Summary

  • LINX’s story has several beginnings: a January 1994 pilot announcement, an August meeting, first traffic in November and a company incorporated in December 1995.
  • The exchange’s operating surface, its members’ coordination and its later legal entity were related, but they were not the same thing.

Analysis

On 8 November 1994, according to LINX’s anniversary history, traffic passed between networks at Telehouse in London’s Docklands. The company now known as London Internet Exchange Limited did not yet exist as an incorporated company. Companies House records its incorporation more than a year later, on 14 December 1995.

That gap changes how the beginning should be told. An Internet exchange is first a place where networks can interconnect. A company can own or steward that shared facility, but the legal form is not the packet path itself. LINX’s early chronology shows the technical arrangement moving before its permanent corporate wrapper.

The first public step came earlier still. On 24 January 1994, PIPEX and JANET announced resources for a pilot London Internet Exchange. The notice invited interested service providers and proposed charging only marginal line and connection costs during 1994. It also described a route server as a plan, dependent on the necessary software and infrastructure. The message therefore records an intended pilot, not proof that every proposed component was already operational.

LINX later placed its first member meeting on 5 August 1994 at the University of London Computer Centre. Its 2020 account says the meeting preceded LINX’s formal establishment that November. Then, on 8 November, the anniversary account says the first network-to-network traffic crossed at Telehouse. These dates mark different transitions: discussion, an exchange arrangement and working traffic.

The physical scale was modest. LINX’s institutional history describes a switch supplied by PIPEX and installed at Telehouse; its account identifies a Cisco Catalyst 1200 with eight 10-megabit ports. The 2019 anniversary article says the equipment was donated and that no legal contracts were in place. The original purpose, LINX says, was to keep UK traffic local instead of sending it across the Atlantic and back, adding cost and delay.

That purpose should not be inflated into a claim that LINX carried all domestic British traffic or eliminated transatlantic routing. The evidence establishes a local exchange and a reason for building it. It does not provide a complete 1994 route table, a traffic series or a measure of how much transit was avoided. Nor does a switch, by itself, decide which routes a network will announce or accept. Those decisions remain with the participating networks; the public histories reviewed here do not reconstruct their initial policies.

The legal company arrived after the traffic. Companies House identifies London Internet Exchange Limited as a private company limited by guarantee without share capital, incorporated on 14 December 1995. The register pins down the date of the legal entity. It does not tell us that the exchange had been ungoverned until then.

LINX’s own history says a not-for-profit, member-run model was agreed from the beginning. Its 2019 anniversary account likewise celebrates a neutral, not-for-profit approach. Those are meaningful institutional recollections, but they are not the same as the surviving January announcement or the company record. The January notice supplies evidence of a marginal-cost pilot; it does not spell out a complete member constitution. The sources do not identify who held every operational decision during the interval before incorporation.

Even the founding roster deserves care. LINX’s 2019 anniversary page names five founding members: PIPEX, BT Internet Services, Demon Internet Services, EUnet GB and UKERNA/JANET. The current history page narrates the participants differently, foregrounding PIPEX, Demon, UKnet and UKERNA while noting BT’s interest. That variation is a reason to attribute the lists, not to collapse them into a single unquestioned roll call.

The deeper significance is institutional. A service can become operationally useful before the institution that later formalizes it. In LINX’s case, networks had a shared physical point, a stated cost rationale and a developing member forum before a company was registered. The exchange was not simply “community-owned” because people met, nor was it merely a private switch because one participant supplied equipment. Authority sat across several surfaces: equipment and hosting, each network’s routing decisions, and the members’ emerging rules.

That division also created exposure. A shared facility depends on whoever can maintain and expand it; member coordination determines who can participate and on what terms; each network still controls what it routes. Incorporation can make responsibilities durable, but it cannot retroactively prove that every early participant had equal influence. The public record leaves that question open.

LINX did not begin on one date. A January plan, an August meeting, November traffic and a December 1995 company record describe a sequence in which operation preceded incorporation. Keeping those layers distinct makes the history more precise—and makes clear that a network’s authority is found in the controls that actually shape its paths, not in a single anniversary label.

Sources