Summary
- TELECOM-AS-AP is the registered name of AS137451, associated by APNIC with Hong Kong Business Telecom Limited. Public routing monitors classify that ASN as inactive: it has originated no visible IPv4 or IPv6 space since early April 2025, and its three other registered AS numbers also have no current announced space.
- The organisation's resource footprint has not simply vanished. APNIC records associate it with twelve portable IPv4 /22 allocations, equal to 12,288 addresses, plus 2401:2c80::/32. A route-by-route check found 45 of the 48 component IPv4 /24s visible through 18 other origin networks, while the IPv6 /32 was visible through another origin. Registration, route operation and customer service therefore sit in different hands.
- No public evidence establishes that TELECOM-AS-AP owns or operates local fibre, fixed-wireless towers, poles, ducts, customer-premises equipment, a staffed network operations centre or field-repair depots. Its registered address is an administrative contact point, not proof of a network facility, and the exact company name does not appear in Hong Kong's current public lists of carrier or services-based operator licensees.
- A buyer should not infer route diversity from the number of address blocks or origin networks. Resilience has to be proved for the actual service address: access medium, building entry, upstream paths, power boundary, spare equipment, named repair party, restoration target and an escalation route that remains usable during an outage.
The first surprise is that the network name is not the network
The most precise public description of TELECOM-AS-AP is not a consumer offer or a coverage map. It is an internet registry entry. APNIC's AS137451 record assigns the name TELECOM-AS-AP to Hong Kong Business Telecom Limited and gives Hong Kong as the country. It identifies an administrative organisation and the number under which a routing policy could be presented to the global internet. It does not say that a household, office tower or industrial estate can order a line.
That distinction matters because an autonomous-system number is often treated as a compressed description of a company. In a vertically integrated operator, the shortcut can be useful: the company owns or controls access plant, puts customer addresses into service, operates edge routers and buys transit or peering under one recognisable identity. Here, the shortcut fails. The number named in the title is no longer visible as a route origin, even though address space registered to the same organisation remains heavily used under other autonomous systems.
Hurricane Electric's AS137451 routing profile says the network has not been visible in the global routing table since 7 April 2025. RIPE NCC's independent routing-status observation places the last seen route slightly earlier, on 2 April 2025, and reports zero announced IPv4 and IPv6 space. Collector dates can differ because they observe different peers and sampling windows. Their material conclusion is the same: AS137451 is not a current origin network.
IPinfo therefore labels AS137451 inactive, with zero IPv4 and IPv6 addresses currently attributed to it. Cloudflare still has an AS137451 identity page, but an identity page is not proof of current traffic. A number can remain registered and indexed long after its routes have disappeared. Conversely, an address allocation can stay in use after the holder stops originating it through its own ASN.
The name also needs to be kept separate from similarly named carriers. TELECOM-AS-AP is not a generic synonym for Hong Kong Telecom, HKT, Hong Kong Broadband Network or another operator whose name contains telecom. The authoritative handle is AS137451, and its organisation field points to Hong Kong Business Telecom Limited. Confusing it with a larger carrier would import networks, licences, buildings, staff and service promises that the record does not support.
This leaves a narrower but still important business. The organisation appears to be a steward or commercial controller of internet number resources that are used by multiple origin networks. That can be a legitimate layer of internet infrastructure. Address holders may assign or lease space, allow a customer or hosting provider to originate it, or divide blocks among operating partners. But it is not equivalent to a local ISP that can repair the fibre outside a shop after a cut.
The registered portfolio is large enough to look like an operator
APNIC's inverse organisation records associate Hong Kong Business Telecom Limited with a notable resource portfolio. It includes twelve portable IPv4 allocations, each a /22 containing 1,024 addresses, and one IPv6 /32. The IPv4 total is 12,288 addresses before subtracting addresses unavailable for ordinary host use within each routed subnet. The same organisation record also connects four autonomous-system numbers: AS132422, AS135119, AS137451 and AS64021.
The blocks are not all maintained in the same way. The records for 103.20.220.0/22, 103.48.168.0/22, 43.229.152.0/22 and 43.252.208.0/22 use TELECOM or HKNET-BROADBAND maintainer handles. Others delegate lower-level and route maintenance to handles labelled PACDIA, ASIANET, WGOC or NCEN. A maintainer handle indicates who is authorised to change particular database entries. It does not disclose the commercial agreement, the physical router, the end customer or who will attend a failed circuit.
| Registered resource | Registry label or scope | What the record establishes | What it does not establish |
|---|---|---|---|
| 103.20.220.0/22 | TELECOM-HK | Portable allocation associated with the organisation | A Hong Kong access footprint or one operating origin |
| 103.209.232.0/22 | TELECOM-AS-AP | Portable allocation with PACDIA maintenance handles | Ownership of customer links or field staff |
| 103.225.196.0/22 | TELECOM-HK | Portable allocation with ASIANET maintenance handles | A unified service area |
| 103.228.64.0/22 | TELECOM-HK | Portable allocation with WGOC maintenance handles | That all four /24s are in service |
| 103.234.52.0/22 | TELECOM-HK | Portable allocation with NCEN maintenance handles | Physical location of servers or routers |
| 103.48.168.0/22 | TELECOM-HK | Portable allocation associated with the organisation | One origin, one facility or one product |
| 150.107.0.0/22 | TELECOM-HK | Portable allocation with NCEN maintenance handles | Retail broadband availability |
| 150.129.40.0/22 | TELECOM-HK | Portable allocation with ASIANET maintenance handles | Installed customer capacity |
| 163.53.16.0/22 | TELECOM-HK | Portable allocation with WGOC maintenance handles | A tower, pole, duct or cable route |
| 36.255.192.0/22 | TELECOM-AS-AP | Portable allocation with PACDIA maintenance handles | Control of the route origin or transit |
| 43.229.152.0/22 | TELECOM-HK | Portable allocation associated with the organisation | That every component prefix is announced |
| 43.252.208.0/22 | TELECOM-HK | Portable allocation with HKNET-BROADBAND handles | A current public-facing service |
| 2401:2c80::/32 | TELECOM-HK | Portable IPv6 allocation | Customer IPv6 availability through AS137451 |
The portfolio creates an appearance of scale because a /22 is a familiar commercial unit in address markets and hosting. Yet registered address count is not installed bandwidth. It does not disclose router port speed, committed information rate, oversubscription, usable rack power, access-tail capacity or the number of paying subscribers. Forty-eight potential IPv4 /24s could support many servers, a smaller number of network-address-translation gateways, or no live hosts at all. The addresses are identifiers and routing inputs, not cables.
The four autonomous-system registrations reinforce that lesson. APNIC records AS132422 as TELECOM-HK, AS135119 as TELECOM-KR and AS64021 as Network-Transit, in addition to AS137451. RIPE's current status reports show zero announced space for all four. AS135119 has no observed route history in that dataset. The others were visible in earlier periods but had all ceased announcing by April 2025 or before.
A buyer cannot add those four numbers together and call the result fourfold redundancy. Redundancy exists when the same service can survive a defined failure through a tested alternate path with enough spare capacity. Four dormant routing identities provide no such protection.
The address space is active, but under other origin networks
The strongest evidence of current activity appears one layer away from TELECOM-AS-AP. Each /22 can be divided into four /24s, the smallest prefix commonly propagated across the whole IPv4 internet. Checking all 48 components against RIPE's live prefix view found 45 announced and three unannounced. None of the announced /24s used AS137451, AS132422, AS135119 or AS64021 as origin. Instead, 18 other autonomous systems appeared across the IPv4 set.
The split is visible inside single allocations. Within 103.48.168.0/22, the 103.48.168.0/24 route was originated by AS146952, 103.48.169.0/24 by AS153706, 103.48.170.0/24 by AS140525 and 103.48.171.0/24 by AS141153. Four adjacent address ranges registered under one /22 therefore had four current routing origins.
Another allocation mixes live and absent space. RIPE's 103.228.64.0/24 observation showed AS153706, while 103.228.65.0/24 showed AS132825. The remaining 103.228.66.0/24 and 103.228.67.0/24 were not announced at the cut-off. A /22 allocation is therefore not evidence that its full address capacity is lit, routable or available for sale.
The 43.229.152.0/22 block is similarly divided. Hurricane Electric's 43.229.152.0/24 page identifies Zenlayer's AS21859 as a current origin and also shows larger overlapping announcements. The 43.229.153.0/24 route was originated by Stack Network's AS153706. The next /24 was not visible, and 43.229.155.0/24 used AS131646. That is not one network with three upstreams. It is one registered allocation divided among distinct origin policies.
| Example prefix | Current origin reported at the cut-off | Significance |
|---|---|---|
| 103.20.220.0/24 | AS208628 | Registered holder and route origin differ |
| 103.20.222.0/24 | AS40065 | Adjacent /24s within the /22 use another origin |
| 103.209.232.0/24 | AS146952 | The PACDIA-maintained /22 is operationally subdivided |
| 103.209.233.0/24 | AS401948 | A second origin appears in the same allocation |
| 103.225.197.0/24 | AS62610 | One of two Zenlayer ASNs seen across the portfolio |
| 103.234.54.0/24 | AS9304 | HGC is the route origin, not proof of a retail HGC product |
| 150.107.1.0/24 | AS202736 | Another independent origin in an NCEN-maintained block |
| 163.53.18.0/24 | AS38136 | Current routing says nothing by itself about facility location |
| 43.252.211.0/24 | AS142132 | The fourth /24 in this allocation has its own origin policy |
| 2401:2c80::/32 | AS214266 | IPv6 is visible, but not through any of the holder's four ASNs |
The origin names help describe the route but do not establish a durable commercial relationship with Hong Kong Business Telecom Limited. RIPE's AS-overview service identifies operators such as CNSERVERS, Zenlayer, HGC Global Communications, Stack Network, Akari Networks and several smaller technology companies among the observed origins. The public route does not disclose whether an origin is a lessee, customer, managed-network provider, reseller, temporary operator or an authorised party under another arrangement.
This is also why 18 IPv4 origins must not be reported as 18 upstreams. An upstream carries routes for another network. An origin asserts reachability for the prefix at the end of the path. To determine upstream diversity, each origin's path must be examined separately, including the facility, physical handoff and common providers behind apparent alternatives. Counting origins across unrelated /24s says nothing about whether one customer's circuit has a backup.
The address distribution does establish an economic clue. The resource holder can derive value from portable addresses even when its own ASN is dormant. IPv4 scarcity makes routable /24s commercially useful to hosting providers, security services and networks that need independent address space. But the public records do not show price, contract term, customer identity, renewal risk or the party responsible for route authorisation. They support an address-resource business more strongly than a local access-network business.
A local broadband product is not visible
The planned category for TELECOM-AS-AP is regional ISP, but category is not evidence of a working last mile. No public company page was found offering fibre-to-the-building, fixed wireless, leased access, residential broadband or business internet at a serviceable address. The domain associated with the registry contact, hknetbroadband.com, resolves to a commercial parking page rather than an operating product catalogue. An older administrative contact domain, 31.com.hk, presents a short under-revision message and says the domain is for sale; its certificate also does not match the hostname when accessed directly.
Those web conditions do not prove that every commercial activity has stopped. Infrastructure businesses can operate through private contracts, resellers or another trading site. A parked domain can coexist with active address assignments. It does mean that a prospective customer cannot use the named public domains to verify coverage, tariffs, service levels, acceptable-use terms, support hours, network maps or incident notices.
The regulatory lists add another negative signal. Hong Kong's Office of the Communications Authority publishes a current list of internet service providers, drawing from unified carrier and services-based operator licensees authorised for relevant services. The exact name Hong Kong Business Telecom Limited does not appear. It is also absent from OFCA's downloadable services-based operator licensee list and carrier licensee list at the publication cut-off.
Absence is not a finding that the company has breached a licensing requirement. It could provide a wholesale resource that does not constitute a licensed retail access service, contract through a differently named licensed party, operate outside Hong Kong, or have no current retail service. OFCA's carrier data dictionary is useful because it defines internal fixed services to include broadband internet access, public Wi-Fi and leased lines. A company claiming to own and operate those services in Hong Kong should be able to state the applicable licensee and licence number.
That identity is essential to a bill. A customer needs to know whether the invoice comes from the access owner, a reseller, an address supplier or a managed-service intermediary. The party collecting money may not be able to enter a building riser, dispatch a fibre jointer, replace a rooftop radio or change a route. If the bill says only TELECOM-AS-AP, the customer should ask for the legal contracting entity and every operating party that controls restoration.
The registered office is not a point of presence
APNIC lists the organisation at Suite C1, 6/F, Wing Hing Industrial Building, 14 Hing Yip Street in Hong Kong. That is useful as a contact address. It is not evidence that the building contains a core router, customer aggregation switch, cable landing, network operations centre, warehouse or repair team. Registry country codes and postal addresses describe resource administration more reliably than packet location.
The currently originated /24s reinforce the warning. Their origin networks span different corporate identities and, in third-party geolocation databases, can appear in Hong Kong and other markets. Geolocation is an estimate based on routing, latency, commercial feeds and user observations. It cannot place a rack on a particular floor. A Hong Kong country code in APNIC does not establish that every address is used in Hong Kong, and a low-latency response from Hong Kong does not prove which company owns the server or fibre.
No public facility record ties AS137451 to a named data centre. No exchange membership located for the four registered ASNs shows a live port, speed or switch. No disclosed access map identifies ducts, poles, towers, building entrances or street cabinets. These are meaningful absences because the article's central question concerns the resilience of local connectivity. Without the access medium and handoff location, there is no physical system to audit.
The proper operating boundary may be almost entirely contractual. Hong Kong Business Telecom Limited may control address rights and database maintenance while another network originates the prefix, another carrier provides transit, a data-centre operator supplies power and cross-connects, and a hosting company interacts with the end user. Such chains are common. Their weakness is not the number of parties but the possibility that responsibility falls between them.
Suppose a server on 43.229.153.0/24 becomes unreachable. The route record points to AS153706, while APNIC points to Hong Kong Business Telecom Limited as the allocation holder. Neither record says whether the failure is a server, top-of-rack switch, cross-connect, transit session, route filter, unpaid account or withdrawn authorisation. A support process must identify the party with access to each layer. Otherwise, every supplier can report that its own component is healthy while the customer remains offline.
One historical peer cannot support a diversity claim
Before AS137451 disappeared, Hurricane Electric observed one IPv4 peer: AS64050, BGP Network Limited. That is a historical routing relationship, not a present service. It is still useful because it shows the last visible version of TELECOM-AS-AP was narrow at the edge. A single observed peer means public collectors did not see an independently propagated alternative for those routes at that time.
Collector visibility is incomplete. A network can have private interconnection, a backup session that carries no route until failure, or another provider hidden behind the same path. Yet a claimed redundant design should produce evidence stronger than an invisible possibility. The operator could show two contracts, two router interfaces, physically separate cross-connects, distinct upstream autonomous systems, route advertisements from both, and a failover test under load.
Even two upstream names are not enough. They can share the same fibre entrance, metro conduit, building meet-me room, long-haul carrier or power distribution. A backhoe cut before the paths diverge can remove both. A data-centre power event can stop both border routers. A configuration error can withdraw routes from every provider simultaneously. Route diversity and physical diversity are related but separate controls.
The present address portfolio moves the question down to each origin network. The route for 103.234.54.0/24 through HGC may have a completely different design from 103.234.53.0/24 through CNSERVERS, despite their adjacency in one registered /22. If a customer receives an address from one of them, the relevant upstreams are those behind that exact origin and service location. The other /24s offer no automatic failover because internet hosts cannot simply move between prefixes without configuration, route authority and application changes.
Route authorisation is another dependency. The visible origins need corresponding routing policy and, ideally, valid Resource Public Key Infrastructure authorisations. RPKI helps networks reject an origin that the address holder has not authorised. It does not verify that the authorised operator is solvent, responsive or physically diverse. APNIC's RPKI guidance describes certificates and route origin authorisations as routing-security controls; they are not service-level guarantees.
A route can therefore fail through at least four control points: the origin router stops advertising; an upstream rejects or withdraws the route; an authorisation or filter changes; or the physical circuit carrying the session fails. Restoring only the registry entry will not repair a cut fibre, and repairing fibre will not restore a route rejected by policy. The bill should state who owns each decision.
Field repair is the missing half of the product
Local connectivity becomes real at the point where a cable or radio reaches a customer. Fibre service needs a path through streets, building entries, risers, termination panels and an optical network device. Fixed wireless needs a clear radio path, mounted equipment, tower or rooftop access, power and alignment. Copper or coaxial access has its own cabinets, pairs, amplifiers and environmental limits. TELECOM-AS-AP's public footprint identifies none of these.
That makes repair labour impossible to price from the public evidence. There is no disclosed number of technicians, depot, spare optical units, radios, power supplies or vehicles. There are no published support hours, dispatch zones, mean time to repair or escalation contacts. The absence is particularly important for a small regional provider because labour density often determines restoration more than nominal bandwidth.
A provider can cover a broad area cheaply when failures are rare and installations are scheduled. An outage changes the economics. A technician may need permission to enter a building, a lift to reach a rooftop, a spare matched to an older radio, traffic management for a street cabinet, or a subcontractor authorised to splice carrier fibre. If one crew covers many distant sites, simultaneous faults create a queue. The customer's monthly fee is partly a payment for that standby capacity, whether or not the invoice names it.
The address-resource model can push repair further away from the registered holder. If an origin network uses an allocated /24 in a third-party data centre, remote hands may replace a cable or reboot a device. If the same space serves end users through a reseller, a local access carrier may own the final circuit. Hong Kong Business Telecom Limited might be able to correct a database entry but have no authority to enter either location. A resilient service needs a coordinated path from first alarm to the person holding the correct spare.
Four timings should be separated. Detection is the interval before anyone knows the service has failed. Acknowledgement is the interval before a responsible party accepts the case. Dispatch is the wait for an authorised technician or remote-hands engineer. Restoration is the time until usable service returns. A promise to answer tickets quickly addresses only the second timing. TELECOM-AS-AP publishes none of the four.
Spares are also capacity. A provider may have a working access network but no replacement for a failed optical module, customer router, power injector or radio. A substitute can require reconfiguration, a new mounting bracket or a changed wavelength. Holding every part is expensive, which is why small providers balance inventory against supplier lead times. Customers financing critical connectivity should ask which failure units are stocked inside the service area and which must be shipped across a border.
Power turns route diversity into a physical claim
Every route in the table terminates on powered equipment. At the customer end, an optical network terminal, router, switch or wireless receiver must remain on. At the provider end, aggregation and edge routers need facility power and cooling. Between them, active cabinets or wireless sites may add more power dependencies. An ASN record contains none of this information.
For an office buyer, backup power must be traced across the entire handoff. A generator in a data centre does not keep the customer's router alive. A customer uninterruptible power supply does not keep an unprotected rooftop radio or street cabinet alive. A mobile generator is useful only if connectors, fuel, access and dispatch labour are available. The duration must be stated under realistic load, not only as a battery label.
The split origin model creates several possible power boundaries. The registered holder may maintain route authority from an office while an origin network runs routers in a colocation facility. A separate access provider may terminate the local tail. Each can honestly claim its own equipment has backup while a shared cross-connect room or building riser remains unprotected. Integrated failure testing is the only way to reveal the weakest boundary.
No public material identifies a TELECOM-AS-AP facility, contracted power, generator runtime, dual power feed or tested backup arrangement. It would be wrong to infer these from the capabilities of Hong Kong data centres generally or from the origin network's marketing. Facility-wide redundancy also does not prove a tenant has connected both power supplies, reserved failover ports or purchased diverse cross-connects.
The customer should require a service-specific diagram that begins at its equipment and ends beyond the first independent upstreams. The diagram need not disclose sensitive rack details. It should identify demarcation points, access technologies, powered intermediate nodes, facility boundaries, route origins and restoration owners. Without it, phrases such as carrier grade or redundant are not auditable.
Installed, routed and usable capacity are different numbers
The portfolio permits three tempting but false calculations. Twelve /22s do not equal twelve access networks. Forty-five routed /24s do not equal 11,520 addresses available for new customers. Eighteen IPv4 origins do not equal 18 diverse upstreams. Each number describes a different layer.
Installed capacity is the equipment and circuits physically present. Routed capacity is the address space currently advertised and accepted. Usable service capacity is the throughput and restoration performance available to customers after contention, maintenance reserves and failure headroom. The records give a good point-in-time view of routed identifiers and almost no view of installed or usable service.
Address space can be densely used or mostly idle. A /24 might support 256 public addresses on paper, but network and broadcast conventions, router interfaces, reserved addresses, security policy and customer subnetting reduce the usable total. More importantly, address count says nothing about bits per second. One /24 can sit behind a congested 1 Gbps port while another uses a lightly loaded 100 Gbps fabric.
Capacity under failure is smaller than capacity on a normal day. Two equal links loaded to 60 per cent each cannot absorb one another completely. A backup path may accept routes but lack bandwidth. A wireless sector may reconnect customers after a fibre failure but become unusably congested. A provider selling resilience should disclose normal peak utilisation, surviving capacity after the largest credible failure and the threshold at which it stops accepting orders.
TELECOM-AS-AP has no public traffic graph, port inventory, customer count or utilisation report. Cloudflare Radar and third-party address-intelligence pages may see traffic associated with particular origins, but they cannot reconstruct the holder's contracts or spare capacity. Unofficial estimates of hosted domains, pingable addresses or fraud risk describe the observer's sample. They cannot prove service volume, quality or ownership.
The same caution applies to IPv6. The organisation holds a large 2401:2c80::/32 allocation, currently originated by AS214266 rather than one of its four ASNs. A /32 can be divided into enormous numbers of customer subnets, but design potential is not deployed access. A buyer needs to test whether its actual product receives IPv6, what prefix size is delegated, whether reverse DNS is supported and whether IPv6 survives the same failover as IPv4.
Who is affected depends on which layer the customer bought
The evidence does not support a count of TELECOM-AS-AP retail subscribers. It does show several plausible customer classes with different failure exposure. A holder of an assigned prefix depends on continued authority to use and originate it. A hosting customer depends on the origin operator, facility, server and transit. A business broadband customer, if such a product is supplied through a partner, additionally depends on the local tail and field crew. A reseller depends on every layer and must explain them to its own customers.
An AS137451 outage today would have a different meaning from one in 2024. Because the ASN currently announces nothing, its disappearance from routing is already the baseline; there is no visible customer route to withdraw from it. A failure in the organisation's administrative function could still matter if route permissions, registry maintenance, renewals or abuse handling depend on it. That impact would emerge through other origin networks rather than AS137451 itself.
A failure at one current origin would be narrower but operationally real. If AS153706 stopped announcing 103.48.169.0/24, adjacent 103.48.168.0/24 and 103.48.170.0/24 could remain reachable through their own origins. Hosts in the failed /24 would not automatically move to an adjacent range. Their operators would need another authorised origin, working connectivity and configurations that permit the change.
A failure of the resource holder's contact route can complicate abuse response and emergency coordination. APNIC's live records marked the hknetbroadband.com abuse contact invalid shortly before publication, although the same domain still receives DNS and mail configuration. An invalid validation status is not proof that all messages fail. It is a warning that the registry could not confirm the contact through its normal process. For a portfolio spread among many origins, a working escalation path is a material control.
The parked public site deepens that concern. A customer facing a route leak, denial-of-service incident or mistaken filter needs more than an address on an invoice. It needs a monitored network contact, an abuse contact, a commercial escalation and an out-of-band telephone path. Those contacts should be tested before a fault, not discovered while traffic is down.
The bill needs to identify the service that the records cannot
A credible local-connectivity quotation from TELECOM-AS-AP or a reseller using the name should answer a set of concrete questions. First, what is being sold: address assignment, transit, hosting, a managed router, a wireless tail, dark fibre, lit fibre or a complete internet service? Combining them into the word broadband conceals who controls each failure.
Second, who is the legal service provider and, for Hong Kong access, which current licence covers the offer? If another licensee supplies the tail, the quote should name that dependency and state whether the customer can contact it directly during escalation. A licence number does not guarantee quality, but it resolves responsibility and regulatory scope.
Third, where are the demarcation points? The customer should know whether service ends at an ethernet port, optical terminal, rooftop radio or virtual interface. It should know who supplies and replaces the customer-premises equipment, who powers it and whether configuration can be exported to a spare.
Fourth, what are the route and facility paths? The relevant evidence is not the number of ASNs registered to the resource holder. It is the origins and upstreams for the assigned prefix, the physical facilities where they connect, and the point at which supposedly diverse paths first share conduit, power or equipment. A route view captured during normal operation and a controlled failover provide stronger proof than a network diagram alone.
Fifth, how is repair staffed? The contract should distinguish remote configuration from physical dispatch, state hours and coverage, define the clock start, and name the party holding spares. If a subcontractor performs work, the provider should state whether the restoration target includes the subcontractor's acceptance and travel time.
Sixth, what capacity survives? Committed throughput, latency and packet-loss targets should apply in the normal state and after an agreed failure where resilience is sold. A backup that accepts a route but collapses under customer traffic is not meaningful continuity.
Finally, what happens when the commercial or routing relationship ends? Portable address space can be reassigned or re-originated, but a customer may still have firewall rules, allowlists, DNS records and counterparties tied to an old prefix. Exit terms should provide notice, data and configuration export, renumbering assistance and a period of overlap where feasible. The current fragmentation of the portfolio makes portability planning more important, not less.
Route migration is not service continuity
The portfolio's movement among origin networks shows that routes can change. It does not show that applications, customers or physical access moved without interruption. Bringing a /24 up under a new ASN requires authority, accepted route filters, upstream connectivity and a route that propagates. Keeping the services behind that /24 working can also require moving machines, extending a layer-two network, changing tunnels or rebuilding security controls.
DNS can soften a migration when services accept new addresses, but it does not solve every dependency. Resolvers cache answers. Partners may allowlist the old prefix. Email reputation and reverse DNS can be tied to the previous range. Virtual private networks, certificates, monitoring systems and customer devices may embed addresses. A route that appears in BGP is therefore only one milestone in recovery.
The same limit applies to a backup origin. A second operator can advertise a prefix only if the holder has authorised it and the service can reach that operator. If both origin networks terminate in the same room, depend on the same cross-connect provider or reach the workload through one switch, the route-level alternative may not survive the physical fault. If the workload is copied elsewhere but its data is stale, reachability returns before the business does.
A customer purchasing continuity should ask for a timed exercise that removes the normal path and follows a real transaction from an external network. The result should record route convergence, application availability, degraded capacity and the time required to reverse the change. For a local access circuit, the corresponding exercise may use an independent wireless or second wired tail, but it must also test customer power and equipment. TELECOM-AS-AP publishes no such result.
This is the economic difference between spare ingredients and a resilient product. Portable addresses, alternate origins, spare routers and multiple carriers can all contribute. They create value only when contracts, configurations, physical paths and people have been assembled in advance. Otherwise, migration begins after the outage, when every dependency is already consuming time.
The operating grade must remain weak
There is enough evidence to say that Hong Kong Business Telecom Limited remains associated with valuable internet number resources. The APNIC organisation and allocation records are current. Most component IPv4 /24s are globally routed. The IPv6 /32 is also visible. This is not a claim that the entire footprint is fictitious or abandoned.
There is not enough evidence to say that TELECOM-AS-AP currently operates a regional access network. Its namesake ASN and three sister ASNs announce no space. The public domains do not present an orderable service. The exact company name is absent from current Hong Kong carrier and services-based operator lists. No public source identifies access plant, a service area, customer handoffs, facilities, power design, field crews, spares or restoration performance.
The negative result changes how the title should be read. A local-connectivity bill carrying the TELECOM-AS-AP name would depend on upstream routes and field repair because those functions are not visible inside the registered identity. The live route may belong operationally to one of many other origins. The access tail may belong to an undisclosed licensed carrier. The technician may work for a facility or subcontractor. Until those parties and boundaries are named, the customer is financing a chain it cannot test.
The evidence that would raise the grade is practical: a current product and coverage page; the contracting and licensed entity; an access and facility map; a list of active origins and upstreams by service; valid operational contacts; power and spare policies; a measured failover; and restoration records that separate acknowledgement from repair. None requires disclosing customer identities or sensitive router configuration.
For now, the defensible conclusion is narrow. TELECOM-AS-AP is a durable registry label attached to a dispersed address portfolio, not a publicly demonstrated local network. Its value lies in the resources and permissions around routes. Its resilience, if sold as a service, lies elsewhere: in the origin operator, the physical access path, the power chain and the person who can reach the failed equipment before the customer's business runs out of time.