TELECOM3000 lists a 10Gbps NAP.EC port while UFINET precedes it in 99.7% of a July route snapshot

Summary

• TELECOM3000’s 2021 ARCOTEL title documented one initial Sangolquí node and a proposed 100Mbps fibre connection, while Surnet’s current order form offers plans up to 1Gbps; that proves expansion in the offer, not the capacity or topology now supporting it.
• NAP.EC data lists AS273177 on a 10Gbps exchange port, but a 16 July 2026 RIPE snapshot showed UFINET immediately before TELECOM3000 in 5,706 of 5,723 visible paths; exchange capacity and route visibility do not prove independent physical transit.
• The decisive unknowns are the last-mile architecture, optical split ratios, any microwave relays, route separation, backup-power runtime, busy-hour utilisation, spare stock and restoration targets, so resilience cannot be inferred from advertised access speed alone.

Noon on a roof, and the radio link the record does not prove

At noon in Quito, the light arrives almost vertically. Metal housings become too hot to rest a hand on, cloud edges harden over the Andean basin, and a technician at a rooftop parapet has to shade the alignment display while turning a point-to-point dish by fractions of a degree. Below, offices are returning from lunch, homes are filling with video calls and streams, and the network is moving toward the hours when a marginal link has the least room to hide. The dish may show a strong signal at one moment and surrender several decibels when wind moves the mount, rain thickens the path or a new obstruction intrudes into the clearance zone.

That is the right physical scene with which to test TELECOM3000’s capacity boundary. It is not, however, a documented scene from TELECOM3000’s network. The company’s 2021 internet-access title said its initial application contained no wireless networks and had not initially contemplated spectrum use. Its present retail site, operating under the Surnet name, says the access network is “100% fibra óptica”. The standard customer contract form marks fibre as the access method in its technical annex and leaves the wireless line unmarked. None of those documents rules out a microwave backhaul, a temporary relay or a later wireless extension, but none establishes one either.

That distinction matters because microwave capacity cannot be inferred from a rooftop photograph, a dish diameter or a marketing claim. The current ITU recommendation for terrestrial line-of-sight systems treats path clearance, diffraction, multipath fading, atmospheric attenuation and rainfall as engineering inputs. Channel width, modulation, licensed or unlicensed spectrum, polarization, fade margin and interference determine usable throughput. A nominal radio rate is not the same thing as stable payload capacity at the worst hour, and one visible dish does not reveal whether traffic can move through a second relay when the first path fails.

So the noon alignment is a stress test posed to the public evidence, not a claimed asset inventory. If TELECOM3000 does use microwave anywhere between a customer, a fibre handoff and an upstream node, the operational questions are exact: where are the endpoints; what spectrum and channel widths are used; what is the committed payload after protocol overhead; what fade margin remains in heavy rain; which sites have independent power; and what alternate path exists if a relay or mounting structure fails? If the network is entirely fibre, the same scene still exposes the central issue.

The dish disappears, but contention, route concentration, power and repair remain. The medium changes; the need to separate installed capacity from capacity that survives a fault does not.

The operator behind Surnet is TELECOM3000

The legal and commercial names are easy to blur because the customer-facing brand is more visible than the title holder. ARCOTEL’s 2021 resolution identifies TELECOM3000 S.A., Ecuadorian tax number 1792965667001, as the company receiving a 15-year registration to provide internet access and a framework for non-essential frequencies. It names Marcia Guadalupe Tacuri Pilicita as general manager and describes an initial service area in Pichincha. The company’s current corporate-domain landing page presents the same Surnet retail operation, and its footer identifies TELECOM3000. The contact page gives a south-Quito address at Pueblo Viejo E6-34 and Tosagua and publishes local sales and support channels.

The operator boundary is therefore reasonably firm: TELECOM3000 S.A. holds the title and Surnet or SURNET.EC is the retail identity through which it sells connectivity. That is stronger than an inference from a logo. The contract names both, gives the same tax number, and states that customer equipment remains the provider’s property. ARCOTEL’s January 2025 subscription-television permit again identifies TELECOM3000, the same manager and the Pueblo Viejo address. It authorises a physical-cable audio and video service called SURTV for Quito and notes that the company already held an internet-access title.

The 2025 permit also records sworn declarations made in August 2024 by Tacuri and by representatives of DYCOCET S.A. and COMSADIMTELCO S.A., each described there as a shareholder of TELECOM3000. That supports a dated statement about the parties represented in that licensing proceeding. It is not a current share register, and it does not reveal economic ownership percentages, financing arrangements or control rights. A permit is evidence of legal authority, not a complete picture of governance or balance-sheet capacity.

The asset boundary is less firm. TELECOM3000 is the licensed service provider, but public materials do not identify which ducts, poles, fibre strands, rooftop sites or interoffice links it owns; which it leases; or which are supplied by another authorised carrier. The 2021 title expressly allowed subscriber access through dedicated physical and wireless links using its own infrastructure and/or infrastructure licensed from third parties. The public record therefore supports operation of the service, but not a conclusion that every cable, handhole, pole attachment or backhaul segment bearing Surnet traffic belongs to TELECOM3000.

That ownership distinction becomes material during a cut. A provider can control routing and customer support while depending on another company for the damaged span, access permission, splice crew or replacement optics. Conversely, it can own the last mile but buy upstream transport whose repair priority is set by a wholesale contract. Customers experience one outage; operations staff may be coordinating several legal and physical owners. TELECOM3000’s public pages make local support visible, but they do not publish the responsibility matrix that would show who can enter each site, authorize a splice, refuel a generator or reroute traffic.

A 100Mbps starting point meets a 1Gbps order form

The most revealing comparison in TELECOM3000’s public history is not between two current speed tests. It is between the network described at authorisation and the products now offered for sale. The technical appendix to the 2021 title, based on the original 2019 application, described one primary node called TELECOM in Sangolquí, in the canton of Rumiñahui. It listed a MikroTik router, a generic server, a generic uninterruptible power supply and a D-Link switch. No secondary node was required in the initial stage. The proposed external connection was a 100Mbps Ethernet link over fibre, with the authorised supplier still being selected.

That was an initial design, not a measurement of today’s network. It should not be treated as a continuing 100Mbps ceiling. The current online order form lists seven internet options: 350, 400, 500, 600, 700, 800 and 1,000Mbps. Current promotional cards separately present a 500Mbps plan, a 700Mbps plan, an 800Mbps plan and a 1,000Mbps plan. These are retail access rates displayed to prospective customers, with commercial terms and restrictions attached. They are not statements of a dedicated rate to every destination at every hour.

The simple arithmetic nonetheless establishes that the network offer has moved beyond the original design. A single customer tier of 1Gbps is ten times the proposed 100Mbps external link in the old appendix. The company could not meaningfully sell its current portfolio at scale if that initial external link were still the only operating path at its recorded rate. There must have been expansion, replacement, additional aggregation, caching, local exchange access, or some combination of those changes. This is a justified inference from the contradiction between dated design and current offer.

It does not disclose how much was installed, where it was installed or how much remains usable at peak load.

The contract adds one useful but limited detail. Its blank technical schedule shows fibre access and a 2:1 sharing level on the specimen form, while leaving the plan name and exact up-and-down speeds to be completed for an individual subscriber. A form is evidence of terms TELECOM3000 contemplated using; it is not proof that every current residential plan has exactly that ratio, still less that contention stops at one aggregation point. A customer can encounter sharing on a passive optical segment, an uplink from an access chassis, a metro handoff, an upstream transit circuit or a remote content path.

Each layer can be adequately provisioned on its own and still combine into a poor busy-hour experience.

Nor does “symmetric” settle the issue. Surnet’s home page advertises symmetric service, which speaks to the nominal upstream and downstream access profile. It does not reveal whether the same capacity is reserved through the wider network. A 1Gbps optical port may pass a laboratory test to a nearby server while many subscribers share a smaller upstream interface. Conversely, local caches and peering can make popular content perform well even when international transit is more constrained. The operational number that would connect the retail promise to actual experience is not the sum of advertised plans.

It is the distribution of busy-hour throughput and packet loss across access, aggregation, peering and transit, reported alongside subscriber counts and failure conditions. TELECOM3000 does not publish those figures.

The access network vanishes between the address and the node

Surnet’s fibre claim is commercially clear and physically incomplete. It tells a household that the provider intends to reach the premises over optical plant rather than a copper telephone pair or a fixed-radio subscriber unit. It does not say whether the architecture is point-to-point Ethernet, GPON, XGS-PON or a mixture; where the optical line terminals sit; how many customers share each passive optical segment; which splitters are cascaded; how long the feeder and distribution runs are; or how many homes are passed but not connected.

Those omitted details define usable capacity. The ITU specification for XGS-PON describes a 10Gbps symmetric point-to-multipoint optical system. “Point-to-multipoint” is the important phrase. Even if TELECOM3000 uses XGS-PON—and there is no public confirmation that it does—the nominal line rate belongs to a passive optical network segment, not automatically to each subscriber on that segment. If it uses GPON, the line rates and contention envelope differ. Split ratio, optical budget, scheduling, service profiles, active subscriber count and uplink design turn a standard’s line rate into customer capacity.

The physical route matters just as much. A fibre service can have two logical interfaces and one vulnerable feeder cable. Two upstream contracts can enter the same building through one duct. A ring can be drawn in network software while both sides cross the same bridge or share the same pole line. A street cabinet can contain passive splitters that need no local electricity, yet the optical line terminal, aggregation switch and upstream optics still need power elsewhere. The customer’s optical network terminal and Wi-Fi router need power in the home.

“Fibre” removes some electrical components from the outside plant; it does not remove the power chain.

The public record gives only a few fixed points. The initial node was in Sangolquí at Calle España and Juan Genaro Jaramillo. The current commercial address is in La Argelia in south Quito. The retail site says coverage is in south Quito and solicits an installation address and neighbourhood before accepting a request. No public page joins those points with a fibre route, identifies an access hub, or marks a service polygon. The gap is not permission to draw a line between them. It is an unknown.

Customer-premises equipment creates another boundary. The contract says equipment supplied to the subscriber remains Surnet’s property and must be returned when service ends. That indicates operational control over at least some devices at the edge. It does not identify optical-network-terminal vendors, Wi-Fi standards, Ethernet-port limits, replacement stock or whether remote diagnostics can distinguish a fibre fault from poor in-home radio coverage.

An advertised gigabit plan delivered through an older 100Mbps Ethernet port would be obviously constrained; a modern gateway placed behind thick masonry can also make a healthy optical link look slow. Good support separates those cases before dispatching a crew.

The missing access diagram therefore conceals several different capacity questions. How many customers share a passive segment? What is the busiest segment’s offered load? Which access chassis uplinks are oversubscribed? How many feeder routes have true physical separation? Where can traffic be switched after an optical-line-terminal failure? How many replacement terminals, power supplies and optics are held locally? The current offer proves that TELECOM3000 is selling a substantial access product. It does not reveal the installed access base or the margin left when many of those products are active together.

The 10Gbps NAP.EC port has a narrower meaning than it appears

TELECOM3000’s clearest current capacity datum sits outside the last mile. Euro-IX’s IXP database lists Telecom3000 (Surnet) at the NAP.EC Quito switch with a 10,000Mbps connection. Its entity record associates AS273177 with IPv4 address 200.1.6.46 and IPv6 address 2001:13C7:6006::27:3177:1. NAP.EC’s own public member list dates Telecom3000’s entry to 24 October 2025, and the member detail identifies AS273177 as a full member with the same exchange addresses.

This is strong evidence of a logical exchange attachment and its port rate. It is not evidence that TELECOM3000 has 10Gbps of international transit, that the port is filled, that it is protected by a second physical handoff, or that 10Gbps can reach every access node. An internet exchange lets entities exchange traffic directly or through exchange services. Traffic destined for a network present at NAP.EC may avoid a longer upstream path; traffic for networks not reachable there still needs transit or another interconnection. A port’s configured rate sets an interface ceiling.

It does not report actual load, packet loss, committed traffic or headroom.

TELECOM3000’s PeeringDB network entry supplies a compatible but self-reported signal. It classifies the network as an ISP, reports traffic in the 5–10Gbps band, describes the ratio as mostly inbound and states an open peering policy. The associated organisation entry says the network interconnects at NAP.EC and maintains BGP sessions with multiple local operators and transit providers. PeeringDB is useful because operators publish how they want potential peers to understand them. Its traffic band is not an audited measurement, and the network page itself does not list a second exchange, facility or public interconnection record.

The combination is informative. A 10Gbps NAP port and a declared 5–10Gbps traffic band are plausible together. They suggest TELECOM3000 has grown far beyond the 100Mbps starting design and is operating its own autonomous system at a meaningful local scale. They do not tell us whether the traffic figure is peak, average or rounded; whether it includes customer, peering and transit traffic on the same basis; or whether the exchange port is the constraining interface. If measured traffic actually approaches 10Gbps, a single 10Gbps port would have less room for bursts or failure diversion.

If the declared band is broad and actual NAP traffic is much lower, the port may have ample headroom. Public pages do not settle which case applies.

The same caution applies to resilience. Two BGP sessions over one switch port can provide policy choice without port diversity. Two ports in one building can fail with the building’s power or cross-connect plant. Two facilities can still share a metro fibre corridor. A defensible capacity statement would therefore pair the 10Gbps logical port with time-series utilisation, discard and error counters, physical handoff count, facility diversity and a description of what traffic moves there during an upstream failure. None is public. The port is real evidence, but its meaning ends at the exchange interface.

UFINET dominates the visible path sample

The public routing view adds a different layer. LACNIC’s autonomous-system record shows AS273177 active, registered on 30 January 2024. A RIPE overview response identifies TELECOM3000 as the holder and shows the system as announced. In a RIPE BGP-state snapshot timestamped 16 July 2026 at 15:59:55 UTC, 5,723 route views covered 17 distinct prefixes originated by AS273177. UFINET’s AS52468 appeared immediately before TELECOM3000 in 5,706 of those paths, or 99.703%. EdgeUno’s AS7195 appeared in the remaining 17, or 0.297%.

That is a powerful observation about the sampled control plane and a weak basis for claims about physical monopoly. BGP, as defined in RFC 4271, exchanges network-reachability information and AS paths. The path selected and propagated by a collector reflects routing policy, export rules and the collector’s position. It does not encode duct ownership, fibre length, circuit capacity, commercial commitment or whether two apparent neighbours enter through separate buildings. The snapshot can show that UFINET overwhelmingly preceded TELECOM3000 in the available views. It cannot show that 99.7% of customer packets used one cable.

RIPE’s separate neighbour observation reinforces the concentration. It lists UFINET with a much greater observation “power” than EdgeUno. The metric is a count derived from route visibility, not a bandwidth measure. It says that UFINET is deeply represented in the paths collectors see for TELECOM3000’s announcements. It does not say how many gigabits TELECOM3000 buys from UFINET or whether an EdgeUno circuit sits idle as a protected backup.

UFINET is a credible physical-network counterparty in Ecuador. Its country page lists Quito and Guayaquil offices and advertises capacity, dark fibre, FTTH, towers and data-centre services as part of a regional wholesale fibre business. That establishes what UFINET offers in the country; it does not by itself establish the product TELECOM3000 purchases. The BGP adjacency supplies evidence of a routing relationship. The commercial and physical terms remain private.

There are several plausible explanations for the route distribution. UFINET could be the primary transit provider and EdgeUno a smaller backup. EdgeUno could carry selected prefixes, selected address families or only a narrow policy set. TELECOM3000 could prefer UFINET for inbound reachability while retaining another path for outbound traffic. Some EdgeUno visibility could arise through a local interconnection arrangement rather than a full-capacity transit service. The sample does not adjudicate among them.

What it does reveal is the right due-diligence target. A buyer should not ask only whether TELECOM3000 has “multiple upstreams.” It should ask how much committed and burstable capacity exists on each, which prefixes are normally advertised over each, whether both can carry the busy-hour load, where the circuits hand off, and whether their physical routes separate before reaching TELECOM3000’s network. Without those answers, two AS neighbours can be genuine routing diversity and inadequate failure capacity at the same time.

EdgeUno is visible, but logical diversity is not physical diversity

EdgeUno’s appearance in the snapshot matters because it prevents the public record from being reduced to a one-neighbour story. EdgeUno’s Ecuador page offers managed connectivity, local infrastructure and 24-hour support from Quito and Guayaquil. The route data places AS7195 directly beside AS273177 in a small set of collector views. That is evidence of a live logical relationship at the sampled time. It is not evidence that the path was provisioned to absorb all TELECOM3000 traffic if UFINET disappeared.

RIPE’s routing-status response, queried for 16 July 2026, showed two observed neighbours and near-complete visibility for IPv4 and complete visibility for IPv6 among the participating RIPE collectors. It counted 512 announced IPv4 addresses and 15 IPv6 prefixes. The announced-prefix response listed two IPv4 /24s, the covering IPv6 /32 and fourteen IPv6 /48s, for 17 visible routes in total. These figures establish an active, dual-stack routing presence. They still do not measure subscribers, access capacity or recovery time.

Address registration adds another ownership boundary. LACNIC’s 148.227.173.0/24 record is active and was last changed in November 2025; the 2803:ca50::/32 record is an active IPv6 allocation registered in January 2024. The other visible IPv4 route, 193.160.220.0/24, appears in RIPE’s address-space hierarchy as assigned provider-aggregatable space associated with another organisation and an IPXO geofeed. TELECOM3000 can legitimately originate a routed block without owning the parent allocation. Routed address space, leased address rights and physical network assets are three different things.

The route-origin security picture is clearer. RIPE’s current validation responses mark 148.227.173.0/24, 193.160.220.0/24 and 2803:ca50::/32 as valid for origin AS273177. That is positive operational hygiene. Under RFC 6811, origin validation checks whether the origin AS is authorised for the prefix and permitted length. It explicitly does not attest the full AS path. Valid origins reduce one class of routing risk; they do not prove path diversity, protect a fibre span or reserve failover bandwidth.

The decisive recovery test is therefore counterfactual: if every UFINET-facing session vanished at the busiest hour, which prefixes would EdgeUno accept, how quickly would global routes converge, and how much traffic could the remaining circuits carry before loss or latency rose sharply? The 17 EdgeUno-adjacent views show that an alternative is visible somewhere in the control plane. They do not answer the capacity part. A planned failover test, observed under load and accompanied by interface data, would.

South Quito is a market, not a published route map

TELECOM3000’s geography can be bounded, but not drawn with engineering precision. The 2021 title authorised initial internet service in Pichincha and named one primary node in Sangolquí. Surnet’s current site says it has coverage in south Quito, while the commercial and regulatory addresses sit in La Argelia. The order form asks for a neighbourhood and installation address, which implies serviceability is checked at the premises level. The 2025 SURTV permit covers the city of Quito under a physical-cable service. Together, these facts support an operating focus in the Quito metropolitan area and its south-eastern approaches.

They do not support a claim of continuous coverage across Pichincha or a specific fibre corridor between Quito and Sangolquí.

The difference between a service area and a route is fundamental. “South Quito” can mean a sales territory, a cluster of neighbourhoods where plant is available, or a broader area where installation is considered after survey. A street address can be an office, customer-service point, headend, access node or several of those at once. Coordinates in a licence identify the initial node proposed in that filing. They do not prove that the node remains the main aggregation site in 2026. Nor does the presence of an AS at NAP.EC identify the building or fibre path used for the exchange connection.

No public TELECOM3000 page displays a route-level network map. There is no published inventory of poles, ducts, cabinets, splitters, towers, rooftop relays, access nodes or exchange handoffs. This absence has two consequences. First, exact route claims would be inventions. Second, resilience cannot be inferred from geographic separation between named places. A line drawn from La Argelia to Sangolquí and another toward a Quito exchange would look like a triangle, but the actual fibres could converge on one duct, one wholesale provider or one powered room.

The market context explains why that hidden topology matters. ARCOTEL’s January 2026 statistical report, using December 2025 fixed-internet data, counted 3,405,976 subscriptions nationally. Pichincha represented 27.95% and Guayas 27.24%, with the rest of the country accounting for 44.81%. The report also showed a large combined share for providers outside the largest named operators. A regional ISP can compete by extending plant into neighbourhoods, responding locally and packaging service attractively. Its constraint is often not national demand but the cost of reaching the next block, lighting the next access segment and maintaining enough spare capacity for the customers already connected.

ARCOTEL’s description of the service-regulation framework makes geographic area, network type, quality and subscriber obligations part of the regulatory structure. That does not substitute for a public engineering map. It does mean the operator’s authorised area and technical filings should exist in regulatory form even when customer-facing detail is sparse.

For a prospective customer, the practical map is an address-level answer: serviceable now, survey required, or unavailable. For an enterprise or public-sector buyer, that is limited public evidence. It needs the serving node, entrance route, shared-risk spans, upstream handoff and restoration arrangements. TELECOM3000’s public geography establishes locality. It does not establish route diversity.

Power turns advertised speed into a duration question

The initial technical appendix’s generic UPS is the only explicit power-protection item in TELECOM3000’s public network description. It belongs to the old starting configuration, not necessarily to the present network. No current page states UPS rating, battery age, generator availability, fuel autonomy, automatic-transfer design, remote power monitoring or runtime at access and aggregation sites. The honest capacity figure during a power event is therefore unknown.

Power affects each layer differently. Passive fibre splitters can continue passing light without local electricity, but the optical line terminal and aggregation switch cannot. A rooftop radio, if one exists, depends on site power even if both endpoints retain line of sight. The NAP.EC port is useful only while TELECOM3000’s router, cross-connect and path to the exchange remain active. At the customer end, an optical terminal and Wi-Fi gateway normally go dark with the premises unless the user provides backup power.

A provider may keep its network alive and still see customers disappear from monitoring because their homes have lost electricity.

That makes outage duration more useful than a binary claim of backup. A small UPS might bridge a brief interruption but not a long regional cut. A generator can extend runtime but introduces fuel, maintenance, starter-battery, ventilation and access dependencies. A remotely managed site may report low battery; a crew still has to reach it. If multiple sites draw on the same field team and the same stock of batteries or generators, a widespread event changes restoration from a technical problem into a prioritisation problem.

Ecuador’s regulator maintains a public page for provider contingency plans and issued a specific 2025 submission notice for telecommunications and subscription-broadcast providers. The existence of that obligation is relevant: TELECOM3000 should plan for continuity. It does not make the company’s plan, site inventory or achieved runtime public. Compliance evidence and operational performance are separate questions.

The customer impact also varies by service. A residential user may tolerate a short interruption and rely on mobile data. A business using cloud applications, payment terminals or security cameras may need power and connectivity to recover together. SURTV adds another service riding on physical infrastructure authorised in 2025. Bundling can improve the economics of the same access plant, but it can also put more household functions behind one optical path and one powered device.

The permit gave TELECOM3000 one year to install and begin the cable service; it does not disclose whether every advertised television bundle is now available at every internet address.

A serious resilience disclosure would list the critical powered locations, protected load, tested runtime, generator coverage, refuelling arrangement and latest load-bank or battery test. It would distinguish headend, core, access and customer equipment. Without that information, “1Gbps” is a rate with no duration attached. The commercial promise is instantaneous; resilience is measured in minutes and hours.

Recovery depends on crews, spares and rights of way

Network recovery begins where automatic protection stops. A routing session can converge in seconds, but it cannot splice a severed feeder. A passive optical segment can be rehomed only if fibres, ports and configuration exist for the move. A misaligned radio can be retuned only after a technician reaches the roof, gains access and works safely in the weather. TELECOM3000’s visible local support channels indicate that customers can reach the company, yet the public record gives no mean time to acknowledge, dispatch, repair or restore.

The likely failure paths divide into distinct work queues. A customer drop cut may affect one premises and require a short replacement cable. A feeder cut can isolate many splitters and require location, excavation permission, splicing and optical testing. An access-chassis failure needs a compatible spare, saved configuration and trained staff. An upstream loss may be handled remotely if another circuit has sufficient capacity and route policy is ready. A rooftop fault, if wireless infrastructure exists, adds keys, landlord permission, climbing safety, alignment tools and possibly replacement radios or power units.

Those queues compete for the same people. “Local support” has economic value because short travel times and knowledge of neighbourhood plant can reduce restoration time. It is also a capacity constraint. The relevant number is not total staff but simultaneous incidents that can be handled by skill type and geography. One fibre crew cannot repair two distant cuts at once. One network engineer may manage several logical failures remotely, but only if monitoring identifies the affected layer correctly. A stockroom with customer routers does not solve a failed core power supply.

Rights of way create another operator boundary. If TELECOM3000 owns the electronics but leases poles, ducts or fibre, restoration may require another party to approve access or perform the repair. If a wholesale carrier delivers the upstream handoff, TELECOM3000 can open and escalate a ticket but may not control the repair crew. Public BGP data cannot reveal those contractual dependencies. The same is true for exchange cross-connects and building access.

Recovery capacity should be tested under the failure, not inferred from normal operation. A second route that carries little traffic on an ordinary day may congest immediately when the primary fails. Spare optical ports are useful only if fibres reach them and configuration is current. Batteries are useful only if they have been tested under load. A contingency document is useful only if contact lists, escalation paths and access credentials are current. The strength of a regional provider often lies in practical readiness that never appears on a glossy site; the risk lies in assuming it exists without evidence.

For TELECOM3000, the public record supports active customer sales, a local support presence, its own autonomous system, a NAP.EC attachment and at least two visible routing neighbours. It does not publish field-force size, vehicle coverage, spare-part holdings, maintenance windows, restoration performance or wholesale escalation terms. None of those absences proves poor operations. They define the questions that remain before a customer can convert “support personalizado” into a recovery expectation.

The noon rooftop scene returns here with its proper boundary. If a microwave relay is part of the network, the spare path must survive the same weather, power event and site-access restrictions as the primary. If no microwave relay exists, crews still face fibre, power and premises faults. Either way, a resilient design needs both an alternate technical path and the human ability to put that path into service.

The due-diligence questions are measurable

TELECOM3000 is not an invisible operator. Its legal authority is documented, its Surnet retail service is active, its order form reaches 1Gbps, its autonomous system is globally visible, its route origins validate, and its NAP.EC membership includes a listed 10Gbps port. Those are substantive facts. They are enough to reject the old 100Mbps design as a description of the present offer. They are not enough to calculate end-to-end usable capacity or recovery performance.

The next disclosure should connect the layers. On access, the company could identify the technology used by service zone, the split or sharing design, the number and location class of active access nodes, the busiest-segment utilisation and the customer equipment supplied for gigabit tiers. It need not publish sensitive street-by-street security detail to state whether a service area is GPON, XGS-PON, point-to-point fibre or wireless, or to disclose the maximum engineered split and monitoring threshold.

On aggregation and interconnection, it could publish total lit uplink capacity, peak utilisation by broad route class, NAP.EC port utilisation, purchased transit commitments and failover headroom. It could explain whether UFINET and EdgeUno circuits enter different facilities and follow physically separate paths. Because the July route snapshot is dominated by UFINET, the most useful test would show what happens when UFINET-facing sessions are withdrawn under representative busy-hour load. Successful route convergence without sufficient remaining bandwidth is not successful service continuity.

On facilities and power, the measurable questions are runtime and scope. Which core, exchange and access locations have batteries? Which have generators? What runtime has been verified at current load? How often are batteries tested and replaced? Can a single power event disable both upstream handoffs or both sides of an access ring? At the customer boundary, what backup options are offered for the optical terminal and gateway?

On repair, buyers should ask for achieved rather than aspirational figures: median and high-percentile acknowledgement times, dispatch times and restoration times; the number of fibre and network crews available locally; after-hours coverage; spare optics, access devices and power supplies; and the escalation path when a third-party carrier or property owner controls access. Enterprise customers should request the serving-node and shared-risk analysis for their own circuit, because aggregate network claims may not apply to one address.

The public evidence also supports a disciplined reading of what is not known. There is no verified current microwave route, so radio capacity and relay failover cannot be claimed. There is no public fibre map, so route separation cannot be drawn. There is no current access-equipment inventory, so passive-optical capacity cannot be calculated. There is no published upstream contract, so UFINET’s dominant route visibility cannot be translated into purchased gigabits. There is no power inventory or restoration history, so backup duration and repair performance remain open.

That is not a verdict that TELECOM3000 lacks capacity or resilience. It is a finding that three visible numbers—1Gbps at retail, 10Gbps at an exchange port and 99.703% UFINET adjacency in one route snapshot—measure three different surfaces. The first is a sold access profile. The second is a logical interface ceiling at a local exchange. The third is a distribution of observed AS paths. A reliable service depends on the links between them: fibre or radio from the premises, powered aggregation, sufficient transit and peering, physically credible alternatives, and crews able to repair what automation cannot.

At noon, the technician’s final adjustment is not the end of the capacity story. The meaningful test begins when the city below becomes busy, one path disappears, and the network has to preserve useful service rather than merely advertise a rate. TELECOM3000’s public record shows that it has grown into that test. It does not yet publish the measurements that show how well it passes.