Summary
- Noakhali Internet Exchange holds AS139875 plus an ASSIGNED PORTABLE /24 and /48 in APNIC, yet the route object on its /24 names AS139009 as origin and the route6 on its /48 names AS136151; neither record names its own ASN as origin.
- Its own ASN is observed announcing three IPv4 prefixes that APNIC registers to Today Vision and Lakshmipur Online, while its self-maintained PeeringDB entry asserts 6,000 IPv4 and 36,000 IPv6 prefixes against 0.0% visibility among RIS full peers at one dated instant.
- Holdings, route origin, collector observation and self-description are four records answering four different questions. The article prints them side by side and converts none into another.
Two dates sit two days apart in the APNIC registry. AS139875 was registered at 08:38:30 UTC on 11 December 2019; the exchange's /24 was registered at 07:07:17 UTC on 13 December 2019 (aut-num AS139875; 103.146.134.0/24). Six years on, both still exist. Neither is routed under the ASN registered first.
I went looking for the other story — the one where a directory row names an "internet exchange" and no numbers follow, membership running ahead of build-out. The records refuse it. This entity holds an ASN, an ASSIGNED PORTABLE /24 and an ASSIGNED PORTABLE /48, all in APNIC under its own organisation handle, ORG-NIE2-AP. The finding is not absence.
It is a separation the premise did not anticipate: the blocks the exchange holds are routed under two other networks' ASNs, the prefixes its own ASN is observed announcing are registered to two more organisations again, and the prefix count it publishes about itself is orders of magnitude away from anything observed on the wire. Holdings, origin, observation and self-assertion are four records, and none of them converts into another.
What APNIC says it holds
The aut-num record is short and precise: handle AS139875, name NIE-AS-AP, country BD, status active, registration 2019-12-11T08:38:30Z, last changed 2024-12-21T07:25:30Z, and exactly one remark, a description reading "Noakhali Internet Exchange". The registrant is ORG-NIE2-AP; the administrative and technical contact is a separate group object rather than the registrant organisation itself; the abuse role object carries a remark dated 2026-07-07 recording that its mailbox was validated (APNIC RDAP aut-num AS139875). The response contains no route object and no route6 object. Read on its own, it says who holds the ASN and says nothing about what the ASN originates.
The IPv4 block is equally plain: inetnum 103.146.134.0 – 103.146.134.255, netname NIE-BD, description remark "Noakhali Internet Exchange", country BD, status active, type ASSIGNED PORTABLE, registration 2019-12-13T07:07:17Z, last changed 2024-12-21T07:36:11Z, holder ORG-NIE2-AP (RDAP ip 103.146.134.0). APNIC's RDAP prefix responses carry no route objects, so this record also ends without an origin.
The IPv6 block reads the same way in mirrored whois text: inet6num 2001:df2:6080::/48, netname NIE-BD, description "Noakhali Internet Exchange", holder ORG-NIE2-AP, ASSIGNED PORTABLE, last modified 2024-12-21T07:28:08Z (mirrored whois for the /48). That last date is a mirror's last-modified field and not an allocation date, and the APNIC RDAP record for the /48 was not retrieved in this reading.
Three objects, one holder, no route attached inside the registry response itself.
What the routes name
The route objects live outside APNIC's RDAP prefix responses and had to be read from whois text reproduced by other sites. On the exchange's own /24 the object reads: route 103.146.134.0/24, description "WinStream", origin AS139009, maintained by MAINT-NIE-BD, last modified 2025-06-15T11:15:20Z (mirrored whois for 103.146.134.0/24). On its own /48: route6 2001:df2:6080::/48, description "Jatrabari Online", origin AS136151, maintained by MAINT-NIE-BD, last modified 2025-06-15T11:16:15Z (mirrored whois for 2001:df2:6080::/48). Neither object names AS139875.
The inverse arrangement appears one prefix to the west. For 103.135.78.0/24 the whois text shows a route object with description "LKN", origin AS139875, maintained by MAINT-LAKSHMIPURONLINE-BD, last modified 2025-06-24T16:42:17Z, served under a heading that pairs that prefix with that origin (mirrored whois for 103.135.78.0/24). That is the only place in anything read for this article where a route object names AS139875 as origin.
The maintainer fields deserve reporting exactly and no further: the route object and route6 on the exchange-branded blocks sit under MAINT-NIE-BD, while the one route object naming its ASN sits under the Lakshmipur Online maintainer. A maintainer is a label attached to an object. It is not a person who can be shown to have approved anything at a stated hour, and nothing here treats it as one.
What the ASN is seen announcing
The routing observation for AS139875 prints two counters that do not agree with each other: "Prefixes Originated (all): 1" and "Prefixes Announced (all): 3", with RPKI counters of 0 valid and 1 invalid. The announced table lists 103.99.248.0/24 under the label "Today Vision" and both 103.135.76.0/24 and 103.135.78.0/24 under "Lakshmipur Online" (bgp.he.net AS139875). Every one of those three prefixes is registered in APNIC to someone else. 103.99.248.0/24 is ALLOCATED NON-PORTABLE, netname TDV-BD, description "Today Vision", registration 2017-09-11T11:39:06Z, with no registrant organisation entity in the response (RDAP ip 103.99.248.0). The other two sit inside a single /22: 103.135.76.0 – 103.135.79.255, netname LAKSHMIPURONLINE-BD, holder ORG-LO1-AP, ALLOCATED PORTABLE, registered 2019-02-22T04:01:51Z. There is no separate /24 inetnum for either address, and a query for 103.135.78.0 returns the same /22 object (RDAP ip 103.135.76.0).
The arithmetic inside the RPKI block is mine, not the page's: 0 valid plus 1 invalid equals 1, which is the originated count rather than the announced count of 3. The RPKI block therefore describes one prefix while the table lists three. That is a statement about which counter the certificate status is attached to, and nothing more.
The page is also unstable in one specific field. Two retrievals of the same URL returned different peer counts: one showed a single peer labelled Today Vision, the other showed two, AS138636 and AS136948. Neither reading is adopted here, because a number that changes between two reads of one URL is evidence about the instrument before it is evidence about the network.
The count it publishes about itself
PeeringDB, whose records networks maintain themselves, prints for this entry: organisation "Noakhali Broadband Network", also known as "NIX-BD", long name "Noakhali Internet Exchange", ASN 139875, IRR as-set AS139875:AS-ALL, network types "Route Server", IPv4 prefixes 6000, IPv6 prefixes 36000, peering policy "Open" (PeeringDB net/21789). Its Public Peering Exchange Points table prints "No filter matches", while facility rows extracted from the page name BDIX Main Node - Panthapath and NIX-BD, the latter spelled the page's own way.
Six thousand plus thirty-six thousand is forty-two thousand. That sum is my arithmetic: PeeringDB prints the two components and no total, and defines neither. Set beside three announced IPv4 prefixes, the 6,000 would be two thousand times the observed figure if both counted the same object. No document read for this article defines either counted object, so the two numbers are printed side by side and no reconciliation is asserted. A self-maintained profile records what an operator asserted. It is not a capacity measurement, a coverage figure, a growth rate, or a denominator for anyone's share.
One dated instant of visibility
RIPEstat gives its observation with the instant attached: at 2026-10-30 00:00, AS139875 was visible to 0.0% of 323 IPv4 and 0.0% of 312 IPv6 RIS full peers; last seen 2026-09-27 08:00 announcing 103.135.76.0/24; "Does not originate prefixes visible in RIS"; "Not seen in RIS as transiting" (RIPEstat AS139875).
That is one measurement at one timestamp by a stated set of collectors. It says what those 323 and 312 peers saw then. It does not say the network is idle, that nothing is connected to it, or that the exchange has no traffic; no member list, switch list, port count or traffic figure was obtained for this article at all.
Two retrieval routes that disagree
The IRR mirror reports "registered in source · 1 route objects originated (0 v4 / 1 v6)" for AS139875, listing 2404:76c0::/48 with source APNIC (IRRd3 AS139875). That is incompatible with the mirrored whois text showing an AS139875-origin route object for 103.135.78.0/24.
Both readings are printed; neither is adopted as the complete inventory. The registry text and the IRR listing cannot both be complete, and no document read explains the difference.
The country frame, used narrowly
The national picture comes from one dated survey: an APNIC Blog piece of 26 March 2025 records that Bangladesh "currently has eight operational Internet Exchange Points (IXPs), all licensed by the Bangladesh Telecommunication Regulatory Commission (BTRC) as National Internet Exchanges (NIXs)" and that "only 7.57% of registered networks peering at NIXs" (APNIC Blog, 26 March 2025). It does not name Noakhali Internet Exchange, and that silence is context: it is not evidence about this exchange's licensing status, which nothing read here establishes one way or the other. The 7.57 per cent is a country-level share dated March 2025 and is never used here as a measurement of this network.
The APNIC public member directory table carries columns for Name, Membership category and Economy, but the retrieved text stops at an entry beginning with the numeral 2, before the alphabetical M entries (APNIC member directory). No membership category and no economy for this entity are reported in this article, because none was observed.
Where the accountability actually sits
The separation documented above has a consequence that is easy to miss: a registry entry, a route object, a collector observation and a self-maintained profile are four different claims, produced by four different processes, and each answers for something different.
A registry record answers a registration question: which party is recognised as holding this identifier. A route object records a declared origin and the maintainer label attached to that declaration; it does not, by itself, show who approved the declaration or on whose instruction. A collector records what one vantage saw at one time. A self-maintained profile records what the operator chose to assert. When those four disagree, the disagreement is a verification trigger — not a verdict, and not a finding about anyone's intent.
Nothing read here establishes why any of these records differ, and no motive is alleged anywhere in this article.
What a reader can do with the split is narrower and more useful than a scandal: ask for the authorisation. A route object naming AS139009 or AS136151 as origin for these blocks is a declaration; an authorisation, routing agreement or contract naming those origins would be the explanation. The registry cannot supply that document, the collector cannot, and the profile cannot.
What remains unknown, in plain terms
Whether Noakhali Internet Exchange operates switches, a fabric, a route server or a single port is not established here. No member list, switch list, traffic record or peering session was obtained. Whether the announcements described above were made under any arrangement with the registrants of those prefixes is not established either: an RPKI-invalid status and an origin differing from the registrant are observed statuses, and no document read characterises intent. Whether AS139009 or AS136151 holds an authorisation covering the exchange's own blocks is likewise unestablished.
And nothing here infers ownership, control or location from personal names, postal addresses or maintainer handles.
Boundaries of this reading
- The reading is bounded to thirteen documents retrieved on 20 November 2026 between 22:33Z and 22:51Z, each retained as a snapshot; most were read as page text rather than through a live browser session, so dynamic elements or footnotes may sit outside the retained text.
- APNIC's whois lookup page for AS139875 was not readable in this reading. Registry values come from APNIC's own RDAP responses for the aut-num and four prefixes; route-object values come from mirrors of whois text, because APNIC's RDAP prefix responses carry no route objects.
- Every route-origin statement rests on a mirror or another site's quotation of whois text; the APNIC whois service itself was not queried directly.
- The routing observation is point-in-time and internally ambiguous: "Prefixes Originated (all): 1" sits beside "Prefixes Announced (all): 3", and the observed peer count differed between two retrievals of the same URL.
- PeeringDB's figures are self-asserted, are reported exactly as printed, and are never used here to compute a share, a growth rate or a population.
- The 0.0% is a visibility share among 323 IPv4 and 312 IPv6 RIS full peers at 2026-10-30 00:00 — one dated instant measuring what those peers saw.
- No party's conduct is alleged. An RPKI-invalid status is an observed status, and one network appearing to announce prefixes registered to another is described as the records show it, nothing more.
- Nothing here states whether the exchange operates switches, a fabric, a route server or a port.
- The APNIC member directory row for this entity was not reached, so its membership category and economy are not reported.
- The APNIC RDAP record for 2001:df2:6080::/48 was not retrieved; the /48's inet6num and its route6 are read from mirrored whois text only.
- Whether an APNIC route object for 103.99.248.0/24 naming AS139875 exists could not be checked; only the IRR mirror's listing was available, and it shows none for AS139875.
- Dates are kept to their fields: 2019-12-11T08:38:30Z and 2019-12-13T07:07:17Z are RDAP registration events, while the /48's 2024-12-21 date is a mirror's last-modified field and not an allocation date.
- The March 2025 APNIC Blog survey is used for the national picture only; its silence about this exchange is context, never proof of licensing status.
- Prior BTW coverage of 4 November 2026 recorded the directory row and set falsifiable tests for rows of this kind. This article runs that test on this entity, reports that its working expectation — a member with no attributed numbers — fails here, and states the failure in the body rather than quietly revising the expectation.
An exchange can hold its numbers and be routed through someone else's ASN, and the four records can be read without any one of them becoming the others. The fact that would change this picture is not a better adjective; it is a document — a dated route-object history, or an authorisation naming AS139009 or AS136151 as approved origin for these blocks. Until one appears, the split itself is the finding, and it is worth more than a guess about what sits behind it.
Member Briefing
Read the full trend briefing
Members read the full briefing with the source notes behind every finding.
Strategic Circle
A daily briefing in your language, built from changes we have verified.
Join for USD 20/monthCancel anytime.
Members also get the full research library, a daily briefing on the networks they follow and the private forum.






