Tierranet’s Three /24s Reach Three Upstreams—but Puerto Tejada’s Exit Is Unmapped

Summary

• Three publicly announced IPv4 blocks from Tierranet’s AS269746 are presently associated with three different adjacent networks, a useful form of logical separation that does not establish separate fibre entrances, ducts, buildings, power feeds or metropolitan routes.
• Colombian programme records place Tierranet in a substantial fixed-access build across northern Cauca and connect part of that work to INTERNEXA service nodes, yet the public documents do not reveal the company’s actual trunk geometry, lit throughput, oversubscription, backup power or restoration performance.
• The decisive resilience test is therefore physical: customers benefit from the three routes only to the extent that Tierranet can keep access electronics powered, isolate a cut, reach spares and reroute traffic beyond any shared local bottleneck.

Three Routes, One Physical Question

Tierranet’s public routing picture has a satisfying symmetry. The company originates three IPv4 /24s, and a current view of those routes associates each block with a different network immediately upstream. In mid-July 2026, RIPEstat’s announced-prefix view lists 45.181.204.0/24, 45.181.205.0/24 and 45.182.41.0/24. Its neighbour view identifies AS262186, AS264646 and AS272942 on the side from which the wider internet can reach Tierranet. The accompanying routing-status view counts 768 routed IPv4 addresses, three visible IPv4 prefixes and three observed neighbours. All three facts describe the public internet at the same moment. None describes a trench.

The distinction matters because routing diversity and physical diversity answer different questions. Routing data asks whether several autonomous networks can propagate Tierranet’s destinations. Physical evidence would ask whether the circuits enter different rooms, cross different bridges, use different poles, terminate on separate equipment, draw from independent power and leave Puerto Tejada along paths that cannot be severed by one excavation or collision. Three names in a routing table can coexist above one vulnerable span.

Conversely, two fibres in different ducts can still converge on one unprotected router or upstream aggregation site. A robust design needs both kinds of separation, and the public view verifies only the first.

Prefix-level observations sharpen the logical picture without resolving the physical one. RIPEstat’s distributed view of 45.181.204.0/24 includes paths that approach the Tierranet address space through AS262186. The corresponding view of 45.181.205.0/24 includes AS272942, while the view of 45.182.41.0/24 includes AS264646. These are observations from internet vantage points, not contract schedules. They can change with route selection, filtering, maintenance and the monitors available at a particular hour. They do, however, indicate that Tierranet is not simply announcing all three blocks along one visibly identical path.

That is already operationally useful. A fault confined to one external routing relationship may leave the other prefixes reachable. Tierranet can potentially shift announcements or customer traffic if its equipment, commercial arrangements and remaining links permit it. Yet an address block is not a customer circuit, and the survival of one prefix does not guarantee that every household can be moved onto it. Customer addressing, network segmentation, capacity headroom and the location of the fault all determine whether logical alternatives become usable service.

The headline question is therefore deliberately narrower than a verdict on Tierranet’s reliability: where do the three exits become physically distinct? The public material does not identify the relevant building, handhole, pole line, road crossing, river crossing, splice enclosure or upstream handoff. It does not state whether all three carriers reach Tierranet over separate last-mile construction or over shared facilities. It does not reveal whether a single local power event would darken every border session. The three routes are evidence of a considered internet edge.

They are not evidence that Puerto Tejada has three independent ways out.

The Company Behind AS269746

The legal and numerical records establish a firmer boundary around the operator than its sparse public-facing material does. LACNIC’s AS269746 registration identifies Tierranet S.A.S. as the holder of the autonomous-system number, gives Puerto Tejada in Cauca as the registrant location and records a direct allocation dating from August 2019. The related Tierranet registrant record provides a Puerto Tejada address, names the legal representative and links the organisation to its number resources. Those records support a specific proposition: Tierranet is not merely a sales name attached to someone else’s public routing identity. It controls the registered identity that originates the network under examination.

The address holdings reinforce that boundary. LACNIC records 45.181.204.0/23 and 45.182.41.0/24 as Tierranet IPv4 resources. It also records an allocated 2803:3960::/32 IPv6 block. Ownership of number resources is not ownership of every cable over which they travel, but it gives the company a portable internet identity. It can originate its own prefixes and form routing relationships under its own autonomous-system number rather than appearing only as a hidden retail customer of a larger provider.

Corporate continuity requires more care. A historical Colombian regulatory provider listing includes “Tierranet Ltda.” with the tax identifier 900068083. A May 2026 MinTIC expression-of-interest list names “TIERRANET S.A.S.” with the same identifier. Taken together, the documents indicate a longstanding provider identity followed by the present corporate form. They do not, by themselves, explain every corporate change between the two dates, and the article does not assume that every old asset or obligation passed unchanged into the current company.

The company’s current public website confirms an active public presence in July 2026, but offers little engineering disclosure. It does not publish a network diagram, point-of-presence list, upstream capacity schedule, restoration history or power-reserve statement. The older tierranet.net.co address displays an expiry notice rather than the operator’s service pages. A lapsed web domain is not a lapsed access network: DNS, hosting and fibre operations are separate things. The contrast is useful chiefly because it warns against treating an old marketing address as a current operational authority.

The boundary that can be defended is thus precise. Tierranet S.A.S. is the registered holder of AS269746 and named address space, a Colombian communications provider active in recent public programme records, and a company publicly associated with fixed-internet work in Cauca. The evidence does not show that Tierranet owns every pole, duct, feeder, building, long-haul circuit or service node involved in that work. Some assets may be leased, shared or supplied under programme arrangements.

That distinction becomes especially important when assigning responsibility for a failure: the customer sees Tierranet, while restoration may require action by a fibre owner, an upstream network, a utility, a municipal authority or a field contractor outside Tierranet’s corporate perimeter.

What Public Records Place in Northern Cauca

Government records provide the strongest public evidence of Tierranet’s access footprint, though they describe programme obligations and reported progress rather than a complete commercial network. In the 2023 “Conectividad para Cambiar Vidas” process, the proposal-receipt record shows multiple Tierranet submissions received on 29 November 2023. The later final evaluation records legal and technical compliance and identifies allocations for Puerto Tejada, Santander de Quilichao and Villa Rica. The counts attached to those allocations—2,006, 2,290 and 922 respectively—describe eligible household connections in that call, not the number of active retail subscribers and not simultaneous network throughput.

A later official response gives a broader and more granular snapshot. The Colombian government’s response to Congress lists Tierranet against six Cauca municipalities under the 2023 call: Puerto Tejada, Santander de Quilichao, Villa Rica, Padilla, Corinto and Caloto. Across the six entries, the table assigns 3,888 connections, records 3,368 as reported and 1,648 as approved at the date represented in the document. Those categories should not be collapsed. “Assigned” is an obligation or programme quantity; “reported” is a claim of work submitted into oversight; “approved” is the subset accepted at that stage. None is automatically equivalent to a paying, connected household on 17 July 2026.

The same response separately reports the newer call. For the three municipalities named in the final evaluation, it shows 5,218 assigned connections, only 45 reported at that snapshot and none yet approved. That timing gap is not proof of abandonment or failure. It indicates that the second group was at a much earlier stage when the information was compiled. It also illustrates why infrastructure claims require dates: a programme table can be accurate and still be a poor description of today’s installed base.

The geography is nevertheless meaningful. Puerto Tejada, Villa Rica, Padilla, Caloto, Corinto and Santander de Quilichao form a northern Cauca operating surface rather than a nationwide abstraction. Service there depends on distribution fibre or wireless links that reach neighbourhoods and premises, aggregation points that collect those access links, and an onward connection to larger networks. The public programme documents refer to fixed residential access and, in the 2023 evaluation material, to FTTH submissions. They place Tierranet close to the physical work of passing and connecting homes.

The Findeter contract page adds another part of the operating burden. Its contract description covers planning, installation, commissioning, operation, maintenance and support of residential fixed internet in regions 28 and 30, using wired or wireless technologies. It gives a contract amount of COP1.681 billion. The description matters more than the generic geography shown elsewhere on the page: Tierranet’s responsibility was not limited to delivering boxes. It extended into keeping service working and supporting households during the contract period.

What remains absent is a route-level map tied to Tierranet. Municipality names do not reveal which roads the backbone follows, whether fibre crosses the Cauca River or other waterways, which bridges or poles it uses, where aggregation takes place, or which parts are radio-fed. Counts of households say nothing about feeder length, splitter depth or the number of customers behind one cabinet. The records therefore locate the consequences of a failure better than they locate the failure domains themselves. They tell us which communities could depend on Tierranet’s work.

They do not tell us which single cut, flood, collision, fire or substation event could affect several of those communities at once.

The INTERNEXA Layer Beneath the Access Footprint

One programme requirement brings a larger carrier into view. A published response concerning the evaluation discusses the need for proposed household coverage to connect through an INTERNEXA service node. It also records an evaluator’s concern that another bidder had submitted the same coverage information as Tierranet for Puerto Tejada, prompting questions about whose mapped network was being presented and what relationship existed between the parties. That exchange is evidence of scrutiny, not evidence that Tierranet misrepresented an asset. The final evaluation selected Tierranet, and the public documents do not expose the underlying commercial agreements or the submitted geographic files in a form that would settle ownership.

Number-resource records corroborate a working interface with INTERNEXA more directly, but still at a different layer from fibre ownership. LACNIC shows 179.1.123.48/28 as address space reallocated to Tierranet in June 2024. The parent 179.1.0.0/17 record belongs to INTERNEXA S.A. E.S.P. Tierranet’s registrant record lists several such small reallocations. These blocks can support managed interfaces or services. They do not establish that INTERNEXA is one of the three present adjacent networks visible for AS269746, and they do not show the bandwidth, physical medium or route of any circuit.

The congressional response supplies geographic context for the public programme. It says eight INTERNEXA service nodes in Cauca entered service on 19 March 2024 and that 4.4 kilometres of new fibre in the department were completed on 21 June 2024. That is meaningful regional infrastructure. Yet the document does not say that all 4.4 kilometres belong exclusively to Tierranet, identify the exact Tierranet-facing segment, or trace a continuous route from a Tierranet access node to each service node. The nodes create potential aggregation points; the missing link is the physical chain between those points and the households.

This layered arrangement explains why the word “network” can conceal several owners. Tierranet can own customer-facing electronics and distribution fibre, lease transport to a service node, receive addresses for a managed connection, and originate its public prefixes through other adjacent autonomous networks. A programme administrator can fund connections while a carrier supplies regional transport and a local contractor maintains a span. Each party may control a different repair clock.

A damaged drop cable is a different event from a severed municipal feeder, a failed access switch, a dark service node or a route withdrawn by an external carrier.

It also explains why INTERNEXA should not casually be labelled a fourth present upstream for AS269746. Publicly reallocated address space and programme-node use demonstrate a technical or contractual relationship of some kind. Current autonomous-system observations, by contrast, show three different adjacent networks. The two sets can coexist because managed transport need not appear as an AS neighbour, and an address assignment need not carry Tierranet’s public internet traffic. Treating every named infrastructure partner as an internet-transit provider would blur the very boundaries that resilience analysis needs.

The defensible conclusion is limited but valuable. Tierranet’s access obligations in Cauca were designed to meet a regional service-node layer, and public address records are consistent with a direct service relationship involving INTERNEXA. The physical handoff, capacity commitment and ownership boundary remain undisclosed. If that handoff is a common dependency for multiple municipalities, it could dominate the resilience of the access build even while AS269746 retains three external routes elsewhere. If the access regions reach separate nodes over separate paths, the design could be stronger.

Public evidence does not allow either assumption to become fact.

Allocated, Announced, Installed and Usable Are Different

Small-network capacity is often discussed with numbers that measure unlike things. Tierranet’s registered IPv4 holdings comprise a /23 and a /24: 512 plus 256 addresses, or 768 in total. The current routing view shows the /23 divided into two /24 announcements alongside the separate /24, yielding the same 768 addresses across three routes. This is address capacity and routing granularity. It is not 768 customers, 768 megabits per second, three circuits, or a guarantee that every address is in use.

Independent public summaries broadly agree on that outward shape. The Hurricane Electric BGP view presents the autonomous system and its visible prefixes and adjacencies. CIDR Report likewise shows three adjacent networks, while warning through the nature of its topology analysis that labels such as “upstream” describe observed routing relationships rather than necessarily disclosing commercial contracts. IPregistry’s AS269746 page reports three IPv4 /24s, 768 IPv4 addresses and no announced IPv6. These views make the internet edge observable from several vantage points; they do not measure purchased gigabits, packet loss under load or customer demand at the evening peak.

The IPv6 contrast is especially instructive. Tierranet holds a /32, an enormous logical space suitable for many customer subnets. Current public route observations show no IPv6 prefix originating from AS269746. “Allocated” therefore means the resource is assigned to the company, while “announced” means the wider internet can presently find a route to it. Even an announcement would not prove that residential access, support systems and customer equipment can deliver usable IPv6 end to end. The gap warns against reading a registry allocation as deployed service.

Another independent measurement demonstrates how observation windows differ. The CAIDA AS Rank record attributes three provider-side adjacencies but, in the returned snapshot, counts two prefixes and 512 IPv4 addresses rather than three and 768. That discrepancy need not mean either system is wrong. Route collectors see different peers, refresh on different schedules and apply different aggregation rules. It does mean a single count should carry its observation date and method, particularly when the network is small enough that one prefix changes the apparent total by a third.

Installed capacity is harder. The programme records count connections assigned, reported or approved, but publish no Tierranet circuit rate, committed information rate, oversubscription ratio, access-port count, optical split ratio, radio-channel plan, router forwarding limit or peak traffic curve. They do not disclose how much external capacity is reserved on each of the three adjacent networks or whether one link could carry the normal load of all three prefixes after a failure. A backup path that remains reachable but saturates immediately is technically alive and commercially painful.

Usable capacity adds still more conditions. Fibre must be lit, terminal equipment powered, software configured, optics within budget, external circuits paid and field faults repaired. Capacity must exist in the right place: spare headroom at one edge does not help customers stranded behind a dead aggregation node. It must also exist at the right time, because a planned alternative that requires manual reconfiguration can leave a long interruption before it becomes useful.

For Tierranet, the public record supports a clear capacity statement and an equally clear limit. The company currently has 768 visible IPv4 addresses split into three /24 announcements and maintains three observed external routing relationships. Public programme records associate it with thousands of household connection obligations. No public document reviewed here quantifies the throughput installed between those households, the access aggregation layer, the regional service nodes and the three external neighbours. The former facts are hard numbers. Turning them into a claim about available bandwidth would be invention.

Why Three Upstreams Can Share One Failure

The three adjacent networks are identifiable. LACNIC registers AS262186 to TV Azteca’s Colombian branch, AS264646 to Dobleclick Software e Ingeniería, and AS272942 to Global Raices S.A.S. Their records place the organisations in Colombia and provide administrative contacts. Those addresses are not network locations. A registrant’s office in Bogotá, Popayán or Guachené does not reveal the handoff site, the path to Tierranet or the ownership of the intervening fibre.

The observed arrangement nevertheless has plausible benefits. Prefix-specific propagation can contain some external incidents. If one neighbouring network withdraws Tierranet’s route or loses reachability beyond its own edge, the other two prefixes may continue to be seen. Tierranet may be able to change announcements, provided its border equipment and agreements allow it. Separate neighbours also reduce dependence on one carrier’s routing policy and can improve the options available during maintenance or congestion.

But several shared dependencies can sit below those distinct names. The three services could enter the same Tierranet room through one duct. Two providers could lease strands in the same cable. All three could cross the same bridge before separating. One wholesale carrier could supply hidden transport to more than one named neighbour. The handoffs could terminate on separate routers plugged into one unprotected power strip, or on one router with three interfaces. Even genuinely separate paths could converge at the same upstream building.

None of these arrangements is established for Tierranet; each is a common enough possibility that AS-level diversity cannot exclude it.

The absence of a public exchange record adds uncertainty without proving isolation. PeeringDB’s AS269746 query returns no listed network. That means the voluntary service has no discoverable entry for the number at the time checked. It does not mean Tierranet has no private peering, no shared facility and no exchange connection. Many small operators disclose little, and private handoffs do not require a public profile. The result removes one potential way to verify facilities and interconnection points; it is not negative proof.

Physical diversity also has a scale problem. A route can be diverse at the national level and common at the last kilometre. Tierranet might buy services from carriers with entirely different backbones, yet reach both carriers through one local access cable. Or the local entries could be separate while both upstreams converge on one regional corridor exposed to landslide, flood or civil works. The relevant scale depends on the event being considered. Protection against a failed border port is easier than protection against a severed bridge crossing; protection against a citywide electrical disturbance requires another design again.

The 3-to-3 prefix pattern should therefore be read as an engineering clue, not a resilience certificate. It suggests intentional distribution: each /24 is visible through a different adjacent network in the sampled routes. That may simplify fault isolation and give the operator more levers during an external incident. Yet it provides no reason to assume that a household on one /24 has a physically independent path from a neighbour on another, or that Tierranet can renumber or reroute customers quickly.

A strong claim would require Tierranet or its carriers to show where the circuits hand off, the fibres’ ownership and duct paths, their shared-risk groups, the buildings and power systems they traverse, and the capacity each retains after another fails. Until then, three upstream names accurately describe logical reachability. “Three physically independent exits” would go beyond the evidence.

Power and Facilities Sit Outside the Routing View

Every route visible on the internet rests on equipment that consumes electricity in a place. For a local fixed network, that chain may include customer terminals, powered radio sites, optical line terminals, aggregation switches, border routers, cooling, security systems and the upstream carrier’s handoff equipment. A route collector can continue to display a recently observed path without revealing whether the customer-facing electronics have enough reserve power to survive the next outage. It cannot show a depleted battery or a generator that cannot be refuelled.

No reviewed Tierranet material states the location or protection class of its principal network facility. The LACNIC address in Puerto Tejada is an administrative address and should not be promoted into a point of presence without corroboration. The government records name municipalities and service nodes but do not locate Tierranet’s border routers, optical terminals, warehouses or field depots. The public website offers no facility inventory. This missing geography prevents a basic shared-risk assessment: it is impossible to determine whether the public internet edge, access aggregation and support operation occupy one site.

Power is similarly opaque. There is no disclosed utility feeder, battery autonomy, generator rating, fuel arrangement, transfer-switch design or load priority. The absence of those data is not evidence that protection is absent. Small providers often operate backup systems without publishing them. It does mean an outside analyst cannot convert “three neighbours” into “service survives a local blackout.” If all external links terminate beside the same unpowered switch, their upstream diversity becomes irrelevant to customers until power returns.

The household end has its own dependency. Fibre can remain intact while a customer’s optical terminal and Wi-Fi router lose power. A programme can install a technically sound FTTH link that the household cannot use during an area outage. Business or public-service users may have local backup, while residential users often do not. This is one reason outage reports must distinguish access-network failure, customer-premises power loss and upstream loss rather than treating every loss of connectivity as the same event.

Facilities also shape environmental risk. Northern Cauca networks may use aerial plant, buried routes, road verges, bridge approaches and shared utility corridors. Without route geometry, it is not possible to assign Tierranet exposure to any particular floodplain, crossing or construction zone. The municipality list establishes broad service geography but not the vulnerable segments within it. A photogenic map of towns joined by straight lines would imply routes that the public evidence does not show.

There are still useful deductions, if clearly labelled. A network serving several municipalities must aggregate traffic somewhere, and its external routes must meet physical handoffs somewhere. Those necessities create a finite set of facilities even though their number and locations are unknown. Maintenance obligations imply access to spares and field labour, but not the staffing level or travel time. Three routing relationships imply at least three logical interfaces, but not three chassis or three buildings. These are structural inferences, not claims about Tierranet’s disclosed design.

Power and facility disclosure would substantially improve public understanding without requiring publication of security-sensitive detail. Tierranet could report whether border and access functions share a site, give ranges for battery autonomy, state whether generators are tested, identify how many independently powered aggregation locations exist and disclose whether carrier entrances share a duct. The company could describe resilience classes rather than exact coordinates. In the present record, the responsible position is neither to assume fragility nor to award resilience.

It is to identify power and co-location as the unmeasured dependencies that sit between a healthy routing table and a customer’s working connection.

Recovery Depends on People, Spares and Access

Resilience is revealed after protection fails. A provider must detect the event, identify the layer, reach the site, obtain access, find the damaged span or device, secure replacement material, coordinate with another owner if necessary, restore traffic and then confirm that customers have actually recovered. The speed of that sequence is shaped by local support labour as much as by route design. A spare fibre path that requires a manual patch in a locked building is only as fast as the person with the key.

Public contract material makes maintenance part of Tierranet’s role but does not publish the company’s operational resources. A September 2025 MinTIC management report identifies Tierranet’s contract for regions 28 and 30, with a start date of 27 November 2024, a stated end date of 26 January 2026 and a value of COP1,681,607,847. At the report’s snapshot, one region had a target of 601 connections, 592 approved and eight under review; the other had a target of 600, 451 approved and 25 under review. The report also noted that a maintenance plan version remained under correction and recorded a financial enforcement item associated with installation.

Those entries should be kept in proportion. They are oversight facts from a dated programme report, not a direct measurement of Tierranet’s current repair performance and not evidence of a present customer outage. The near-target approved counts in one region and the lower count in the other show differing progress at that moment. A plan under correction shows that maintenance documentation had not yet cleared every review. It does not show whether technicians responded quickly to an actual cut, how many vehicles or fusion splicers the company held, or whether spare optics were stocked locally.

Colombian communications rules establish the customer-facing stakes. CRC’s 2022 quality resolution treats achieved transmission speed and one-way delay as relevant fixed-data measures and provides for geographic reporting and incident obligations within its scope. A later CRC consumer-protection explanation discusses compensation when service is unavailable beyond specified thresholds, subject to the regulation’s conditions, and addresses questions involving fibre theft and power failures. These rules do not prove that Tierranet has breached a standard. They show why restoration records, not route counts alone, determine the practical value delivered to users.

Different fault types demand different recovery chains. A failed border session may be resolved remotely by changing routing or contacting a neighbouring network. A router hardware failure requires a redundant unit or a spare and someone able to install it. A cut feeder requires location, permits or safe access, a crew, cable and splicing. Damage on another owner’s plant adds an organisational handoff. A widespread power event may require batteries, generators and fuel at several sites. The slowest dependency sets the customer’s outage.

Local knowledge can be an advantage for a regional provider. Technicians who know the pole routes, landowners, cabinets and municipal access procedures may reach faults faster than a distant national desk. The same small scale can concentrate knowledge in a few people and spares in one depot. The public record offers no staffing roster, service-level history or inventory from which to judge which effect dominates at Tierranet.

The most useful evidence would be aggregated and retrospective: median and high-percentile restoration times by fault class; the share of incidents on Tierranet-owned versus third-party plant; spare-equipment locations; after-action summaries for major cuts; and the proportion of external capacity available during maintenance. Such disclosures would link the physical network to human response without revealing sensitive site coordinates. In their absence, Tierranet’s maintenance obligations are established, while its demonstrated recovery capability remains unresolved.

Who Carries the Cost of an Outage

The customers implied by the public programmes are not abstract endpoints. They are households in municipalities where fixed access can support schoolwork, remote services, job searches, payments, communications and small businesses. The consequence of a failure depends on timing and alternatives. An evening capacity shortage may degrade video and study without producing a clean outage. A daytime feeder cut can interrupt commerce. A prolonged failure may force users to buy mobile data, travel for connectivity or abandon a task entirely.

Programme counts offer scale but must be used carefully. In the six-municipality group under the 2023 call, 3,888 connections were assigned to Tierranet, 3,368 were reported and 1,648 had been approved in the cited government response. In the three-municipality 2024 group, 5,218 were assigned, 45 reported and none approved in that same snapshot. The populations overlap geographically, and the categories describe programme states rather than unique active customers. Adding all of them and calling the result Tierranet’s subscriber base would double-count concepts and misstate what the tables measure.

Even the approved count is not a live-service counter. Approval can confirm an installation against programme requirements at a particular stage. It does not state whether the household remains subscribed, whether the line is currently lit, what speed it receives or how often it fails. Conversely, Tierranet may have commercial customers outside the programmes who do not appear in those tables. The public numbers define an exposure range and a public-interest obligation, not a complete retail census.

Costs also fall unevenly along the network. A household may receive a regulatory adjustment while still bearing the time cost of disconnection. Tierranet may lose revenue, incur crew and material expense, or face programme consequences. An upstream or infrastructure owner may be responsible for repair under a contract that is not public. Government administrators bear the risk that connectivity targets are counted as delivered without remaining consistently usable. The surrounding municipalities bear a broader economic cost when many users lose access together.

Capacity shortfalls can distribute harm less visibly than cuts. If one of three external links fails and the remaining links lack spare throughput, the network may remain reachable while applications become unreliable. Customers with lower-quality Wi-Fi or longer access paths may experience the degradation first. Basic messages can work while video calls fail, creating a confusing incident in which operator dashboards show connectivity but users experience practical loss. This is why post-failure usable throughput matters more than the simple survival of a route.

The public-service context heightens the need for honest status labels. “Assigned”, “reported” and “approved” should remain distinct in public reporting. Installed equipment should be separated from activated service; activated service from sustained use; address holdings from bandwidth; a contract’s monetary value from capital invested in any particular route. Each distinction prevents a real achievement—such as connecting a home or obtaining a second external route—from being stretched into a claim it cannot support.

For affected users, the decisive metric is service experienced over time. That includes speed during busy periods, latency, frequency and duration of interruptions, repair communication and the ability to obtain help locally. Tierranet’s three upstream paths can contribute to that experience, but only as part of a chain that begins at the premises and passes through access plant, aggregation, power, transport and support. The public record is strongest at the two ends: it names communities and shows internet routes. It is weakest in the middle, where most shared failures and recovery delays occur.

Where the Evidence Conflicts or Stops

The evidence is unusually revealing for a small private operator, but it is not seamless. The clearest discrepancy concerns the company’s tax identifier. The historical provider list and the May 2026 interest document give 900068083. The January 2024 proposal-receipt document prints 900068093-9 for Tierranet entries. One digit differs. Because the former number appears across records separated by many years and is attached to the current corporate name in the newer official list, the proposal document’s number may be a transcription error.

Public material reviewed here does not provide enough corporate documentation to settle it, so the discrepancy should be preserved rather than silently corrected.

Routing counts also differ by observer. Current RIPEstat and IPregistry views show three /24s and 768 IPv4 addresses, while the cited CAIDA response reports two prefixes and 512 addresses for its snapshot. CIDR Report and Hurricane Electric provide their own collected paths. Small networks are particularly sensitive to collector reach: one missing route changes the total sharply. The correct response is not to average the counts. It is to prefer dated, multi-vantage observations for the present description and retain the conflicting result as a warning about visibility.

The registered IPv6 /32 and the absence of a visible IPv6 announcement are not a conflict. They measure different stages. An allocation can be held before public origination, used in limited tests or left undeployed. No reviewed material establishes which explanation applies to Tierranet. It would be equally unwarranted to call the block active residential IPv6 capacity or to declare it abandoned.

Programme documents contain another apparent tension. The 2023 final evaluation assigns 5,218 connections across Puerto Tejada, Santander de Quilichao and Villa Rica in the newer group, while the later congressional response shows only 45 reported and none approved at its snapshot. That is best understood as a timing and status distinction. The same response reports much greater progress for the earlier six-municipality group. It should not be read as a single cumulative construction total without preserving the call, date and status of each line.

The evaluator’s question about identical Puerto Tejada coverage information submitted by Tierranet and another bidder is a genuine unresolved issue, but the record stops before the underlying network relationship is visible. The public exchange asks who owned the geographic information and what arrangement connected the bidders. The final selection supplies an administrative outcome, not the requested asset map or contract. It would be irresponsible to infer fraud, common ownership or sham competition. It is equally important not to erase the question, because ownership and access rights determine who can repair and expand the network.

Most consequentially, no reviewed record supplies route geometry or capacity. There is no Tierranet fibre map, shared-risk register, handoff schedule, circuit inventory, peak-traffic series, power-reserve statement, restoration log or customer-performance history. Public autonomous-system data cannot fill those blanks. Government household tables cannot fill them either. Each provides a different, valid view of the operation.

These limits shape the article’s confidence. Confidence is high that Tierranet controls AS269746 and the named address resources, currently announces three IPv4 /24s and is visible beside three external autonomous networks. It is high that official records associate the company with fixed-access work in the named Cauca municipalities. Confidence is medium that the prefix-specific neighbours represent a stable intentional arrangement rather than a temporary observation, because several public monitors broadly agree but commercial contracts are private.

Confidence is low on physical route independence, usable failover capacity, power autonomy and restoration performance because direct evidence is absent.

What Would Prove Real Resilience

The missing proof can be described without demanding sensitive coordinates. First, Tierranet could publish a simplified physical-diversity statement for its external links: how many carrier entrances, ducts, buildings, border devices and independent power domains exist; where circuits share a risk group; and whether each surviving link has enough committed capacity for essential traffic. Carriers could attest to separation under confidentiality while Tierranet reports the result in ranges or classes.

Second, the access network needs the same discipline. A municipality-level diagram could show aggregation zones, the number of independently fed nodes, major route pairs and whether each zone has a secondary path. Exact pole numbers and splice locations need not be public. The key is to distinguish ring from spur, owned plant from leased plant, fibre from radio, and a truly alternate path from a cable that returns to the same crossing.

Third, installed and usable capacity should be reconciled. Tierranet could report external committed capacity, peak utilisation bands and the headroom available after losing the largest link. At the access layer, it could state active versus installed ports, typical optical split ranges and the share of programme connections presently active. The already public distinction among assigned, reported and approved connections should remain intact. IPv6 could be described separately: allocated, publicly announced, available to access customers, or not yet offered.

Fourth, power and recovery deserve measurable claims. Useful disclosures would include battery-autonomy bands for major nodes, generator coverage, testing frequency, fuel arrangements and the percentage of equipment monitored for power state. Restoration reporting could separate customer drops, feeder fibre, active equipment, utility loss and third-party transport. Median time alone is limited public evidence; a high-percentile figure would show how long difficult incidents last. Major events should receive short after-action accounts explaining what failed and what changed.

Fifth, contractual boundaries should become legible. The INTERNEXA relationship can be described in terms of function—service-node access, managed transport or another service—without publishing prices. Tierranet could state which party maintains each segment and how escalation works outside business hours. Equivalent clarity is needed for the three autonomous-system neighbours: a route observed through a network is not necessarily a direct paid-transit circuit, and the physical provider may differ from the routing neighbour.

Finally, public performance should connect technical evidence to users. Municipality-level measures of achieved speed, latency, interruption duration and complaint resolution would show whether network design produces dependable access. They should be dated and accompanied by sample size and measurement method. Results during a major external-link failure would be especially informative because they reveal whether the remaining routes offer practical capacity rather than symbolic reachability.

Until such evidence appears, Tierranet deserves neither the easy dismissal reserved for a one-route local provider nor the confidence granted to a demonstrably diverse regional network. It has a real autonomous-system identity, its own address space, three currently visible IPv4 routes through three adjacent networks, recent public programme responsibilities and signs of integration with a larger regional service layer. That is a more developed operating surface than a simple reseller’s.

The unresolved issue is the one hidden below every route advertisement. If all three external relationships leave Puerto Tejada through a common cable, room or power domain, the internet can display diversity until the instant that shared dependency fails. If Tierranet has separated those elements and retained enough failover capacity, the same three routes could provide meaningful protection to households across northern Cauca. The difference cannot be read from the three /24s. It lies in fibre paths, powered facilities, repair access and the people who restore them.