Summary

  • Internet Talk Radio’s regular 1993 programme was pre-recorded audio distributed as files. Its launch notice said FTP was sufficient and explicitly said IMS was not using the MBONE for the service.
  • The plan put production at one source and invited networks to relay copies through UUNET, regional systems, EUnet, WIDE and IIJ. It did not publish a complete mirror inventory or traffic measurements.
  • A separate live NPR/MBONE special followed in May 1993. That experiment did not turn the weekly “Geek of the Week” file series into a continuous live broadcast.

An announcement before the first file

On 22 March 1993, a message forwarded to RIPE’s mailing list carried Carl Malamud’s plan for a new service. The name was Internet Talk Radio; the first programme would be “Geek of the Week,” an interview show for people in the networking community. The note said it would launch on 31 March. In the surviving programme record, 31 March is also the release date of the first listed episode, an interview with Marshall T. Rose about, among other subjects, OSI, X.400, IETF politics and SNMP version 2. The dates are unusually close because this was not a later story about a distant milestone: the announcement and the programme’s own archive describe the same working service. The RIPE-list announcement and the episode record are the clearest starting points.

Malamud called “radio” a metaphor. The product first being offered was a professionally produced audio file placed on the Internet. A listener could retrieve the file, keep it, and play it later. The March announcement described multicast as one possible way a receiving network might distribute the programme, but immediately drew a boundary: IMS itself was not using the MBONE for Internet Talk Radio. A service could permit a multicast relay without making the show a live multicast transmission.

That distinction changes the opening scene. There was no need for thousands of listeners to tune in at one shared instant. The recording could arrive at a regional site first, then be copied to a local machine; each person chose when to fetch and listen. The service treated a broadcast schedule as a useful cultural analogy while accepting that the computer network gave the listener control over playback.

One programme, several paths through the network

The first show was planned as roughly half an hour. Malamud’s March note estimated 15 megabytes in standard PCM at 8,000 samples per second and 8-bit mu-law. His 1993 IETF presentation put the encoding in network terms—64 kilobits per second—and described a choice of audio formats. A later account by the creator recalled a half-hour file as 30 megabytes. These are not interchangeable measurements: one is the launch announcement’s estimate for a stated encoding, while the other is a retrospective description. No surviving transfer log or file manifest settles a single size for every copy, so the article keeps both claims attributed.

Size made the relay plan part of the programme. The message proposed starting at UUNET and fanning out to regional networks to avoid needless duplicate transfers. It named mcsun at EUnet as an initial European spool point and WIDE and IIJ for distribution in Japan; Alternet customers could use anonymous FTP from UUNET. Local networks were told to contact their service provider, and providers could request instructions. Those details describe an intended operating arrangement, not a verified census of live mirrors. The record does not say that every named network installed a copy, how many listeners used each site, or how much traffic moved.

The surviving IMS archive shows what this arrangement became for readers. The first “Geek of the Week” release is divided into seven audio segments, with links for .au, .gsm and .ra files. Later weekly episodes appear with dates and guest names. By 1994, the index lists interviews with network researchers and engineers, including Steve Crocker, Radia Perlman, Scott Bradner and Tim O’Reilly. In other words, the archive preserved an editorial schedule as well as audio. Its page was a catalogue; the recording itself remained something to fetch in a compatible format. The Geek of the Week index shows the dates, guests and formats.

A file could be staged where the network could afford it. That made the path less dependent on one global stream, but it did not make access equal. At 64 kilobits per second, a 15-megabyte file takes hours over a modem-speed link even before retries and protocol overhead. Breaking the programme into parts made partial retrieval and selection possible; it did not remove the cost of carrying the bytes. A regional copy could reduce repeated long-distance transfers when listeners fetched it locally, but the available sources contain no before-and-after traffic measurements to quantify the savings.

“Multicasting” was an option, not a receipt

The initials IMS stood for Internet Multicasting Service, and the same organization did experiment with live audio. That is not evidence that every IMS channel used the same delivery method. In a separate May 1993 announcement, Malamud described an NPR “Science Friday” programme recorded on 21 May and released on 31 May. The special joined live national radio to the MBONE and allowed Internet participants to ask questions. The announcement identifies the channel as Internet Town Hall, distinct from the regular Internet Talk Radio series. The archived notice documents that live event and its later release date.

The distinction survives in the service’s later history. The IMS archive grew to include National Press Club events, public-affairs programmes, TechNation, NASA material and other channels. A 1994 Washington Post account describes plans for separate live feeds from the House and Senate alongside an archive service. Those later projects show an organization trying different audio uses; they do not rewrite how “Geek of the Week” was first delivered. The contemporary record is more specific than a retrospective label: one service could publish files, arrange live special events and keep recordings for later retrieval.

The launch note also tried to make distribution acceptable to networks with their own rules. It promised short, National Public Radio-style acknowledgements for sponsors Sun Microsystems and O’Reilly & Associates. Malamud said at least two large government networks had indicated that the planned service complied with their appropriate-use policies. The message did not name them or publish formal approvals. It is evidence that policy questions mattered to the project, not evidence of a universal waiver.

The IETF slides likewise described a rights model in which noncommercial copying could be unlimited while derivative works remained restricted. An audio file was technically easy to copy; the right to redistribute every recording still depended on terms.

A distribution design, not a claim of universal access

The 1993 experiment joined one editorial source to many possible receivers. IMS chose and produced the programme; regional networks could choose a protocol, encoding or local staging arrangement; listeners chose when to retrieve the file. The design separated the person who made content from the institutions that carried it. That flexibility was an engineering advantage, not proof that all the affected networks had authorized the same service or that all listeners could reach it.

Malamud’s later retrospective says the service had listeners in a few dozen countries by the end of 1993 and was using substantial bandwidth. That is a useful participant account of ambition and load, but it is not an audited audience survey. The launch proposal named regional contacts without listing completed mirrors; the surviving archive proves that programmes and format links existed, not that a particular listener downloaded a particular copy. The May live event proves a separate MBONE use. None of those facts supplies a full audience denominator.

This is why Internet Talk Radio matters as a history of the Internet. In 1993, “radio on the Internet” did not have to mean a single stream pushed to a mass audience. It could mean a recording converted into files, staged by networks that volunteered capacity, and played on the listener’s own schedule. That model exposed the Internet’s practical strengths—copying, multiple formats, regional cooperation and user control—alongside its limits: a large file could still be costly to retrieve, rights still had to be negotiated, and an announced relay plan was not the same thing as measured reach.

The strongest historical claim is therefore modest. Internet Talk Radio made an Internet-native programme available as a set of retrievable recordings, and its operators invited other networks to help distribute them. The evidence does not require calling it the first station, a universal broadcaster or an early version of any later product. What it shows is an early service built around a choice that remains recognizable: send one live signal, or let each receiver decide which file to fetch, from where, and when.

Records that define the boundaries

The distinctions above can be checked against the original announcement, surviving programme records, later recollection and separate live-event evidence: