A price list for joining the internet's routing core
Start with the two prices, because everything else in this story hangs between them.
The official rate for a seat at the internet's coordination table in Europe is set by the RIPE NCC, the Amsterdam-based registry that allocates addresses and autonomous system numbers for Europe, the Middle East and Central Asia. Its Charging Scheme 2026 puts the annual fee for a member account at EUR 1,800, on top of a one-time EUR 1,000 sign-up fee. That is the wholesale door: become a Local Internet Registry, deal with the NCC directly, and hold resources in your own name with no intermediary.
The retail door costs a thirtieth of that. Openfactory GmbH, a Swiss network operator from Frauenfeld, publishes a tariff on its FREETRANSIT project page: it will sponsor a globally unique autonomous system number for EUR 90 of setup and EUR 60 per year thereafter. Of that EUR 60, EUR 50 is simply the NCC's own per-ASN fee passed through — the charging scheme bills sponsoring members EUR 50 a year for each ASN they maintain on behalf of an end user. The sponsor's gross margin on the recurring piece is roughly the price of two coffees. A competing sponsor, LIR Services, charges EUR 150 a year with no setup fee and throws in a /48 of IPv6 space. Both are list prices, published on the suppliers' own pages and current as of this writing; no transaction documents exist in this market because the deals are too small to leave any.
Those two published figures — EUR 1,800 a year direct, EUR 60 to 150 a year sponsored — are the price pair this essay turns on. They describe the same underlying good: the right to originate routes on the public internet under your own number. The 30-to-1 spread between them is not arbitrage waiting to be closed. It is a deliberate structure, built into registry policy, that lets individuals hold real routing resources without carrying the fixed costs of registry membership. And thousands of Europeans have walked through the cheaper door. The RIPE NCC's own staff counted 4,391 autonomous system numbers registered to natural persons as of September 2024 — about one number in nine, against a regional total of roughly 38,600 implied by the registry's companion review, which records the 1,700 out-of-region holders as 4.4 percent of all ASNs — and reported that end-user ASN requests had doubled between 2019 and 2024, from about 500 per half-year to about 1,000.
One of those 4,391 numbers belongs to a research network in León, in Spain's northwest. It has no customers, no revenue, and as far as the public record shows, no ambition to acquire either. That makes it nearly ideal for the question this piece asks. When a market good is stripped of every commercial motive, what remains is the pure economics of the thing itself: what it costs to exist, and what its owner gets in exchange.
An entity that is a person, and why the registry says so
The name in the directory is finitud labs research network, recorded as a company in Spain. The public record says something more precise, and the precision matters.
The name appears verbatim in exactly one place: the network's PeeringDB entry, the industry's self-service database of interconnection data, where it is registered as an Educational/Research network with an open peering policy, a website at as214930.net, and an operator listed under the long-form name of the holder. The RIPE database tells the rest. The autonomous system object, created on 10 May 2024, carries the network name FINITUD-LABS-AS and points to an organisation object registered to a natural person — Alice Wyan Garcia Martin, with an address in León — whose company-registration field reads, in the registry's own words, "Not Applicable." The organisation type is OTHER, not LIR: the holder is not a registry member but a sponsored end user, with Openfactory GmbH recorded as the sponsoring member.
Spanish corporate records agree by their silence. Full-text searches of the Boletín Oficial del Registro Mercantil — the state gazette in which every Spanish company formation, dissolution and officer change must be published — return no entity named finitud labs, either through the gazette's own archive at boe.es or through the open mirrors that index it. There is no Sociedad Limitada behind the name, no tax identity, no filed accounts. That is not a gap in diligence; it is the finding. The "company" in the directory is one person's laboratory, operating under a project name, and every subsequent judgement in this essay flows from that fact. The directory's category — a regional internet service provider — describes the shape of the record, not the substance of the operation. Nothing in the routing table, the registry, or the operator's own pages suggests service provision to anyone.
The operator's site says so plainly. labs.finitud.org describes finitud labs as "a collection of personal projects relating to computers, networking, electronics and radio," and lists four of them: the global autonomous system, a twin on the dn42 experimental network, a LoRa satellite ground station, and an APRS position-reporting gateway in León. The network's own page is a single sentence of self-description — "an education network operated by Alice Wyan. It is currently IPv6-only" — followed by peering details. The word finitud is Spanish for finitude. As corporate names go, a one-person network called "finiteness" is at least honest about its scale.
Identity, then, reconciles cleanly, which is itself unusual in this directory's evidence class: one person, one project name, one domain matching the alias on file, one registry organisation object, one PeeringDB record, all mutually consistent, all created within a two-week window in May 2024. What does not exist — a company — is the most economically informative absence in the file.
What two years of routing announcements reveal
Registry paperwork describes intentions. The routing table records behaviour, and for this network it records twenty-six months of it.
The ASN was assigned on 10 May 2024. RIPEstat's routing history shows the first prefix — a /48 of IPv6 space carved from Openfactory's own allocation — appearing in the global table five days later. A PeeringDB record followed within the week. By mid-June 2024 the network was announcing an IPv4 /24 as well, and by the following month it had settled into roughly the shape it holds today: four IPv6 blocks and one IPv4 block, visible worldwide.
The composition of those five announcements is a tour of the hobbyist supply chain. The /48 comes from the sponsor's space. A /40 arrives via HyeHost, a small hosting operation whose parent /32 is registered, in turn, under yet another micro-provider's maintainer handle. A /41 comes from space held by Cloudie Networks, a US LLC. A second /40 descends through a /36 sub-allocation labelled, in the registry, as an "australia network." Four blocks, four chains of custody, spanning at least four countries before reaching a bedroom in León — not because the operator needs the diversity, but because in this market IPv6 space is so abundant that micro-providers hand out /40s the way conference booths hand out stickers. The IPv4 story is the mirror image. The single /24, 44.32.90.0/24, is not purchased at all: it sits inside the amateur-radio block registered to Amateur Radio Digital Communications, the Californian foundation that has stewarded network 44 since the early 1980s and assigns it to licensed radio amateurs free of charge, with written permission required before any block is announced to the global table. At a moment when a leased IPv4 /24 commands serious monthly money on the open market, the only IPv4 this network touches costs nothing and cannot be sold — an elegant sidestep of the scarcest commodity in networking.
The upstream arrangements complete the picture. The autonomous system object lists import and export relationships with six networks: Openfactory's FREETRANSIT project, the cloud host Vultr, two networks belonging to the Swiss host iFog, the tunnel-based exchange fabric BGP.Exchange, and Vanvik Internet, a Norwegian sole trader's network. Three of those six publish a price for what they provide, and the price is zero. FREETRANSIT offers free transit to non-commercial networks, explicitly without service guarantees. BGP.Exchange states that membership and tunnels are free of charge and asks only for an ASN and an email address to register. FogIXP, iFog's exchange platform — 414 connected networks on PeeringDB's count — declares 1-and-10-gigabit ports "free of cost in core locations" until further notice.
Physical footprint is observable, within limits. The network's public looking glass — a standard bird-lg installation of the kind mid-sized carriers run — lists five router nodes: Frankfurt, León, Madrid, Helsinki and St. Louis. Consulted twice on 3 July 2026, hours apart, it told two different stories: on the first pass the León, Madrid and St. Louis nodes refused queries or timed out; by the second, all five answered, and the Frankfurt node showed live sessions to FogIXP's route servers, iFog transit, the FREETRANSIT service, BGP.Exchange peers and a cluster of dn42 neighbours. An earlier description of the network listed Amsterdam, Helsinki and León as its points of presence, so nodes have come, gone and moved over the two years — consistent with the routing history, which shows /48s announced for a few weeks in mid-2024 and then withdrawn as larger blocks arrived, two /44s carried for nearly two years and withdrawn in June 2026, and even a two-day flirtation with a /39 of ARIN-region space in June 2024. This is what laboratory traffic looks like in a global routing table: constant, low-stakes renumbering that a commercial network could never afford, because renumbering a network with customers is surgery, and renumbering a network of one is a weekend.
For completeness: the same operator runs a parallel autonomous system, AS4242423606, on dn42, the closed experimental network where addressing is free and mistakes are consequence-free. The progression — dn42 first, global table second — is the standard curriculum of this subculture, and the fact that both registrations share a maintainer handle makes the learning path legible in public data.
The arithmetic of a network with no customers
Now assemble the cash flows, keeping evidence and inference strictly separated.
The recurring cost that is documented beyond argument is the ASN sponsorship. Openfactory publishes EUR 90 setup and EUR 60 per year; the RIPE NCC's charging scheme confirms the EUR 50 per-ASN annual fee that sits inside it. Whether this operator pays Openfactory's list price is not public — sponsors waive fees for friends, and this corner of the internet runs substantially on favours — but list price is the conservative assumption, and EUR 60 a year is the ceiling for the resource that makes everything else possible.
Address space, at the observed scale, is free to nominal. The /48 travels with the sponsorship. The 44net /24 is free by policy to licensed amateurs — the operator's lab page lists an amateur-radio station and a position-reporting gateway in León, consistent with holding a licence — and ARDC charges nothing for the letter of authority that permits announcement. The /40s and /41 from HyeHost, Cloudie and the Australian chain carry no published price; in this market such blocks are commonly gifted or bundled with services costing a few euros, but that is a characterisation of the segment, not a verified invoice, and it is flagged here as exactly that. A single-sourced corroboration: LIR Services bundles a free /48 with its EUR 150 sponsorship, which indicates where the market clears — IPv6 space at this scale is a giveaway, not a product.
Transit and interconnection are free at the margin, by published tariff rather than by favour. FREETRANSIT's zero, BGP.Exchange's zero and FogIXP's zero-until-further-notice are all documented on the suppliers' own pages, cited above. The network's six upstream relationships include at most two that plausibly involve money: Vultr, and iFog's paid tunnel service, whose price is not published on any page this research could reach — the attempt was made against both the live site and its archived copies, and only a FAQ without prices survives — so no number is asserted for it.
Compute is the one line with a hard published price and a genuine inference gap. Vultr's price list — the live page refuses automated readers, so the figure comes from an archived June 2026 copy — starts at USD 2.50 a month for an IPv6-only instance with one processor core and half a gigabyte of memory, and Vultr permits customers to announce their own prefixes over BGP from such instances at no stated extra charge, subject to an ownership check. An IPv6-only network run by a cost-conscious hobbyist on Vultr is, with high likelihood, sitting on that plan or near it; that the instance exists is evidenced by the upstream relationship, while its size is inference. The remaining nodes — Helsinki, St. Louis, the Spanish pair — run on hosts whose identities and prices the public record does not settle. The honest bound is this: if all five looking-glass nodes were rented at Vultr's entry price, the compute bill would be USD 150 a year; the León node more likely rides a residential line whose cost exists regardless of the hobby, and the true rented-node count on the reporting day was somewhere between two and five.
Add a domain registration of order EUR 10 and the totals converge. Bottom-up, at list prices: EUR 60 of sponsorship, roughly EUR 30 of documented compute at the observed minimum of one paid instance, zero for transit, zero for exchange ports, zero for the /24, zero to nominal for IPv6 space — call it EUR 100 a year, with a defensible upper bound near EUR 250 if every node is rented and every supplier bills at list. Top-down, as a cross-check by an independent method: the all-in sponsored bundles on the open market (LIR Services at EUR 150 a year including number and address space; Openfactory at EUR 60 plus free transit) imply that a complete laboratory of this design clears between EUR 100 and EUR 200 a year before compute preferences. The two methods, built from different documents, reconcile on the same band. The gym-membership comparison in this essay's opening is not rhetorical: EUR 100 to 250 a year is precisely the range of a municipal gym in a provincial Spanish city.
Revenue, for completeness, is zero on every visible axis: no commercial page, no price list, no customer testimony, no job postings, no marketplace presence, and a registry footprint that would have to change shape before a single euro could lawfully flow through it. There is no earnings band to triangulate because there are no earnings. What the arithmetic actually establishes is the inverse finding: the entire cost side of a globally routed, five-city, dual-registry network now fits inside the discretionary budget of a single salaried adult, with a factor-of-two uncertainty that depends mostly on how many virtual machines she chooses to keep running.
Suppliers who sell for almost nothing, and why
A EUR 60 product with a EUR 50 pass-through inside it raises the obvious question: why does anyone bother supplying this market?
The answer visible in the record is that the suppliers are the hobby, industrialised. Openfactory — sponsor of this network and operator of FREETRANSIT — was already the second most prolific sponsor of personal ASNs in an independent 2022 census, with 53 to its name, behind Securebit's 60 and ahead of SnapServ's 45. iFog, whose exchange fabric and transit tunnels appear throughout this network's sessions, ranked in the same census's top ten. These are small Swiss companies whose commercial businesses — colocation, transit, VPS hosting — sit one shelf above their free tiers, and the free tier functions as the cheapest customer-acquisition engine in networking: every hobbyist who takes a free port or a EUR 60 sponsorship becomes a config-literate prospect for paid services, a contributor to the exchange's network effect, and occasionally an employee. The sponsored-ASN business, taken alone, is barely a business. A RIPE Labs analysis found the top fifty sponsoring registries averaging 135 sponsored ASNs each — at least 6,750 in total — which at a EUR 10 net margin per number implies that even the largest specialists are earning restaurant-bill money from sponsorship fees themselves. The margin is not in the fee. It is in the relationship, the funnel, and in some cases the donations.
That structure has a fragility the buyers understand imperfectly. Free tiers are revocable by construction: FogIXP's port pricing is free "until further notice" by its own wording, FREETRANSIT explicitly offers no service guarantee, and the sub-allocated IPv6 blocks in this network's announcements depend on the continued goodwill — and continued registry standing — of at least three intermediary companies on three continents. The registry record of this one network already shows the churn: address blocks announced, withdrawn and replaced as suppliers changed over 2024 to 2026. For a laboratory, that churn is tuition. For anything resembling a business, it would be an availability crisis, which is one clean way to state why this tier of the market stays non-commercial: the supply chain's price is low precisely because its promises are.
Supplier concentration deserves its own line, because it is the nearest thing this network has to a balance-sheet risk. One Swiss company — Openfactory — is simultaneously the sponsoring member of record, the source of the foundational /48, and the operator of a free transit service the network uses. If that single relationship soured or the company failed, the ASN would need a new sponsor within the NCC's grace procedures, the /48 would be renumbered away, and a transit session would drop. None of that is fatal — re-sponsoring is a routine, priced transaction across the market, as the competing tariffs cited above show — but it is the correct answer to the diligence question "what single counterparty can disturb this operation most," and it is knowable entirely from public registry attributes.
What the fee actually buys
The demand side is the more interesting economics, because the good being purchased is not connectivity. Nobody needs an autonomous system to reach the internet; a EUR 20 residential line does that better than five VPSes ever will. What EUR 100-a-year buyers are purchasing divides into three assets, none of which appears on any invoice.
The first is non-simulated experience. Routing on the global table differs from routing in a simulator in the same way flying differs from a flight simulator: the failure states are real, other operators' route filters are real, the registry paperwork — route objects, reverse delegation, RPKI signing — is real, and mistakes propagate to strangers. This network's public artefacts show the full curriculum being worked through: consistent route6 objects across four address suppliers, an as-set registered for peer filtering, a looking glass and a status page maintained for an audience of peers, and the dn42 twin where the syllabus started. The RIPE NCC's registry chief has described the growth of exactly this cohort — natural persons registering numbers "to learn BGP" or to help with IPv6 deployment — as exponential since 2021. The learning is the product. Everything else is packaging.
The second asset is a credential, and the labour market prices it even if no registry does. A personal autonomous system is a standing, publicly verifiable demonstration that its holder can design, deploy and operate multi-site routing — verifiable not by reference check but by typing a number into any route collector on earth. In a hiring market where network engineering interviews lean heavily on claimed experience, a candidate who can say "query my ASN" has converted EUR 100 a year into a portfolio piece that cannot be faked and does not expire. The signal's value is impossible to measure directly from public data, and this essay does not pretend to; what can be said from the record is that the cohort paying for the signal has grown from roughly 750 identifiable personal networks in 2022 to a natural-person registry population of 4,391 by September 2024, while the price of admission fell. Goods whose demand rises as their price collapses are usually being under-produced or newly discovered; hobbyist routing looks like the latter.
The third asset is membership. The maintainer references on this network's registry objects — a half-dozen handles belonging to other micro-operators who can update or vouch for pieces of the setup — sketch a mutual-aid web that behaves like a guild: space is gifted, tunnels are traded, route objects are co-maintained. The economics of guilds are old and well understood; members pre-pay in favours and collect in opportunities. The registry simply makes this one legible to outsiders.
Against these assets stands one genuine cost the buyers rarely price: privacy. A natural person who holds registry resources places a legal name and a service address into a public database forever queried by strangers — the registry community itself has flagged the tension between accuracy obligations and the reasonable reluctance of individuals to be doxxed by design. The operator studied here publishes under the registry's rules like everyone else; this essay uses the record's name once, for identity reconciliation, and otherwise treats her as what the record shows: an operator. But the structural point stands — the hobbyist tier's true price is EUR 100 plus a permanent entry in a public ledger, and the second component is the one no sponsor can discount.
Competition, substitutes, and the line where a lab becomes a business
Every market this cheap invites the question of substitutes, and hobbyist routing has three.
The closest substitute is dn42, the invitation-free overlay network where the same protocols run over tunnels with private numbering, at a price of exactly zero. This operator, like most, runs both in parallel — but the fact that thousands still pay for the global version when the free version exists is itself market information: simulated stakes are not stakes, and the EUR 60-a-year premium over dn42 is the observed market price of consequence. The second substitute is the training industry — vendor certifications and lab courses — which sells structured knowledge without public accountability; the routing cohort's growth suggests the two are complements more than rivals, the course teaching vocabulary and the ASN teaching judgement. The third substitute is employment itself: waiting until a job supplies the routing experience. The 500-to-1,000 doubling in end-user number requests says a growing slice of the profession has decided not to wait.
Competition among suppliers, meanwhile, has already driven the price to its structural floor. With the NCC collecting EUR 50 of every sponsored-ASN fee, no sponsor can durably charge much less than Openfactory's EUR 60 without paying to serve, and the observed spread — EUR 60 to EUR 150 across the published tariffs cited above — is a spread in bundling and hand-holding, not in the underlying good. When a market's retail price sits within pennies of its regulated wholesale input, the commercial story is over; what persists is infrastructure philanthropy with a business-development rationale.
Which leaves the boundary question the assignment poses: where does a learning network end and a business waiting to happen begin? The record of this network offers a crisp answer, because every attribute points the same way. A business would hold its resources through a legal entity to survive its founder's circumstances; this one is a natural person with "Not Applicable" in the company field. A business would buy guaranteed transit; this one rides tariffs that promise nothing. A business would keep its address space stable; this one renumbers casually. A business would sell something; this one's only public offer is an open peering policy, which is to say, a standing invitation to exchange traffic for free. The distance between this laboratory and a minimal commercial ISP is not technical — the routing is the same routing — it is contractual: a registered company, paid transit with remedies, provider-independent space, and customers. Each of those steps has a published price, and the sum of them is roughly the gap between EUR 100 a year and the EUR 1,800-a-year membership tier, which is the registry's own quiet way of pricing the difference between learning and selling.
Reading the quiet signals
Thin-file subjects reward attention to small signals, and this file contains several worth reading carefully, none of them official.
The operator's blog has exactly one post, titled "Applying for an ASN," dated twelve days after the number was assigned — and its body is empty, a zero-minute read, a title with no text beneath it for two years. The lab site's network page still describes a three-city footprint that the looking glass has long since traded for a different five. Three of the five looking-glass nodes refused queries at one point on the reporting day, then answered hours later. The PeeringDB record declares capacity for ten IPv4 and ten IPv6 prefixes; the table carries five in total. Two /44 announcements that had run for nearly two years were withdrawn a few weeks before this essay's date.
A commercial analyst would read decline into that list. The better reading, given the segment, is ordinary hobbyist entropy: documentation is the first casualty of any personal project, nodes in a consequence-free network drop and return on their operator's schedule rather than a customer's, and declared prefix capacity in PeeringDB is aspiration, not audit. The June 2026 withdrawals coincide with continued announcement of the underlying /40s, which suggests consolidation rather than shutdown; the ASN's fee was evidently paid through the current cycle, since unwanted numbers now leave the registry at scale — 150 to 200 handed back per month once per-ASN charging was announced, the highest rate the registry had recorded — while this one remains registered and announced. Silence in this market is compatible with contentment.
What the signals genuinely leave open is trajectory. An expanding laboratory and a quietly wound-down one look identical from outside for months at a time. Three observations would settle it in either direction: whether the looking glass's five-node roster holds, shrinks or grows across the coming months; the next annual fee cycle, which now removes lapsed hobbyist numbers visibly and without delay; and any change to the one-post blog, in either direction — a burst of documentation being the classic mark of a project entering its teaching phase, and a lapsed domain the classic mark of the other thing. None of these requires privileged access to check, which is one of the quiet virtues of studying a subject whose entire existence is public by construction.
The fees, votes and policies that could reprice the hobby
For a network with no revenue, this one carries a surprising amount of regulatory exposure — all of it upstream, in the institutions whose pricing decisions created the niche.
The nearest event is already booked. At the NCC's general meeting in Edinburgh this May, members chose the flat-fee option for the 2027 charging scheme — EUR 1,894 per member account, first ASN included, EUR 50 for each additional — over a category model that would have priced accounts by resource holdings, and the result was announced to members as a near-even split, 51 to 49. The direct effect on sponsored hobbyists is small; the signal is not. A membership that divides that evenly over how to distribute costs is a membership still arguing about who subsidises whom, and the hobbyist tier — thousands of EUR 50 pass-through fees riding on infrastructure funded by EUR 1,800 memberships — is an obvious line item for a future proposal to revisit. The per-ASN fee itself only arrived with the 2025 scheme, and its first documented effect preceded the first invoice: with charging announced, 150 to 200 numbers a month came back — the registry's way of discovering how many of its assignments had been worth less than EUR 50 a year to their holders. Every future euro added to that fee will re-run the same experiment at a higher strike price.
Policy risk runs deeper than fees. The NCC convened an Open House on personal numbering in September 2024 and reported the disagreements unresolved: whether hobbyist use fits assignment policy at all, whether natural-person records degrade registry accuracy, and what to do about the roughly 1,700 numbers registered to holders outside the service region entirely — 700-plus in the United States, some 300 in China — for whom a European registry number has become a mail-order collectible. A community critique published the same summer accused sponsoring registries of hollowing out the multihoming requirement with pre-filled forms, and the registry's own staff floated the question of whether the requirement still serves a purpose. The hobbyist tier exists inside that unresolved argument. A future policy round could formalise it — a cheap, honest personal-use category — or could tighten eligibility and price the marginal learner back out. Both outcomes have advocates on the record.
The operational risks are the free tier's own terms, already described: transit without guarantees, ports free until notice, address space held at three removes. The geopolitical risks are minimal in the way only genuinely small things achieve — an IPv6-only laboratory with an amateur-radio /24 attracts no sanctions exposure, no interception obligations, and no infrastructure-security designation. On the national layer, Spanish telecommunications law regulates providers of public electronic communications services, with registration duties at the regulator for those offering service to the public; a non-commercial personal laboratory with no subscribers sits outside that perimeter on any plain reading, and consistent with that reading, no Spanish regulatory filing for this operation was found — an interpretation offered here as analysis, since no regulator has had occasion to rule on it. The one legal thread that does bind is the amateur-radio licence behind the 44net block: ARDC's permission regime means the /24 exists at the pleasure of a Californian foundation's policy, and the foundation's process for direct announcements explicitly reserves scrutiny for exactly this kind of use. Lose the licence or the permission, and the network loses its only IPv4 — which, for an IPv6-only education network, would be an inconvenience approximately the size of a footnote.
What would change this judgement
The judgement, stated flatly: finitud labs research network is a genuine one-person educational autonomous system, representative of a several-thousand-strong European cohort, operating at a documented cost of very roughly EUR 100 to 250 a year, with zero revenue, zero customers, real but revocable free-tier dependencies, and a value proposition — skill formation, credential, community — that the labour market prices even though no invoice ever will. The evidence for identity is strong, for costs moderate and bounded, and for trajectory weak in both directions.
Specific, checkable facts would move each pillar. A Spanish company formation naming the operator or the project — one gazette entry would do it — would convert the analysis from hobby economics to startup economics overnight, and the boundary section above already prices that conversion. The appearance of provider-independent address space, paid transit contracts, or a second sponsoring relationship in the registry objects would signal the same shift from the supply side. In the opposite direction, a lapsed domain, a deregistered number at the next fee cycle, or the /24 returning to ARDC's free pool would mark the laboratory's orderly end, and the only story left would be the cohort's. On the phenomenon itself, the numbers to watch are the NCC's: another doubling in end-user requests would confirm that learn-by-routing is becoming a standard career stage rather than a subculture; a policy change from the unresolved personal-numbering debate — in either the formalising or the restricting direction — would reprice the entire tier at a stroke; and any future charging scheme that moves the EUR 50 pass-through meaningfully would test, at population scale, exactly how much Europeans will pay to keep their names in the routing table. The present answer, documented across every tariff this essay could reach, is: about as much as the gym, and apparently worth it.
Evidence register
The RIPE database objects for the autonomous system and its holder (aut-num, organisation) carry the identity, sponsorship, upstream and date facts. RIPEstat supplies announcement history and current prefixes. The PeeringDB record documents the network's name, type, policy and exchange presence. The operator's pages (as214930.net, labs.finitud.org, the network page, the empty first post) establish self-description, project scope and the documentation signals, and the looking glass evidences the five-node footprint and session states. Pricing rests on primary tariffs: the RIPE NCC Charging Scheme 2026 and adopted 2027 flat-fee document with the members-list result thread; Openfactory's FREETRANSIT sponsorship tariff; the LIR Services tariff; FogIXP's port terms; BGP.Exchange's free-membership terms; Vultr's archived June 2026 price list and BGP conditions; and the 44Net FAQ with ARIN's registration record and the ARDC allocation help page for the amateur-radio /24. The cohort statistics come from the RIPE NCC's Open House deck (the 4,391 natural-person count, September 2024) and its RIPE 89 review (out-of-region holdings, monthly returns), the registry chief's half-year figures, the sponsoring-registry analysis on RIPE Labs, and an independent 2022 census of personal networks; PeeringDB's exchange record sizes the free exchange fabric. The dn42 registry entry documents the experimental twin. Spanish corporate-registry searches through the state gazette and its open mirrors returned no company, an outcome reported in the identity section above.