Blog

• News
• The Shelf Now Starts in 1981
• IETF 126
• Browsing the New Corpora
• Internet Talk Radio
• New on the Shelves
• The IHI Reading Room
• Collection Growth
• Envisioning an Internet Data Trust
• PingER Rescue, Continued
• IHI in The Register
• APRICOT 2026 Keynote
• Internet Data Trust Workshop in Barcelona
• CAPIF 4 in Almaty
• Spring Meetings
• Baltic Sea Cable Cuts
• Fall Talks
• Meetings
• PingER Rescue
• Internet Exchange Point History Collection
• Berkman Klein Fellowship
• IHI Joins DNS-OARC
• Mirroring RIPE RIS
• Initial resource index
• IHI assigned NAAN 26338
• Welcome to InternetHistoryInitiative.org
• Releases
• Practice Notes
• Fiber route maps in the conference record
• Reading a journal that has no text
• Reconstructing the missing decks
• Using collection context to reduce errors in automated transcription of historical audio

News

The Shelf Now Starts in 1981

The collection’s operator record now reaches back to 1981. We’ve accessioned the TCP-IP list — 39,123 messages, 13 October 1981 through December 1994, from the securitydigest.org archive — along with its digest edition (1981–1983) and the utzoo Usenet tapes, whose unix-wizards traffic begins in June 1981. It is indexed and searchable in the Reading Room alongside the later lists, so the arguments of the ARPANET-to-Internet transition can be read next to the operator communities that inherited them.

IETF 126

IETF 126 met in Vienna, 18–24 July. The collection accessioned its presentation materials as the meeting proceeded; session recordings and minutes follow after adjournment.

Browsing the New Corpora

In June we promised more on how to browse the newest accessions. The Cypherpunks main list (1992–2016), PacNOG, AfNOG, and W3C www-talk are now indexed and searchable in the Reading Room alongside the rest of the collection. The Africa Internet Summit record (2017–2026) is live as well, with per-year meeting pages and nearly two hundred talk transcripts drawn from the community’s own video archive.

Internet Talk Radio

We’ve accessioned the Internet Talk Radio collection — Carl Malamud’s pioneering audio program (1993–1996), 567 items mirrored from archive.org. To make the audio searchable and citable we’re generating transcripts, using the collection itself to correct the errors that speech recognition makes on the period’s names and acronyms. How that works is the subject of our first practice note.

New on the Shelves

The catalog has added some ninety Works over the past two months. A few worth browsing: the Soviet Coup Internet Archive — the Relcom/Demos traffic that carried news in and out of Moscow during the August 1991 coup, while the broadcast media were seized; ConneXions: The Interoperability Report (1987–1996) and its editorial descendant, the Internet Protocol Journal (1998–present), both edited by Ole Jacobsen; the Internet Experiment Notes (1977–1982), Jon Postel’s RFC-parallel series from the DARPA Internet project; and the Three-Network TCP Demonstration of November 22, 1977, when TCP traffic from a packet-radio van on a Bay Area freeway crossed PRNET, ARPANET, and SATNET on a single end-to-end path.

The IHI Reading Room

We’ve soft-launched the IHI Reading Room, a prototype public home for the collection we’ve been assembling: the mailing lists, conference talks, and documents in which the Internet’s operator and standards communities argued the network into existence.

The holdings currently include nearly four million messages of mailing-list correspondence (NANOG, the TCP-IP list, Internet-History, Interesting-People, the regional NOGs, the IETF lists, and many more), tens of thousands of conference presentations and transcripts, and a hand-curated catalog of Works, Authorities, and open research Inquiries, each with a durable ARK identifier under NAAN 26338.

You can search the shelves directly, or ask the librarian a real question and watch it read the primary sources. There’s also an open API and an orientation for AI agents — research assistants are welcome, and their corrections enter the curation queue with honest provenance.

While it’s in prototype, access is by recognition rather than accounts: no passwords, just an arrival link. Get in touch and we’ll send you one. It’s a construction zone, updated daily — have a look, and tell us what’s wrong. That’s what the correction queue is for.

Collection Growth

Recent accessions to the IHI collection: the Cypherpunks main list (1992–2016), PacNOG, AfNOG, and W3C www-talk, with Africa Internet Summit materials (2017–2026) staged for processing. More soon on how to browse all of this.

Envisioning an Internet Data Trust

The writeup promised after December’s Barcelona workshop is now available. Envisioning an Internet Data Trust (January 2026) sets out the charter and operational principles for a public-interest institution dedicated to century-scale preservation of the Internet’s historical datasets, and defines the priority tiers that now organize the IHI catalog. Published under CC BY 4.0; comments welcome.

PingER Rescue, Continued

An update on the PingER rescue: a first tranche of historical PingER measurement data has been transferred from SLAC into IHI custody, verified, and accessioned into the IHI catalog (ark:/26338/w/pinger). The project’s servers have since been shut down, and there is no longer anyone at SLAC to ask about the rest — so what we recovered may be all that survives. Decades of end-to-end measurements to and from the Global South, no longer at risk of disappearing entirely.

IHI in The Register

The Register covered the initiative this week: Internet history is vanishing. Researchers want to save it, by Simon Sharwood — the APRICOT keynote, the replica-hosting model, and the PingER recovery. Nice to see the story reaching a wider audience.

APRICOT 2026 Keynote

Jim delivered a keynote at APRICOT 2026 in Jakarta: Internet History Initiative: Preserving our Collective Data Legacy (slides). The Internet’s operational exhaust — routing tables, measurements, mailing lists, meeting archives — is disappearing faster than we are preserving it, and the fix is institutional as much as technical. The concrete ask to the room: IXPs, NRENs, and universities willing to host replicas. Get in touch if that could be you.

Internet Data Trust Workshop in Barcelona

As part of the Internet Society’s inaugural Pulse Research Week in Barcelona (December 8–11), ISOC Pulse and IHI co-organized a hybrid workshop on December 10: Internet Data Trust: Preserving the Internet’s Historical Data Legacy. More than sixty members of the global Internet measurement community, joined by professional librarians and archivists, shared their institutional experiences and gave feedback on an initial set of community principles. The afternoon covered institutional perspectives on preservation, how to support the custodians we have, and the research horizons that open up if the record survives — working toward a public-interest institution dedicated to century-scale preservation of the Internet’s historical datasets. Thanks to the Internet Society Pulse team for hosting the conversation. A writeup of the ideas is in progress; more soon.

CAPIF 4 in Almaty

At CAPIF 4 in Almaty, Kazakhstan, Jim presented DNS Without Borders: Uncovering Regional Hubs and Dependencies in K-Root Traffic (slides), work with the RIPE NCC continuing the DNS watersheds theme: where Central Asia’s root-server queries actually flow, and what that says about regional interdependence. A recap of the meeting is on RIPE Labs.

Spring Meetings

Two talks this spring. At RIPE SEE 13 in Sofia: Assessing the Internet in the SEE Region (slides). And at ZANOG25 in Durban: Contemplating Internet History: Where Do We Go Next? (slides) — the preservation story, told for the operator communities whose mailing lists and meeting archives are the history in question.

Baltic Sea Cable Cuts

Together with Emile Aben of the RIPE NCC, Jim co-authored a RIPE Labs analysis of November’s Baltic Sea submarine cable cuts: A Deep Dive Into the Baltic Sea Cable Cuts. A reminder that today’s measurements are tomorrow’s history: reconstructing events like these depends on the survival of the measurement archives.

Fall Talks

Slides from two fall presentations are now available. At the RIPE 89 MAT Working Group in Prague: Measuring and Visualizing DNS Watersheds, on how geography and network proximity shape where DNS queries flow. And at the IRTF GAIA session at IETF 121 in Dublin: Internet History Initiative: Overview and Exhortation, a tour of endangered and extinct measurement projects, and a plea to preserve what remains.

Meetings

Jim Cowie will be attending DNS-OARC43 and RIPE89 in Prague the week of October 25th, and IETF121 in Dublin the first week in November. Have a story to tell about preserving (or failing to preserve) the Internet’s historical datasets? Stop by and say hi!

PingER Rescue

After the shutdown of the PingER project at SLAC, decades of measurement data dating back to the late 1990s were feared lost. We’ve contacted SLAC IT and are in the process of recovering and archiving this collection for posterity. PingER was notable for its emphasis on creating measurements both
to and from institutions in the Global South, and for its early exploration of possible relationships between national development and quality of Internet connectivity. More details to follow!

Internet Exchange Point History Collection

New link to the IXP History Collection, an ongoing project which seeks to record and document global histories of computer networking and internet exchange points (IXPs).

Berkman Klein Fellowship

Jim Cowie will be joining Harvard’s Berkman Klein Center as a 2024-2025 Fellow, working with the Harvard Law Library Innovation Lab (LIL) to pursue the Internet History Initiative research themes. Read more here.

IHI Joins DNS-OARC

The Internet History Initiative has officially joined DNS-OARC . The history of DNS operations is, in a real sense, the history of the Internet. We’re looking forward to connecting with the DNS-OARC community to explore and celebrate this legacy in coming years!

Mirroring RIPE RIS

We’ve completed our first offsite preservation mirror of the RIPE RIS historical BGP datasets, all collectors, from inception, and continue to mirror with a 1 day lag. More sources available under Sources.

Initial resource index

First sets of links to public data repositories are now available under Sources. Please
suggest additional public data resources that you’d like to see linked here.

IHI assigned NAAN 26338

The ARK Alliance has assigned the Internet History Initiative a unique Name Assigning Authority Number (NAAN) for use in its issuance of unique ARKs,
joining more than 1200 participating institutions in the ARK Alliance.

Welcome to InternetHistoryInitiative.org

Feel free to get in touch if you or your institution would like to learn more about participating in the preservation and curation of the Internet infrastructure’s historical datasets.

Releases

Practice Notes

Short writeups of methods we use in the collection work, recorded so that others can reuse or improve them.

Fiber route maps in the conference record

This note is also available as a typeset PDF.

Operators have been putting maps of their fiber builds on conference slides for thirty years. Most of those maps were shown once and filed away. This note describes finding them in the collection, judging which of them carry real geographic information, and tracing that information back out into standard geographic form — KML aligned against present-day OpenStreetMap data.

From roughly 1.2 million archived slide and document pages, a text pass over the page descriptions produced 3,806 candidates; a vision triage kept the 1,708 that actually depict fiber or cable routes rather than logical topology diagrams; and a stricter screen — could a cartographer produce a useful KML by tracing this image against real map data? — kept 54. Of those 54, nine now align to draft KML: three fully automated, five with a librarian reading the map and placing control points by hand, one model-assisted. The rest have their routes traced in pixel coordinates and are waiting for a person to place three or four intersections.

The tracing is ordinary image processing, and it works because the material is ordinary: a presenter draws their fiber in one saturated color over a lighter basemap. Georeferencing is the part that resisted automation longest, and the note is candid about why. A vision model asked to name control points across a whole slide drifts into plausible invention — it proposed a tidy series of numbered-street crossings, most of which were not printed on the map at all, and which were collinear, so they could not have constrained the transform even if they had been real. Asking about each quarter of the map separately, verifying every proposed point against a small crop, and pooling several rounds fixed most of that.

Two traps are recorded for anyone attempting this work. Street grids are self-similar: pure geometry matching locked confidently onto an alignment 25% off in scale, with every road landing on a road, just not the right one. And with enough junk control points, some subset will agree — a set of local exchange names that geocoded, unbounded, to random places worldwide produced a fit with innocent-looking residuals. Consistency alone is not evidence. A smaller lesson, but a real one for archives: landmark names rot. An arena named on a slide from the 2010s had been renamed by 2025 and no longer resolves. Street intersections are the durable currency.

These maps are historical documents, not engineering data. A route on a conference slide is what an operator chose to present, at a moment in time, at whatever simplification served the talk, and alignment accuracy is bounded by the slide’s own drawing accuracy long before it is bounded by any fitting. Nothing here has been field-verified and none of it should be relied on for anything operational. Each map links back to its source talk so the geometry can be read in its original context.

Geocoding and reference geometry are © OpenStreetMap contributors, ODbL. We would welcome collaboration on any of this.

Reading a journal that has no text

This note is also available as a typeset PDF.

ConneXions — The Interoperability Report (1987–1996) came to the collection as page scans with no text layer at all: 115 issues, 3,573 pages, unreadable to search and to everything downstream of search. The pages are 1990s desktop publishing — a wide margin column carrying side-headings, a main body column, boxed insets, mastheads, rules, and figures. Handing a whole page like that to a vision model works badly: the reading order scrambles across the columns, and figure axis labels arrive in the middle of the prose as token soup.

What worked was to infer the page’s rectangles first, transcribe each region separately at crop resolution, and reassemble in reading order with the region structure tagged. The segmentation is classical image processing rather than machine learning — ink profiles locate the column gutter, decorative rules are excluded before the gutter is found, scan borders are masked. The model is then given a vocabulary of the journal’s own era, drawn from the collection itself, as a prior: protocol names, organizations, the names of the people writing. On a controlled comparison of a single region, that prompt recovered 22% more body text than a neutral one and turned a figure from noise into a description a reader can use. The whole corpus cost about two hours across two mid-size vision-language model instances, roughly 65 seconds a page.

There is a symmetry in the material worth noticing. ConneXions documented the protocols of the young Internet and was produced with the desktop-publishing tools of its own moment — the PageMaker-and-PostScript era that put multi-column layout and margin apparatus within reach of a small trade publisher. The features that made the journal handsome in 1990 are exactly the features that defeated optical character recognition for the following three decades.

Two things in the note matter more than the method. The first is a survey of forty years of prior art that we built without and should have read first: every segmentation rule we derived empirically was already documented, with better tooling attached. The second is an audit of our own output — sixteen reviewing agents compared 791 sampled pages against their stored extractions, grading archival fidelity and asking whether a blind patron could reconstruct the page from our record alone. Of the 424 ConneXions pages sampled, 95.8% graded A or B. The audit also caught a fleet-scale defect that no aggregate statistic had surfaced: a generation ceiling was silently truncating about 7,100 dense pages mid-transcription, before their figure descriptions were emitted — deleting precisely the content the method exists to produce. All flagged pages were re-run at a higher ceiling the same day.

Reconstructing the missing decks

This note is also available as a typeset PDF.

Of the 11,688 video recordings in the collection’s talk holdings, 11,240 — ninety-six percent — have no accompanying slide deck. The deck is usually where the substance is: the numbers, the network diagrams, the timeline the speaker is narrating. A video without its deck is searchable only by its title. The archive knows a talk happened, but not what it said.

The method has two steps. The first is a hunt: a per-talk web search for the original deck, escalating to event-page fetches and Wayback lookups. When the talk comes off the conference circuit this often works on the first query, and it doubles as a metadata audit — one recovered deck exposed a misspelled presenter name in our own catalog. When the source is a streaming channel the hunt mostly reports honest absence: of forty talks from one such channel, one deck found, twelve confident verdicts that no deck ever existed (panels, welcome remarks, an award induction), and twenty-seven misses. A well-evidenced “probably no deck” is a finished piece of work, not a failure.

Failing the hunt, we reconstruct the deck from the video. Sample a frame every five seconds, cluster the frames into slide-states by perceptual hash, keep the sharpest frame of each state, and drop states too short to be slides — those are camera cuts. Each surviving page records the timestamp span over which it was visible, so every reconstructed page links back to the second of video where the speaker is saying it. About four minutes of CPU per talk.

A vision pass over 1,023 reconstructed pages found the population divides three ways: 282 switched-feed slides, where the production cut to the deck; 347 projections in the scene, where the room camera happens to see the screen; and 377 pages with no slide at all. The projection class was the surprise — talks whose metadata said “panel” turned out to carry legible keystoned decks on the room’s screen.

The finding we did not expect: for the one talk where we hold both the found deck and the reconstruction, the reconstruction is the better artifact. The archived PDF is a print-resolution two-up handout; the video’s switched feed shows each full slide at 720p. Reconstruction is not only a fallback for missing decks. For switched-feed recordings it can be the archive’s best edition, with the found deck serving as the authority on slide count and order.

A reconstructed deck is a derived artifact and must never pass as the original. Each carries the source video’s URL and hash, the full method string, per-page timestamp spans, the model that transcribed each page, and a verdict distinguishing a reconstruction from a confirmed absence of slides. In the catalog they bind at receipt tier, honest about what they are.

Using collection context to reduce errors in automated transcription of historical audio

This note is also available as a typeset PDF.

Background

We recently accessioned the Internet Talk Radio collection (Carl Malamud, 1993–1996; 567 audio items mirrored from archive.org). To make the audio searchable and citable we need transcripts. We ran a single-episode pilot to test a transcription workflow before processing the full collection, using the first “Geek of the Week” episode: an interview with Marshall T. Rose recorded March 31, 1993 (48 minutes).

The error pattern

Automated speech recognition handles ordinary speech in this material well, but systematically mis-renders the specialized vocabulary of the period: acronyms, protocol names, product names, and personal names. The model substitutes phonetically plausible guesses. From the raw output of this episode:

• “ISO-DE” for ISODE (a software package the guest authored)
• “the IEPF” for the IETF (six occurrences)
• “X400” for X.400 (six occurrences)
• “Marty Shostal” for Marty Schoffstall
• “Keith McLaughery” for Keith McCloghrie

Name errors are the most serious for our purposes, since names are access points. These errors are not correctable by a general-purpose model working alone: nothing in the audio distinguishes “McLaughery” from “McCloghrie” without outside knowledge of who worked with the guest.

Method

The workflow has two stages. Both use information already held in the collection.

Stage 1: speech recognition with a vocabulary prompt. We transcribed the audio with Whisper (large-v3 model, one GPU). Whisper accepts a short free-text prompt that biases its vocabulary. We filled it with the episode metadata (show, guest, interviewer, date) and roughly forty period terms. This reduces, but does not eliminate, the errors above. The output — 741 timestamped segments — is saved unchanged as the raw transcript.

Stage 2: constrained correction using a brief derived from the collection. We then passed the transcript in chunks to a language model (Claude Sonnet) with a correction brief and narrow instructions: correct only mis-transcribed technical terms, acronyms, and proper nouns that the brief supports; do not rewrite, paraphrase, or improve the speech; do not alter timestamps.

The brief (about 1,100 characters) was assembled from two collection sources:

1. The guest’s authority record from our catalog, which lists his projects, publications, employers, and collaborators. This is what allows “Shostal” to be corrected to Schoffstall (a colleague at the guest’s employer, PSI) and “McLaughery” to McCloghrie (a co-author of the SNMP standards).
2. Vocabulary mined from contemporaneous holdings. We extracted the most frequent capitalized terms and acronyms from the subject lines of our 1992–1994 mailing-list corpora (tcp-ip, com-priv). This produces the working vocabulary of the period as its participants actually wrote it, with no manual glossary-building.

The corrected output is saved as a second file alongside the raw one. Both are kept permanently: the raw transcript is the preservation copy; the corrected transcript is the access copy; a line-by-line comparison of the two shows every change made.

Results

The correction stage changed 41 of 741 segments. Examples, with timestamps into the recording:

We reviewed the full diff. All 41 changes were within the instructed scope (terms, acronyms, proper nouns); none rewrote the wording of the speech.

Processing cost for the episode: about 9 minutes of GPU time for transcription and 8 language-model calls for correction. Extrapolated to the full 567-episode collection: roughly 10 GPU-hours and a few thousand model calls, which is within our normal batch budgets.

Limitations

• The correction stage can only fix what the brief covers. A name absent from the catalog and the mailing-list corpora will stay as the model heard it. We do not ask the model to guess beyond the brief.
• Errors outside the vocabulary class (mis-heard ordinary words, speaker attribution, overlapping speech) are untouched by this workflow.
• This is a one-episode pilot. Error rates will vary with audio quality and speaker; we will re-check a sample as the batch runs.
• The corrected transcript remains a machine product and is labeled as such in our provenance fields, at the same “receipt” tier we use for other machine-generated finding aids. It is not a human-verified transcript.

Adapting this elsewhere

The workflow assumes nothing specific to internet history. It requires: (a) an ASR system that accepts a vocabulary prompt; (b) a language model that will follow narrow correction instructions; and (c) collection materials contemporaneous with the recordings — correspondence, minutes, newsletters, an authority file — from which to build the brief. For archives that hold such materials, the marginal effort per recording is small: the brief for this episode was assembled automatically from existing records.

Contact: the IHI Reading Room. Tooling: Whisper large-v3 with initial_prompt; Claude Sonnet for correction; brief assembly scripted against our catalog API and message index.