root is no longer usable. The decision can be operationally correct even when the root is still powered and the physical cause is unknown. Treating `GLOBALLY DOWN` as a forensic verdict would…
Leaders
John Ousterhout and the Fast Write That Became Cleaning Debt
The log-structured file system made a small write look almost effortless by turning it into part of a long sequential transfer. Its deeper lesson was not that the cost had vanished, but that systems must account for where the cost returns: in cleaning, spare capacity…
History
RFC 2350 Described the Door, Not What Happened After the Knock
A CSIRT can publish a careful service description and still leave a particular incident entirely unproven. RFC 2350 made the public interface legible; it did not turn contact details, service names or cooperation policies into case evidence.
History
The Server Drew Three Mail Maps. It Did Not Prove a Mailbox Was There: RFC 2342
RFC 2342 let an IMAP server disclose the naming conventions for personal, other-users and shared mailboxes. A prefix and hierarchy delimiter could tell a client how to ask the next question. They could not answer whether a particular mailbox existed, appeared in a list, accepted…
History
One Path, Four Algebras, and One Contagious Unknown: RFC 2215
A network path can be described by many numbers, but those numbers do not obey one law. RFC 2215 made that distinction operational: a break propagates by logical OR, aware hops are counted, bandwidth and packet size are narrowed by minima, and latency accumulates. More…
IETF
The Backup Nameserver Answered. The Cached Key Did Not Fit: RFC 8901
A second authoritative DNS provider can keep answering after the first one fails and still leave validating resolvers without an acceptable answer. RFC 8901 shows why DNSSEC redundancy is a synchronization contract across signers, not a count of nameservers.
History
The Packet Contained Picture and Sound. It Did Not Prove a Programme Played: RFC 2343
RFC 2343 put MPEG-2 video slices and audio frames into one RTP payload so that a stored programme could travel as one coordinated stream. The packet carried useful timing and boundary evidence. It still could not testify for the network, decoder, renderer, loudspeaker, screen or…
History
The Service Had a Name. The Name Proved Almost Nothing: RFC 2216
In 1997, the Internet engineering community gave quality-of-service offerings a namespace. But RFC 2216 made the harder point first: a service number was only an address into a specification. The proof began after the name—with the caller, the traffic contract, the element’s…
History
One Clean Match Hid the Hard Questions: RFC 2345 and the Company-Name Lookup
A company name went in; a name and a URL came back. RFC 2345 made that exchange intentionally small enough for a 1998 browser add-in and a port-43 server. The economy was the point. Yet the response line carried no proof of who registered the domain, which legal entity the name…
History
The Error Terms That Made a Promise Calculable: RFC 2212
Guaranteed Service did not promise that a network would behave like a perfect wire. RFC 2212 did something more useful: it required every participating element to put a maximum price on its departure from that ideal, then showed how those prices could be added into a delay bound…
History
The Map Was Not the Road: What RFC 2333 Let NHRP Resolve—and What It Could Never Prove
A clean NHRP cache row could look like the end of a problem. It joined an IP destination to an NBMA address, carried an authority class and retained a holding time. Yet RFC 2333 placed that row in a much longer chain. Routing had to choose a path first. NHRP could then resolve a…
IETF
The Root Queries Went Quiet. The Traffic Bill Did Not
LocalRoot can keep ordinary root questions inside a resolver and away from the Root Server System. It also creates a different network obligation: discover a publication source, check for change, fetch zone data, verify it, activate it and retreat before it expires. A four-day…
History
The Bit That Invalidated a Path: What RFC 2210's ADSPEC Could Actually Prove
RFC 2210 gave an RSVP receiver a compact account of the path in front of it. The account could include hop count, bandwidth, latency, MTU and service-specific terms. Yet the most important field was not a large number or a precise bound. It was a break bit: a one-bit warning that…
CASE FILE
The resolver address was not the resolver
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…
History
When a Successful ESRO Result Still Ends in Failure
In ESRO, a performer could finish an operation, send a successful result, and later receive a failure indication even though the invoker had already accepted that result. RFC 2188 made room for both observations to be true. Its most durable lesson is not about a forgotten 1990s…
Leaders
Adrian Farrel and the Path That Exists Only on the Controller
A positive PCEP reply can put a clean route on an operator's screen before any router has accepted it. Adrian Farrel's PCE architecture matters because it lets us say exactly what has happened—and what has not.
CASE FILE
WebAuthn’s Remote Client Can Validate the RP ID. The Local Permission Cannot Prove It Did
The first public draft of WebAuthn Level 4 proposes a careful handoff for remote-desktop sign-ins: a local browser may use client data supplied by a remote machine, but only after a per-origin grant. That grant opens the local capability. It does not preserve evidence of the…
Leaders
Cristiano Amon and the Missing Receipt in Qualcomm's Oryon Story
Qualcomm closed its Nuvia acquisition in March 2021, put Oryon inside Snapdragon X Elite in October 2023 and reached retail shelves with a multi-OEM PC wave in June 2024. A court ruling arrived later. Each event strengthens Cristiano Amon's strategic case, but none proves the…