Time Horizon
Multi-year
Within the Time Horizon facet, Multi-year time-horizon intelligence organises articles by the period over which a signal is expected to matter. The page helps readers distinguish immediate operational changes from longer-cycle governance, investment, standards, and infrastructure shifts that may unfold across quarters or years. It connects timing assumptions with public evidence, related actors, market context, customer exposure, policy pressure, and infrastructure planning so readers can judge whether a development is urgent, strategic, or still waiting on confirming evidence. The page also explains how time horizon changes the meaning of a signal, which organisations may be exposed, and which infrastructure decisions require short-term action or long-cycle monitoring.
Global Regional ISP Trends
A Second Remote Peering Service Is Not a Second Path
Remote peering can put an exchange within logical reach without putting a second physical route under the buyer’s control. Resilience begins by tracing each service through every handoff, not by counting green entities in a portal.
History
The Session Died. Route Memory Became an Explicit Exception: RFC 1267
An incremental protocol carries an unspoken dependency: both sides must remember the same beginning. RFC 1267 sent a complete routing table once, then only the changes, and required each BGP speaker to retain its peer’s current table for the life of the connection. Later BGP…
CASE FILE
The Route Key Stayed the Same. Its Meaning Changed at the Border: RFC 9871
Domain Two calls low delay C2. Domain One calls it C1. RFC 9871 lets the route cross that border without pretending the two administrations share a dictionary: the route key stays `(E2,C2)`, while a separate Local Color Mapping record carries the receiving domain's meaning.
History
The Specification Called Zero VAR. Running Code Called It VALUE: RFC 1408
One octet crossed a Telnet connection. The new specification said zero introduced a variable name. A family of existing programs read the same zero as the beginning of a value. Nothing in the packet announced which dictionary had produced it. RFC 1408 had given the exchange a…
CASE FILE
The Prefix Was Right. The Packet Took the Wrong Uplink: RFC 9872
A multihomed host can learn a perfectly valid NAT64 synthesis prefix and still send the resulting address toward the wrong provider. RFC 9872's consequential move is therefore not a new prefix format; it is to prefer a signal that can retain which first-hop router supplied the…
History
The Report Counted 56 Routers. It Refused to Call the Stress Case Normal: RFC 1266
Fifty-six was not the number of production routers. RFC 1266 first counted 56 BGP border routers in seven autonomous systems, then removed the seven-router T3 NSFNET Test Network from the operational denominator: 49 routers, six systems. That correction is the most important line…
History
The Graph Could Travel. Its Measurement History Had to Travel With It: RFC 1404
A line on a screen looks self-contained. In 1993, RFC 1404 treated it as the end of a chain: a named interface had been polled at an actual interval, its counter had been turned into a delta, the result had been mapped to a continuing network resource, and older samples may…
CASE FILE
The Registry Marked a Second Email Primary. It Still Had No Delivery Receipt: RFC 9873
RFC 9873 gives an EPP contact a second address and lets the extension call it primary. That is useful coordination, but it is easy to promote a preference flag into a claim the protocol never makes: neither address has thereby been shown to exist, receive, forward, be read or…
History
The Checklist Was Retired. The Evidence Question Moved: RFC 1264
In 1991, a routing protocol did not advance merely because its specification looked complete. RFC 1264 asked for separate receipts: independently written code, tests for every feature, security demonstrations, operational history, and an account of where the design would break.…
History
The Route Tag Could Admit It Had Forgotten the Path: RFC 1403
A border router in 1993 could squeeze a route’s history into thirty-two bits only by throwing most of that history away. RFC 1403 made the loss visible: its OSPF tag could say whether path knowledge was complete and roughly how long the path had been. When the missing detail was…
CASE FILE
The Registrar Deleted One Domain. Other Domains Inherited the Risk: RFC 9874
A registrar can be entitled to delete its own domain and still be unable to contain the consequences. RFC 9874 exposes the awkward case: a subordinate host sponsored by one client can remain the name server for domains sponsored by others, so a local disposal request reaches…
NPNOG
Two Hours Online: What npNOG’s Pandemic Experiment Changed About Access
npNOG kept a public technical programme alive on 9 December 2020 by moving it online. The official record supports a narrower achievement than the language of digital inclusion often implies: no one had to travel to a named physical event venue to join, and npNOG says more than…
History
The Relay Changed the Message. It Had to Call Itself a Gateway: RFC 1344
A carrier can claim neutrality while the entity in its custody remains untouched. The claim becomes harder after it changes an image, replaces a reference with a copy or divides one message into several. RFC 1344 gave that moment a name in 1992: the intermediary was no longer…
History
The Test Bed Could Demonstrate a Future. It Could Not Decide the Public Network: RFC 1259
RFC 1259 imagined the NREN as a new subdivision built before the surrounding city. Its mistakes could teach later builders, but its streets, prices and residents could not decide the whole metropolis. The crucial Internet-history lesson was not prediction. It was the design of a…
IETF
Eric Rescorla and the Encrypted Hello That Could Not Name the Server
An encrypted ClientHello can make a familiar public clue disappear from the first TLS flight. That does not turn the remaining handshake into an identity ledger. RFC 9849, co-authored by Eric Rescorla, specifies a carefully bounded privacy envelope: it can hide the private…
History
The Network Did Not Have to Keep the Beat. The Receiver Rebuilt It: RFC 1257
In 1991, a voice sample could arrive early, the next one late, and the listener could still hear a regular cadence—if the receiver knew the deadline, kept enough memory and woke at the right time. RFC 1257 used that construction to move a costly promise out of every subnetwork…
History
The Packet Got a New Lifetime at Every Border: RFC 1326
A tunnel is often described as a clean bargain: put one protocol inside another, cross a network that understands the outer wrapper, then recover the original packet at the exit. RFC 1326 asked what happens when two such bargains face each other and routing sends the packet back…
History
The Analysis Could Not Stand In for the Test Network: RFC 1245 and RFC 1246
In July 1991, OSPF arrived with more than a protocol specification. Its supporting case was deliberately divided into an analysis and an experience report. That division preserved a distinction Internet engineering still needs: a model can bound expected behaviour, but only…
History
The NIC Owned the Question. It Did Not Own the Network: RFC 1302
A help desk can keep faith with a user without pretending to control every system behind the problem. RFC 1302 made that distinction unusually clear in 1992. A Network Information Center was expected to take responsibility for an inquiry until it was resolved in some way. Yet…
IETF
Deborah Brungard and the Transport Profile That Did Not Choose a Network
A standards requirement can make a capability available without dictating the network that will use it. RFC 5654, edited by Deborah Brungard and four colleagues, describes requirements for an MPLS Transport Profile. Its most important restraint is plain: the requirements concern…