In 1998, twelve months after deploying RFC 2058 RADIUS for dial-up roaming and issuing a combined calling-card and dial-up roaming product on a single PIN, Tempest Telecommunications attempted to extend the same architecture to a third service: public Internet kiosks. The vehicle was a custom protocol Tempest named Public Access Terminal Networks (PATN). The intent was a roaming network for kiosks — a customer with a prepaid PATN credential would walk up to any participating kiosk in any country, authenticate, get a session, and have per-minute charges deducted from a shared balance, with settlement reconciled between Tempest and the kiosk operator. The effort failed commercially. PATN never reached the scale required to make federation valuable, and the underlying per-operator kiosk billing product that survived was a far smaller business than the network-effect product Tempest had hoped to build.

The Road Ahead inspiration

Bill Gates published The Road Ahead in November 1995. The book's most-cited passages described “friction-free capitalism” — a near-future world in which information access, payments, and transactions would happen without the intermediaries and frictions that defined the existing economy. Among the specific examples Gates discussed was the public information terminal: a screen, in a public place, that any passing person could use to look up flight times, hotel rates, or local information, and that would handle whatever micropayments the interaction required. The terminals were a small portion of the book, but they caught the imagination of a generation of late-1990s technologists, alongside the broader predictions about wallet PCs, telecommuting, and the disintermediation of retail.

The Road Ahead was theory. The implementation question — what would actually have to be built to make Gates's public-terminal vision work commercially — was an open one through the late 1990s. Several distinct components were required: hardware (the terminals themselves), a billing platform (per-minute or per-transaction settlement), an identity layer (so a customer's balance and preferences travelled with them between terminals), and a payment rail (to top up the balance and settle with terminal operators). Of those four, the billing platform and identity layer were structurally similar to the AAA infrastructure Tempest had just finished standing up for ISP roaming in 1997. PATN was the attempt to apply the same shape to the kiosk problem.

The 1998 kiosk landscape

In 1998 the public Internet kiosk industry was almost entirely unaggregated. The branded cyber-café chains that would later define the category — EasyEverything (Stelios Haji-Ioannou's London flagship opened in June 1999; the Times Square store followed in November 2000), Cyberia, Easy Café, Net@ — either did not exist yet or were one- or two-store operations. The vast majority of public Internet access in 1998 came from independent cyber cafés in university towns and tourist districts, hotel business-center kiosks, copy-shop add-ons (Kinko's in the United States, similar chains in Europe and Asia), library and university computer rooms, and a long tail of one-off implementations at airports, ferry terminals, and conference centers.

Each of these operators ran its own billing. A typical independent cyber café sold time in fifteen- or thirty-minute blocks at the counter; the customer paid in cash and was issued a slip with a username and password to log into a workstation. Hotel business centers charged sessions to the room. Copy-shop kiosks took coins or credit-card swipes. The economics were local: the operator's costs were the rent on the terminal floor space, the workstation hardware depreciation, and a flat-rate Internet line; the operator's revenue was per-minute retail pricing against walk-up demand. There was no reason for any of these operators to participate in a federation that would route a portion of their per-minute revenue back to a roaming clearinghouse, unless the federation brought them traffic they would not otherwise have had.

The PATN protocol

PATN, as Tempest specified it, was the kiosk analogue of the RADIUS-based dial-up roaming network Tempest had launched the previous year. A participating kiosk operator installed a Tempest-supplied software agent on the kiosk workstations. A travelling customer carrying a PATN credential — in practice the same prepaid PIN that authenticated the customer's dial-up Internet sessions and PSTN voice calls — walked up to a participating kiosk, entered the PIN, and was authenticated against Tempest's central account database. The kiosk agent reported session start and stop events back to Tempest in real time. The customer's balance, shared across voice, dial-up, and kiosk service types, decremented per minute of kiosk session as it had for the other two. The kiosk operator's portion of the per-minute revenue was reconciled monthly through a settlement statement, mirroring the inter-ISP settlement model Tempest already operated for RADIUS roaming partners.

The technical pieces fit together. Tempest's existing balance ledger handled three service types instead of two. The accounting reconciliation that already settled dial-up minutes between member ISPs settled kiosk minutes between member kiosk operators using the same bilateral-statement mechanics. The customer-side experience was the natural extension of the combined calling-card-and-dial-up product: one credential, now usable in three modes instead of two. The architectural insight that had made the 1997 product compelling — a single prepaid balance, accessible across heterogeneous service types — carried over cleanly.

Why the federation failed

The technical extension was sound. The commercial extension was not. The 1997 dial-up roaming network worked because the ISP industry had pre-aggregated through years of consolidation and through the roaming clearinghouses (iPass, GRIC) that had been brokering authentication between member ISPs since 1996; signing up an ISP to a roaming federation, by 1997, was a known motion with a known payoff. The 1998 kiosk industry had no equivalent pre-aggregation. Each cyber café, hotel business center, and copy-shop kiosk had to be signed up individually. The payoff was uncertain: roaming traffic on top of an operator's walk-up demand might be meaningful, or it might be negligible, depending on whether anyone outside the operator's local catchment carried a PATN credential.

The customer side faced a mirror version of the same problem. The value of a PATN credential to a traveller was a function of how many kiosks accepted it. With a small footprint, the credential was useless in practice. Without traveller demand, the kiosk operator had no reason to install the agent. The federation depended on simultaneously solving both sides of the two-sided market, in an industry that had not yet consolidated to the point where any single rollout could reach a useful number of endpoints. The marketplaces that would have made PATN worth joining — the branded chains, the city-center kiosk clusters, the airport concession networks — emerged a year or two later, by which time Tempest had moved on and the broader cyber-café industry was already approaching its own decline.

A secondary cause was the friction in the kiosk-side software install itself. Dial-up roaming required only a configuration change on the customer's laptop (a phone number to dial, a username to present). Kiosk roaming required physical access to the operator's machines, installation of a software agent on each workstation, and a maintenance commitment to keep the agent running across the operator's typical machine churn. Operators with a dozen workstations weighed the install effort against an unknown traffic upside and, predictably, deferred.

What survived

The PATN federation never reached commercial scale. The underlying kiosk-billing component — the software that handled per-minute billing on a single operator's workstations, without the cross-operator roaming layer — shipped as a standalone product to individual kiosk operators starting in 1999. Stripped of the federation ambition, what remained was a competent per-operator kiosk billing platform: account creation at the counter, time-blocked sessions, per-minute decrementing balances, end-of-session printout. The product served the operators that adopted it for several years. It was not the network-effect business Tempest had planned to build. It was a point solution that happened to be one of the early commercial implementations of a Gates-flavored micropayment concept — deployed in 1999, three years before the IPO of PayPal in 2002, ten years before the launch of Square in 2009, and eleven years before the founding of Stripe in 2010.

Reading the failure

The PATN effort is a specific dated instance of a pattern that recurs in technology history: the right architectural idea, the wrong infrastructure layer, the wrong moment in the underlying market's consolidation curve. The 1997 RADIUS roaming network worked because the ISP industry was aggregated enough to federate. The 1997 combined calling-card-and-dial-up product worked because Tempest controlled both endpoints. The 1998 PATN kiosk network failed because the kiosk industry had not yet aggregated enough to make federation valuable, and by the time it did the broader category was already in decline. The protocol was not the problem. The category was.

The frictionless micropayments idea Gates had described in The Road Ahead arrived in the consumer market through entirely different form factors: not roaming public-access terminals but mobile wallets, NFC tap-to-pay, QR-code payments, and embedded card-on-file flows. The closest direct descendant of PATN's federated-balance concept may be the modern carrier eSIM marketplace and stored-value travel cards, which solve the same problem — one credential, many networks, one balance — against a market that has, finally, aggregated enough to make the federation valuable.

Sources and further reading

This page is part of an ongoing historical archive of the 1989–2012 international telecom industry, maintained by Jason Jacoby, a former operator at Interglobe (UK phone cards) and Tempest Telecommunications. Corrections and additions welcome via the contact page.