Netcom Internet Service makes Mirpur broadband rest on one visible upstream and local repair

Summary

• Netcom Internet Service can be identified with high confidence as the Mirpur, Dhaka operator behind AS141433. APNIC records place the organisation at Borobag in Mirpur, while a Bangladesh Telecommunication Regulatory Commission tariff approval dated 7 July 2025 explicitly describes it as a district ISP at the same address.
• The routed edge is active. RIPEstat observed 2,560 IPv4 addresses, one IPv6 /48 and complete visibility among its participating IPv4 and IPv6 route collectors on 10 July 2026. Cloudflare Radar also sees traffic and estimates an audience of roughly 10,000 users, although that estimate is not a subscriber count.
• The strongest resilience warning is upstream concentration. RIPEstat, Hurricane Electric and other global BGP views show Summit Communications AS58717 as the one immediate external path. BGP.tools also records domestic exchange peers, but domestic peering does not create an independent route to the wider internet, and no public source establishes physically separate Summit paths.
• Netcom says it operates a battery-backed distribution network, offers call-centre support around the clock and fields technicians from 10am to 10pm. Those are useful service claims, not verified measurements: battery runtime, site count, route diversity, repair staffing, spare inventory and actual restoration performance are not disclosed.
• The regulator, APNIC, ISP association and company website all point to a Bangladesh district ISP centred on Mirpur and advertising coverage in Dhaka. The public geography is Bangladesh within Asia-Pacific, while physical resilience remains explicitly unverified.

The monthly bill hides the topology

A broadband invoice reduces a network to a few familiar numbers: a monthly charge, an advertised speed and a due date. Netcom's regulator-approved schedule makes that compression unusually clear. The BTRC tariff approval published by the company lists retail packages from 10 Mbps at Tk500 a month to 100 Mbps at Tk3,500, with a stated contention ratio of 1:8. It also sets compensation for long continuous interruptions. A user pays half the monthly bill after five days of continuous outage, a quarter after ten days and nothing after fifteen days.

Those terms describe the commercial consequence of a prolonged failure, not the probability of one. The price table does not reveal how a customer's building is connected, where the nearest optical splitter or powered switch sits, how many ways traffic can leave the neighbourhood, or whether two nominal circuits share the same pole line, underground duct or upstream chassis. It cannot show whether batteries last through a local power cut, whether a damaged fibre can be reached in traffic, or whether a replacement optical module is on a shelf nearby.

For a district ISP, these hidden dependencies are the service. A customer packet must cross the router or optical terminal in the premises, the building drop, a local distribution segment, one or more aggregation points and transport to an external carrier. Each stage can fail independently. Several can also fail together: a road excavation can cut power and communications in the same corridor, a flooded cabinet can affect many buildings, and two logical links can converge on one carrier facility.

The bill remains local, but the recovery chain extends from a Mirpur street to another company's transport network and onward to domestic exchanges and international capacity.

Netcom is a particularly instructive case because the public record establishes a real operating network without exposing the physical design beneath it. The appropriate conclusion is neither that the company is only a paper trace nor that a visible ASN proves a resilient access network. The evidence supports a live district ISP. It does not yet support a claim of physically diverse broadband delivery.

Identity resolves to Mirpur, not to a global Netcom brand

"Netcom" is used by unrelated telecommunications businesses in several countries, so a name-only search is unsafe. The identifiers converge on one Bangladesh operator. APNIC's registration for AS141433 names NIS-AS-AP, describes the holder as Netcom Internet Service and gives Bangladesh as the country. The organisation record places Netcom at 22/I/4/1, Borobag, Section-2, Mirpur, with the same domain and contact details used by the company's public site.

The regulator provides the strongest scope evidence. BTRC's 7 July 2025 approval is addressed to the proprietor of Netcom Internet Service at 22/I/4/1, Borobag, Mirpur, Dhaka-1216. Its subject line explicitly calls Netcom a "District" ISP. The Internet Service Providers Association of Bangladesh member listing also identifies Netcom Internet Service as a district licensee and gives a Mirpur-2 address. The difference between the association's West Monipur office and APNIC's Borobag address may reflect an office move or separate commercial and registered locations; it should not be converted into evidence of two network sites.

Netcom's own coverage page says coverage is available across Dhaka. Its about page describes the business as a district ISP providing broadband since 2015. An older company contact page, retained on the same domain, also placed the operation in Mirpur-2 and displayed a Dhaka North City Corporation trade-licence number. Together, these records identify a Dhaka-focused access provider rather than an international carrier.

The age claims require care. The current website says "since 2015" and elsewhere refers to more than 12 years of experience, figures that do not align cleanly with each other or with AS141433's November 2020 registration. None necessarily has to be false: a local access business can predate its own ASN, and staff experience can predate the company. But the public pages do not explain the chronology. The defensible statement is that Netcom says it has sold broadband since 2015, while its independently registered routing identity dates from late 2020.

This geography directly contradicts a Global region label. An autonomous system can announce routes globally without operating a global retail network. The route is visible from global collectors because that is how internet reachability works; it does not turn Mirpur coverage into a worldwide service area. Netcom should be understood as a Bangladesh district ISP with a Dhaka-facing commercial footprint whose exact street-by-street service boundary is not publicly enumerated.

A live routed edge establishes operation

AS141433 is not dormant. RIPEstat's AS overview showed the system actively announced on 10 July 2026. Its routing-status view reported that all 327 participating IPv4 peers and all 321 participating IPv6 peers could see at least one Netcom route. The first route in that view, 103.159.72.0/23, was observed on 27 November 2020. Five and a half years of routing history does not prove uninterrupted service, but it decisively improves on a website-only operating claim.

The commercial surface reinforces that conclusion. Netcom has a customer portal that offers bill status, usage information and support access. Its payment page describes online payments through cards and Bangladesh mobile-money services, including bKash and Nagad. A regulator-approved 2025 tariff, a maintained APNIC contact, live routes and a functioning billing path are mutually reinforcing indicators of an operating retail network.

Cloudflare Radar's AS141433 page observes traffic and, at the research cut, displayed an estimated population of about 10,000 users. That figure must not be read as 10,000 active accounts. Cloudflare says the population measurement comes from APNIC; such estimates describe users seen through measurement systems, not audited subscribers, paid lines, premises passed or average concurrent sessions. The company's own about page claims more than 50,000 happy customers and more than 100 enterprise clients, but no regulator return, audited statement or methodology accompanies those numbers.

The responsible operating-status judgment is therefore stronger than the initial thin-footprint hypothesis but narrower than the company's promotional language. Netcom runs a visible network and current retail service. The size of the customer base, revenue, churn, service penetration and enterprise mix remain unverified. The distinction matters when assessing repair capacity: ten thousand measured users, fifty thousand claimed customers and fifty thousand premises passed would imply very different staffing and network-density requirements.

Address space is not the same as owned capacity

RIPEstat's announced-prefixes view showed eleven IPv4 route entries and one IPv6 /48 in the two weeks ending 10 July 2026. Because the list includes both the aggregate 103.159.72.0/23 and its two component /24s, route-entry count should not be added directly into a capacity figure. The routing-status view deduplicated the visible IPv4 space to 2,560 addresses.

The clearest Netcom-held resources are portable. APNIC's registration for 103.159.72.0/23 names Netcom Internet Service and covers 512 IPv4 addresses. The IPv6 registration for 2001:df1:c240::/48 also names Netcom and is marked portable. Portable resources give an operator more administrative independence than addresses carved from a carrier's allocation. They can, in principle, be originated through a different provider if routing policy, contracts and physical connectivity permit.

Other routes need a more careful ownership boundary. 103.149.105.0/24, for example, appears in Netcom's current origin set, but its RDAP registration names Fastnet rather than Netcom. The record for 103.174.189.0/24 is maintained under another Bangladesh allocation even though its description references Netcom. Several newer /24s originated by AS141433 are non-portable ranges with registry contacts tied to other holders. Originating a route demonstrates routing control at the observation time; it does not by itself prove title to the address block, a permanent lease, or ownership of the fibre carrying it.

This is not merely a registry footnote. A network built partly on assigned or partner-provided address space may be perfectly functional, but continuity depends on contracts and operational coordination outside the ISP. A portable block can survive a carrier change more readily than provider-dependent numbering, yet even portable addresses are useless during a physical transport outage. Conversely, several originated prefixes do not mean several upstreams. All can travel through the same external path.

Current routing-security signals are positive. Hurricane Electric's AS141433 view marked the visible routes it counted as valid under the Resource Public Key Infrastructure, and RIPEstat's validation for the core /23 found a valid route-origin authorization for AS141433. This reduces the risk that a conforming network will accept the prefix from an unauthorized origin. It does not authenticate every hop, prevent route leaks, guarantee traffic delivery, or say anything about battery runtime and cable diversity. RPKI is a routing authorization control, not an uptime certificate.

The internet-facing path is visibly concentrated

The most important public topology fact is simple. RIPEstat's ASN-neighbours result found one unique observed neighbour: AS58717, Summit Communications. Hurricane Electric likewise showed Summit as the sole observed IPv4 and IPv6 peer, while CIDR Report described one adjacent system on the path toward the global table and no downstream adjacency. IPinfo accordingly characterises AS141433 as single-homed.

These sources observe routing, not civil works. They show that global collectors see Netcom behind Summit, but they do not reveal the number of physical ports, handoff sites or fibre corridors between the two organisations. Netcom could have two Summit circuits at different locations and still present one AS-level neighbour. It could also have two logical sessions that share one duct and one powered aggregation site. The public BGP result therefore supports upstream-provider concentration, while physical single-path exposure remains a serious but unconfirmed possibility.

BGP.tools adds useful nuance. Its view lists Summit as the upstream and also records Bangladesh Internet Exchange Trust AS140684 and Rahul Enterprise AS139689 as peers. Those domestic adjacencies may improve access to local content or other connected networks. They do not substitute for a second global transit provider. An exchange route can keep some domestic destinations reachable while international services, cloud applications or remote corporate systems fail through the Summit path.

Peering disclosure is incomplete elsewhere. PeeringDB's AS141433 record links the correct company and website but lists no public exchange point and no interconnection facility. That absence does not disprove the sessions shown by BGP.tools; PeeringDB is voluntary and can lag reality. It does mean there is no public facility and port inventory against which a buyer can test geographic diversity.

This distinction between logical and physical redundancy is the central diligence question. A second BGP neighbour would improve the public case, but it still would not be enough if both carriers entered the same building through the same trench. What would materially strengthen the resilience claim is a documented pair of upstreams, handoffs in separate facilities, transport paths with limited shared fate, independent power boundaries and evidence that failover has been tested under load.

Domestic exchange reach can improve performance without rescuing every outage

Bangladesh's exchange ecosystem matters because a local ISP carries two broadly different traffic classes. Some traffic can be exchanged domestically among local networks and content systems. Other traffic must reach international destinations through licensed gateways and long-haul capacity. Keeping popular domestic paths local can reduce latency and transit cost, but only for the routes available at the exchange.

The BDIX connectivity agreement requires members to have their own global internet connectivity and autonomous system number. That language captures the boundary: exchange membership complements global service rather than replacing it. BDIX also publishes a separate application for redundant port connectivity, which shows that even an exchange-facing connection has its own redundancy decisions.

For Netcom, BGP.tools' observation of AS140684 is a positive operating signal, but the absence of a disclosed port, site and capacity prevents stronger conclusions. There is no public evidence that the domestic exchange path uses a different access fibre from the Summit handoff. If both sessions arrive over one NTTN circuit or one building entry, a local cut can remove both at once. If they terminate at separate facilities, some domestic traffic may survive a transit failure. The routing tables alone cannot settle that question.

The practical effect on users varies by application. A cached video, local game server or Bangladesh-hosted service might remain reachable during a partial international failure, while an overseas software platform, foreign university system or remote-work VPN does not. A speed test pointed at a nearby server may look healthy even when international transit is congested. Resilience assessment must therefore include destination diversity, not only the access line's nominal rate.

The access network is where public visibility fades

Netcom's services page markets packages for homes, small businesses and corporate users. The coverage page says Dhaka, and the registered addresses repeatedly point to Mirpur. Beyond that, the public site does not publish a fibre map, list of points of presence, tower inventory, optical access design, pole agreements or building coverage list. It does not state whether customer access is fibre to the premises, Ethernet over shared building cabling, wireless, or a mixture.

Broadband is the defensible service description; a specific last-mile technology is not. Fibre distribution is plausible for fixed urban access and consistent with the network's scale, but no public record identifies a particular pole, cabinet, splice closure, tower or route as Netcom-owned.

Ownership is likely layered even where Netcom controls the service. A district ISP can own customer-premises equipment and local switches while leasing dark fibre, bandwidth or transport from a nationwide telecommunications transmission network provider. It can use building owners' risers, utility poles, shared ducts and third-party field contractors. The BTRC tariff's quality framework explicitly refers to NTTN path redundancy and separates restoration targets from failures that depend on an IIG or NTTN. That regulatory wording recognises that the retail ISP does not control every link in the chain.

The key asset question is not simply "does Netcom have fibre?" It is which sections it operates, which it leases, where responsibility transfers, and how quickly each owner must respond. A fault in a customer drop may be entirely within Netcom's field team's control. A cut in leased metro transport may require diagnosis by Netcom, dispatch by another company and site access controlled by a third. A Summit-facing outage may be logical, optical, facility-related or upstream of the handoff. Restoration time accumulates across those boundaries.

No route map also means no evidence of a ring. A distribution ring can permit traffic to reverse direction around one break, but only if the optical design, switching configuration and power system all support that failover. A line drawn as a ring in sales material would still be limited public evidence without confirmation that both sides are lit, monitored and dimensioned for the redirected load. Netcom publishes no such claim, so the access plant should be treated as topology unknown rather than assumed radial or assumed redundant.

Battery-backed is a useful claim with an unknown clock

Netcom's homepage says it has a battery-backed distribution network, a 24-hour call centre and field support from 10am to 10pm. This is the company's most specific resilience statement. It suggests awareness that an urban fixed-broadband service can fail when distribution electronics lose utility power even if the fibre remains intact.

The missing variable is time. Battery backup can mean a small uninterruptible power supply holding a switch through a short interruption, or a managed battery system supporting a larger node for several hours. Runtime depends on battery chemistry, age, temperature, load, charging history and whether additional equipment has been connected. A battery may keep an optical line terminal alive while an intermediate switch or customer router goes dark. It may also exhaust before utility restoration, turning a delayed outage into a full one.

No public Netcom material gives runtime, site coverage, maintenance interval, replacement policy or test results. The phrase "battery-backed network" should therefore be reported as a self-description, not converted into a claim that service survives a defined outage. Evidence that would settle the question includes a site inventory, design load, tested runtime, low-battery telemetry, replacement dates and records from real utility interruptions.

The customer side creates another boundary. Even if every Netcom node remains powered, a household optical terminal and Wi-Fi router usually require electricity. A customer without a small UPS experiences an outage while the provider network remains available. A business with backup power may stay online and expose a provider-side weak point that residential users cannot distinguish. Service communication should separate customer-power loss, access-node power loss and upstream facility loss because each has a different remedy.

Power diversity also requires more than a battery. Two upstream handoffs in one building can share the same electrical panel. A route ring can include active cabinets on both arms that depend on one feeder. A call centre may be reachable while the technicians' splicing and test equipment cannot be charged. A robust design combines batteries with monitored load, generator or alternative supply where justified, fuel and access planning, and periodic failover tests. None of those controls is publicly verified for Netcom.

Advertised megabits are not installed headroom

The tariff table is useful because it places concrete retail promises against unknown shared capacity. The approved products range from 10 to 100 Mbps and identify a 1:8 contention ratio. Contention is not inherently a defect; consumer broadband economics depends on users not drawing peak rate simultaneously. The engineering question is whether aggregation and upstream capacity are increased before busy-hour demand overwhelms the shared pool.

A 100 Mbps package is an access-rate ceiling, not proof that 100 Mbps is continuously available to every subscriber. Performance depends on the local segment, optical split or switch uplink, aggregation backplane, transport circuit, transit port, destination and time of day. The BTRC tariff sets a framework, but no public Netcom measurement shows busy-hour throughput, latency, packet loss, jitter or utilisation.

Address counts also do not provide a shortcut. Netcom's 2,560 visible IPv4 addresses could serve fewer customers with public addressing, more customers behind carrier-grade network address translation, enterprise services, infrastructure devices or some combination. The enormous numerical size of an IPv6 /48 is a feature of IPv6 architecture, not evidence of enormous customer capacity. Neither resource measures the number of fibre ports, the speed of the Summit handoff or the amount of spare capacity after a failover.

The website's claims of 99.9% uptime, 50,000 customers and more than 100 enterprise clients are similarly unverified. Ninety-nine point nine per cent availability corresponds to roughly 44 minutes of downtime in an average month, a much stronger claim than the lower grades described in the published tariff framework. Netcom does not disclose the measurement boundary, exclusions, sample, or reporting period. A core router can achieve that figure while an individual neighbourhood suffers repeated access faults.

Capacity must also be considered under failure. A second route is useful only if it can carry the surviving load. If two 1 Gbps links normally run near their limits, losing one may preserve BGP reachability while causing severe packet loss and latency. Installed capacity is the sum of ports and circuits; usable resilient capacity is what remains at the busiest hour after the largest credible failure. Netcom publishes neither figure.

Field repair is part of the network, not a customer-service extra

Netcom distinguishes a round-the-clock call centre from field support available between 10am and 10pm. That separation is candid and operationally important. A fault can be reported at any hour, but the people able to replace a power supply, resplice fibre or gain roof access may not be available until the field window opens. The public site does not state whether severe outages trigger an on-call response outside that window.

Urban repair is not automatically fast. A fibre break has to be detected, localised and matched to a route record. Technicians need access permission, safe working conditions, a splicer, test gear, closures, patch leads and the correct optical modules. Traffic and road works can delay arrival. A fault inside an apartment building may require a caretaker or owner. A leased transport failure must be escalated to the asset owner, which may have its own queue and service target.

The most revealing resilience numbers would therefore be human rather than promotional: median time to acknowledge, median and 95th-percentile time to restore, after-hours dispatch policy, simultaneous incident capacity, number of trained splicers, vehicle coverage, strategic spare locations and escalation times with Summit and any transport provider. None is public. "Local support" is evidence-supported as a dependency and a controlled topic, but not as a demonstrated strength.

Repair quality also affects repeat failure. A temporary aerial span can restore service quickly but remain exposed. A rushed splice can pass light while adding loss that later causes intermittent faults. Batteries can be swapped without addressing charger failure. Meaningful recovery evidence should distinguish temporary restoration from permanent remediation and show whether the same segment fails again.

This matters economically. A district operator has the advantage of proximity: technicians may know the streets, buildings and landlords better than a national call centre. It also has a scale constraint. Keeping several crews, spare optics and backup equipment available is expensive when revenue is spread across low monthly tariffs. The quality of the service depends on whether customer density is high enough to support that readiness without running every link and every technician at full utilisation.

Six failure paths define the real service

The first failure path is the customer drop or building distribution. One damaged cable may isolate a single household, floor or building while Netcom's ASN remains perfectly visible. BGP monitoring will not detect it. Recovery depends on accurate building records, access and a technician with the right materials.

The second is a neighbourhood access cut. If several local branches converge on one feeder without a lit alternate direction, road work or construction can interrupt many customers. A spare fibre in the same cable improves repair options but does not provide route diversity when the entire cable is severed. A ring in the same duct also shares the excavation risk.

The third is loss of power at a distribution or aggregation point. Netcom says batteries are present, but unknown runtime makes the outage threshold unknown. Service may continue initially and then disappear as batteries exhaust. Restoration can require utility power, a replacement battery, a generator connection or a reduction in load.

The fourth is loss of metro transport or the Summit handoff. All globally observed Netcom routes currently pass through AS58717. A failure at that boundary can remove international reachability across the address estate at once. Domestic exchange paths may preserve selected local routes if they are physically and electrically independent, but that independence has not been demonstrated.

The fifth is congestion rather than disconnection. A saturated access uplink or transit port leaves the service technically up while making video calls, cloud applications and interactive work unreliable. The tariff speed does not disclose busy-hour headroom, and an uncongested local speed test may miss an international bottleneck.

The sixth is recovery overload. A storm, power disturbance or common cable cut can create many tickets simultaneously. The call centre may accept them while a limited field team works serially. Spare batteries, optical modules and closures can become the binding constraint. For customers, the outcome depends on incident prioritisation and crew depth as much as on router configuration.

These paths affect different groups unevenly. Households lose work, study, entertainment and communication. Small shops can lose cloud point-of-sale access and digital payments. Offices may lose remote systems even when domestic sites remain available. Enterprise users with a second provider can still fail if both providers share the same building entry or NTTN corridor. The geographic concentration around Dhaka can make a local civil-works or power incident more consequential than the worldwide visibility of the routes suggests.

The economics reward density and punish hidden shared fate

Netcom's approved entry package costs Tk500 a month. At that price, the operator must spread upstream transit, leased transport, equipment, electricity, support, billing and repair across many users. Dense urban coverage can make the model work: short drops and many customers per distribution point reduce capital and travel cost. The same density increases the number of users affected when one aggregation node or feeder fails.

The tension is not unique to Netcom, but the evidence makes it concrete. A district licence limits the commercial frame; one visible upstream concentrates the internet-facing dependency; a 1:8 contention ratio formalises shared capacity; and a stated 12-hour field window defines at least the normal labour boundary. Each choice can be rational in isolation. Resilience depends on how they interact under stress.

Buying a second provider, diversifying routes and holding more spares all cost money before they generate visible customer value. So does replacing batteries before failure. Customers often select on advertised speed and price, leaving less room to monetise unseen redundancy. Regulatory compensation after multi-day outages protects users at the extreme, but the first hours of downtime can still impose losses far larger than the pro-rated bill.

The most valuable commercial evidence would link price to service class. A business product could specify two building entries, a restoration target, monitored power and a second transit path. A residential product could state the normal support window and battery design target. Netcom's public pages instead use broad reliability language. Without technical service definitions, customers cannot tell whether an enterprise plan buys a different physical path or mainly a different support arrangement.

Unofficial network measurements offer a useful signal but cannot fill that gap. IP databases and route collectors consistently see an active Bangladesh eyeball network behind Summit. That supports operation and concentration at the AS level. It cannot establish customer satisfaction, street coverage, fibre condition or repair outcomes. Reviews and isolated speed tests would be even narrower: they can reveal a local experience, but not the design of the network as a whole.

What would raise the evidence grade

Netcom can move the resilience assessment from unverified to credible with a relatively small set of disclosures. The first is an access and transport diagram that marks service areas, aggregation sites, route directions and ownership boundaries without exposing sensitive street-level detail. The second is a list of upstream providers and handoff facilities, including whether the transport paths and building entries are physically independent.

The third is capacity evidence: normal and peak utilisation, port speeds, oversubscription at each aggregation layer and the capacity remaining after one link fails. The fourth is power evidence: the proportion of active sites with backup, tested runtime under design load, monitoring coverage and battery replacement policy. The fifth is recovery evidence: field hours, on-call escalation, crew and spares coverage, and restoration distributions rather than a best-case target.

Routing improvements would be visible too. A sustained second global upstream in multiple collectors would reduce provider concentration, especially if Netcom documented a separate facility and corridor. A current PeeringDB entry with facilities, exchange ports and policy would make interconnection easier to evaluate. Continued valid RPKI for every originated prefix would preserve the positive origin-security signal, while registry records should clarify the contractual status of non-portable ranges.

Customers and procurement teams should ask one practical question repeatedly: what is shared? Two subscriptions may share a pole, duct, building entry, switch, power source, NTTN circuit, Summit handoff or international gateway. Redundancy exists only where the failure domains separate. A second invoice is not a second path unless the physical and operating boundaries differ.

Until those facts are available, Netcom's evidence grade is Medium for a live network and Weak for physical resilience. The company is demonstrably more than a name: it has a current regulatory tariff, an operating commercial surface, portable resources, IPv6 and globally visible routes. But the title's causal claim must remain bounded. Mirpur broadband depends on access plant, power, Summit-facing transport and field repair; public evidence does not show which of those dependencies has a tested alternate.

A local service with a global route and a local point of failure

Netcom Internet Service illustrates how the internet's global appearance can conceal a very local operating reality. Its prefixes are visible from hundreds of route-collector peers, and its customers can reach destinations far beyond Bangladesh. Yet the customer experience begins in a building and passes through equipment that must be powered, maintained and repaired in Dhaka.

The public record supports confidence in identity and operation. AS141433 is active, IPv4 and IPv6 are visible, BTRC approved a current district tariff, and Netcom maintains customer and payment functions. It supports only provisional confidence in continuity. One globally observed upstream, no published facility map, unknown access topology, an unmeasured battery claim and undisclosed field capacity leave common-mode failures unresolved.

That is also why the geography should be corrected. Netcom is not Global in the sense relevant to retail service. It is a Bangladesh district ISP centred on Mirpur, with a route that participates in the global internet. Its resilience will be decided where the abstractions end: at the building entry, the powered distribution point, the transport corridor, the Summit handoff and the moment a technician is sent to restore light through a damaged fibre.