Summary

  • The federal definition of a communications provider is functional, turning on what an entity provides for a fee to unaffiliated parties rather than on what it calls itself, and the companion 911 outage trigger is written at the E-911 tandem and selective-router layer.
  • Intrado's own published description places its routing and location work at that same layer, delivering emergency calls to the public safety agency with location information, yet no document in the reviewed file states whether it meets the provider definition or obliges it to file.
  • The reviewed file contains no outage report, enforcement instrument, state authority report or corrective-action record naming the company, which is a dated statement about this file and not a finding that no failure occurred.
  • A restored call path shows recovery; it does not show repair. A settlement shows a penalty was collectable; it does not show the fault was fixed.

A call fails in segments. A handset loses its network, an originating switch misroutes, an aggregation platform loses its location lookup, a router hands the call to the wrong public safety answering point, premises equipment drops it at the last yard. To the person holding the phone every one of those failures looks the same: silence, then a busy tone, then a decision about whether to dial again. What differs is who is obliged to write it down, and whether the instrument that compels the writing is pointed at the segment that broke.

Intrado Life & Safety, Inc. describes itself as working in the middle of that chain. The federal outage rules define a 911 emergency in the vocabulary of the same layer. Between those two statements sits a question the reviewed record does not answer: whether the party performing the routing is the party obliged to file.

What the instrument says, and what it only appears to say

47 CFR Part 4 is the federal rule set for disruptions to communications, and its definitions are where the accountability questions start. As the text makes available to this review puts it, a communications provider "is an entity that provides for a fee to one or more unaffiliated entities, by radio, wire, cable, satellite, and/or lightguide: two-way voice and/or data communications, paging service, and/or SS7 communications."

That is a functional test. It does not ask what a company calls itself, which box appears on a contract, or which logo a PSAP sees on a screen. It asks what the entity provides, to whom, and for what consideration.

The same material defines the event at a similarly specific altitude. An outage that potentially affects a 911 special facility occurs whenever there is "a loss of communications to PSAP(s) potentially affecting at least 900,000 user-minutes," or "a loss of 911 call processing capabilities in one or more E-911 tandems/selective routers for at least 30 minutes duration."

Read the second limb twice. The threshold is not written at the handset, not at the tower, not at the dispatch console. It is written at the tandem and the selective router — the machinery that decides which answering point receives a call and carries the location data that tells the call-taker where to send help. That is precisely the segment a routing vendor occupies.

One caution belongs here, because it is the difference between an argument and a quotation. The procedural side of Part 4 — notification windows measured in minutes, an initial report due within hours, a final report weeks later said to carry a root cause and a sworn attestation by someone authorised to bind the provider — is described in a secondary summary of the part rather than in the definitional text quoted above. This article therefore reports that timetable as reported and does not put it in the rule's mouth.

The distinction matters for the argument that follows: an obligation to state a cause is exactly what would make a filing usable as evidence of repair, and it is the element this review could not verify from operative section text.

A call has a chain of custody, and a duty attaches to one link

Trace a single 911 call. A handset originates it. An originating network carries it. An aggregation and routing layer resolves the location and selects the answering point. A PSAP's premises equipment accepts it. A dispatcher's screen displays it. Five parties, five sets of logs, at least four handoffs where a timestamp, an event identifier and a next-hop acknowledgement either exist or do not.

Now overlay the rule. The event definition reaches the routing layer; the duty attaches to whoever satisfies the provider definition. Those two descriptions need not name the same party, and nothing in the reviewed record establishes that they do.

This is not a technicality. A reporting duty without technical control produces an obligation to explain a failure the obligated party cannot independently diagnose. Technical control without a reporting duty produces operational leverage with no corresponding public exposure. When a multi-party call path fails, the surviving record is not a neutral sample of what happened; it is a sample biased toward whichever segment was required to speak. Buyers of routing capacity — carriers, state 911 authorities, PSAPs — plan against that record.

If it is systematically complete about the regulated segment and silent about the segment that actually failed, then reliability is being measured where it is documented rather than where it is produced.

What the company says it does

Intrado's own service-provider page describes the work in plain terms: the company connects "people, devices, sensors, and networks to emergency assistance with trusted routing and location, text messaging, and data services," and states that its "Routing and Location Solutions ensure emergency calls are delivered to the appropriate public safety agency with accurate location information."

That is a company describing, in its own published words, the delivery of emergency calls to the answering point together with location information. Put beside the rule's event definition, it places the company's self-described function inside the segment the 911 trigger names. It does not place the company inside the provider definition — that determination turns on facts about fees and unaffiliated counterparties that no document in this review supplies — and this article does not make it.

What the pairing does establish is that the question is live rather than hypothetical. A firm that routes emergency calls for a fee to carriers, authorities and enterprises is the kind of entity the functional test was written to capture. Whether it is captured, and whether it files when the router fails, are answerable only from records this review does not contain.

What the reviewed file does not contain

As of 18 October 2026, the review set for this article contained two retained public documents: the definitional text of 47 CFR Part 4 at the Electronic Code of Federal Regulations, and Intrado's own service-provider description. No initial or final communications outage report naming the company was located in that set. No enforcement instrument, consent decree, forfeiture order, bureau advisory, state 911 authority report or PSAP statement concerning the company was located in that set. No record of a named corrective action, its implementation, independent testing or recurrence monitoring was located in that set.

And no synchronised incident timeline, log extract, configuration state, alarm history or handoff acknowledgement for any 911 call path involving the company was located in that set.

That paragraph describes a file, not the world. It is not evidence that no outage occurred, that no filing was made, or that anyone failed in a duty. Public records exist outside this review set, and a document not in front of an author is not a document that does not exist. The honest reading is narrower and still useful: the material available here cannot carry the weight of a claim about this company's conduct, only a claim about what the accountability instrument is built to capture.

What recovery and what repair actually look like

When a call-routing platform fails, the observable sequence is brutally simple: calls stop completing, engineers respond, calls start completing again. The restoration is a fact. It is also the weakest possible evidence of repair, because a restored call path reports only that traffic flows now.

A defensible finding of durable repair needs more. It needs documented fault isolation — a stated account of which component failed and how that was determined. It needs a controlled change, with pre-change and post-change evidence at each affected handoff, not only at the edge where the failure became visible. It needs independent or cross-system validation, because a system confirming its own restoration is not corroboration. It needs a stated observation window long enough to cover relevant load and an actual failover, since many routing faults hide until the next peak or the next maintenance event.

And it needs preserved recurrence evidence, so a second occurrence of the same failure mode can be recognised as a second occurrence rather than greeted as a novel incident.

A settlement, if one exists, answers a different question. A penalty that has been assessed and paid proves the penalty was collectable. It does not prove the fault was repaired, and conflating the two is how a regulatory record gets mistaken for an engineering one.

Why the distinction has consequences

The practical stake is not the reputation of one vendor. It is the reliability of a service that people reach for in the worst minutes of their lives, and the ability of the institutions that buy it to know whether it is holding.

If the routing layer is inside the rule's event definition but outside its reporting duty in practice, several things follow quietly. Outages in that layer can be resolved operationally and leave no public document proportional to their reach. Procurement decisions then rest on vendor assurances and uptime figures rather than on filed accounts of what failed. Regulators looking for patterns see reports from the segments that must file, which is not the same as the segments that fail. And the absence of complaints gets read as the presence of reliability.

None of that requires anyone to behave badly. It requires only that the duty and the control sit in different places, which is the ordinary condition of a layered technical service.