Category: Ancient Web

The most ancient times of the internet: 90s, early 2000s.

Posted on by

Ancient Web: Zusie Built an 8-Bit Computer From 1,500 Salvaged Telephone Relays

Fredrik Andersson wanted to hear a computer think, so he built one from enough relays to make the thinking audible.

Visit Zusie, the relay computer

Zusie is an 8-bit homebuilt computer inspired by the electromechanical machines of Konrad Zuse. Its data path is dominated by relays—the clicking switches used in early computers and telephone exchanges before vacuum tubes and transistors took over.

The specifications are serious: an 8-bit data bus, 16-bit address bus, three 8-bit accumulators, two 12-bit registers, a 12-bit program counter, a 12-bit stack pointer and an ALU capable of logical operations, addition, shifts, increment and decrement.

Memory and microprogram storage use solid-state components, but the machine’s visible logic remains gloriously electromechanical.

The project began with a museum visit and a junk box

Andersson traces the idea to a 2008 visit to Berlin’s technical museum, where he encountered Konrad Zuse’s early machines and became fascinated by the relay-based Z3.

Then the supply problem appeared.

A relay computer needs hundreds of relays, and new parts would make the project absurdly expensive. In 2009 Andersson found old telephone-exchange circuit boards being sold cheaply by a Swedish surplus dealer. He eventually made a 500-kilometer trip each way and bought roughly 100 boards containing relays.

After desoldering and sorting them, he ended up with about 1,500 usable switches.

The extraction process produced another piece of documentation that only a personal project page would preserve: heating old circuit boards with a hot-air gun inside an apartment kitchen worked extremely well and smelled, in his words, approximately like death mixed with poison.

Engineering lessons come in layers.

The finished machine uses a microcoded architecture. Andersson wrote a microcode assembler, a regular assembler and communications software for an old DOS laptop connected through a parallel port. The project page includes architecture notes, a build log, downloadable software and even an online simulator.

That depth matters because a homebrew computer is much more interesting when the design survives alongside the photographs.

The site also links Zusie into a wider Web ring of homebuilt CPUs and relay-computer projects. This was not one isolated eccentric machine; it belonged to a loose community of people rebuilding computation from first principles with whatever technology amused them most.

Modern processors switch billions of transistors silently at impossible speed.

Zusie is slower, larger and less efficient by every practical measure.

It also clicks.

Sometimes that is the feature.

Explore Fredrik Andersson’s Zusie relay computer

Posted on by

Ancient Web: Mother Bedford Used One Pair of Jutti to Explain Centuries of Punjabi Leatherwork

One of the old Web’s better habits is letting an ordinary object become the entire subject for a page.

Visit Mother Bedford’s page on jutti

The June 13, 2014 post begins with a pair of shoes whose toes curl sharply upward. From there it identifies the style as jutti, connects it to the Punjab region of India and Pakistan, and starts taking the object apart with words.

Jutti are generally flat-soled, closed around the heel and richly decorated. The related mojari is described as more sandal-like with an open back. The page notes the use of colored beads, small mirror pieces and shells in decorative uppers.

For men’s examples, the extended toe may curl back over the vamp. The article records a tradition connecting that shape to the curled tips of Punjabi men’s moustaches.

That is the sort of detail a generic “traditional footwear” page usually loses.

The object has a provenance

The pair shown on Mother Bedford is not merely an illustration pulled from a catalog.

The author says the shoes may date to the nineteenth century and had once belonged to a collection displayed at the Crawford Museum in Breezewood, Pennsylvania. When that museum closed and parts of the collection were sold, the pair entered the author’s possession.

The photographs then become evidence.

The soles and heels appear to be made from layers of leather sewn together with leather thread. The uppers combine leather and fabric. Wear, construction and decoration can be inspected directly rather than summarized into one museum-label sentence.

That makes the page a good example of small-scale material-culture documentation.

Collectors and hobby historians often own objects that institutional catalogs never describe in detail. A personal website gives them room to photograph the seams, explain where the item came from, compare terminology and admit uncertainty about dating.

The result is not a formal peer-reviewed catalog entry.

It does not need to be.

For researchers, the useful pieces are still there: object name, regional context, construction method, related terminology, approximate date, former collection and multiple photographs.

This is one reason CacheRat’s Ancient Web project keeps wandering into pages that look almost absurdly narrow. Search value often lives exactly there. Somebody researching jutti construction, Punjabi footwear, the Crawford Museum collection or layered leather soles may care far more about this one page than about a broad history of shoes.

The Web does not have to explain everything.

Sometimes preserving one strange old pair of shoes is enough.

Read Mother Bedford’s jutti article

Posted on by

Ancient Web: Don Knuth’s Homepage Still Treats the Web Like a Personal Research Desk

Donald Knuth’s homepage does not look like a personal brand. It looks like a desk where somebody has been leaving useful piles for decades.

Visit Donald Knuth’s Stanford homepage

Knuth, Stanford professor emeritus and author of The Art of Computer Programming, uses the page as an index into his actual work rather than as a résumé with a headshot and a newsletter signup.

The links jump to frequently asked questions, infrequently asked questions, computer musings, preprints, downloadable programs and data, graphics, his pipe organ, Fantasia Apocalyptica, curriculum vitae and lists of known errors in his books.

There is even a link asking whether you borrowed a video from him.

This is personal-web architecture at a very high intellectual level: the navigation follows the person rather than a content-management taxonomy.

Errata as a public system

Knuth’s book pages are especially revealing.

He publishes lists of corrections and amendments in formats including plain TeX, DVI and PostScript. Readers who are first to report a genuine new error can receive a reward of at least hexadecimal 0x$1.00—$2.56.

That joke is famous among programmers, but the useful part is the infrastructure behind it.

A book does not become frozen when it reaches print. Errors remain discoverable, readers can report them, and corrections become part of an openly maintained record. The website acts as the living appendix to physical books.

The list spans far more than The Art of Computer Programming: The TeXbook, The METAFONTbook, Concrete Mathematics, MMIXware, Literate Programming, the Stanford GraphBase and numerous collections of papers all connect back into their own correction trails and supporting material.

That is what durable scholarly web publishing looks like.

The homepage also refuses to separate technical life cleanly from everything else. Music sits beside algorithms. Religious and literary work sits beside source code. Political questions appear under the deliberately odd heading “Infrequently Asked Questions.”

A university marketing department would probably reorganize this.

That would make it worse.

The old personal homepage model allowed an expert to decide that a downloadable program, a book erratum, a pipe organ and an unfinished question all belonged one click away from each other because they belonged to the same human being.

Search engines benefit from that specificity too. Exact book titles, TeX terminology, algorithm names and program files remain directly reachable instead of being buried under a generic faculty template.

Knuth helped teach generations of programmers to care about structure.

His homepage quietly demonstrates another kind of structure: enough links to be useful, enough personality to remember whose desk you are standing beside.

Explore Donald Knuth’s homepage at Stanford

Posted on by

Ancient Web: The Enron Archive Rebuilt the Company’s Website Before the Collapse

Most surviving Enron material begins with the scandal. The restored Enron corporate website begins before everybody knew how the story ended.

Visit the restored Enron corporate site

EnronCorp.com is a historical reconstruction maintained by the Enron Archive. The project describes the site as a mostly working restoration of Enron’s different websites as they existed before Jeff Skilling resigned on August 13, 2001.

That cutoff is deliberate.

The archive is not trying to retell the bankruptcy, criminal trials or later investigations. It is preserving the company’s own Internet presence before collapse rewrote the meaning of everything on it.

The reconstruction draws on archived web material and files from the Enron email corpus. Its sections include About Us, Products + Services, Press Room, Investor Relations and Work at Enron. Some interactive pieces remain broken because the original systems behind them no longer exist.

That incompleteness is part of the evidence too.

Corporate history before the footnotes arrived

Enron was once one of the largest energy companies in the United States and a major trader of natural gas and electricity. Fortune repeatedly called it America’s “Most Innovative Company.”

A restored corporate site shows how that identity was presented in real time.

The page title still carries the slogan-like language of the era: “discover the power of WHY.” Investor material, product descriptions, career pages and press releases were written to explain a growing, sophisticated company—not a future case study in accounting fraud.

That makes the site useful in a way that a later encyclopedia entry cannot reproduce.

Retrospective history knows which claims deserve suspicion. Contemporary corporate copy does not. Researchers can compare what executives told investors, job applicants, customers and the public with what later investigations revealed.

The contrast is more powerful when the original presentation is preserved rather than quoted selectively.

This also demonstrates a difficult side of web preservation. A corporate website is not just HTML. Search forms, job databases, investor tools, image assets and server-side applications can disappear even when individual pages survive. The Enron Archive openly notes which portions could not be reconstructed.

That is better than pretending an incomplete mirror is a perfect time machine.

For CacheRat’s Ancient Web project, reconstructed sites deserve a place when the restoration itself is transparent. The goal is not nostalgia. It is access to the public-facing record of organizations that later became historically important.

After Enron collapsed, its name became a synonym.

This restoration remembers the awkward period when it was still a brand.

Explore the restored Enron corporate website

Posted on by

Ancient Web: PSDoom Turned Unix Processes Into Doom Monsters

In 1999, Dennis Chao looked at Unix administrators talking about “killing” processes and decided the obvious interface was a shotgun.

Visit PSDoom at the University of New Mexico

His experiment, later known as PSDoom, modified the Linux port of Doom so running processes appeared as monsters. Wound a monster and its corresponding process could be reniced. Kill the monster and the process died.

This was not merely a joke screenshot.

Chao used the project to think seriously about whether abstract computer operations could be mapped into an environment people understand intuitively. The desktop metaphor had already turned files into documents and directories into folders. Why not turn a machine’s process table into a building full of demons?

Unix had done half the naming work already.

kill -9, but with ammunition

The proposal has a surprising amount of internal logic.

A crowded room can represent a heavily loaded machine. Important processes can be represented by tougher monsters. A novice administrator can be issued a weak weapon, while an experienced one gets something stronger. Multiple administrators could cooperate in multiplayer mode.

Chao even points out that drastic actions should require more effort than harmless ones. On a command line, ls and a catastrophically destructive command can be typed with roughly the same physical effort. In a game world, large actions can demand movement, time and risk.

Then reality demonstrates why metaphor-driven interfaces are dangerous.

One screenshot shows Chao being attacked by csh, process ID 18729. He explains that after taking the image, csh was hit by friendly fire—possibly from tcsh or xv—and his session abruptly ended.

That is excellent research documentation because the failure is the point.

The page dates itself October 17, 1999 and links to code plus a CHI 2001 paper and talk. Chao credits the University of New Mexico’s Adaptive Computation Group and thanks the group for providing an environment where playing Doom for two days could plausibly be called research.

PSDoom survives because it is more than nostalgia for a famous game. It captures a moment when Linux, released game source code and academic human-computer-interaction research could collide on a university homepage.

Today system dashboards offer colored graphs, sortable process tables and polished controls.

PSDoom offered a BFG.

There are arguments for both.

Read Dennis Chao’s original PSDoom page

Posted on by

Ancient Web: Paul Rendell Built a Turing Machine Inside Conway’s Game of Life

Conway’s Game of Life has no CPU, no variables and no instruction set. Paul Rendell built a computer in it anyway.

Visit Paul Rendell’s Game of Life work

Rendell’s site documents a Turing machine constructed entirely from patterns inside John Conway’s cellular automaton. The original machine is dated April 2, 2000 and uses the collisions and timing of Game of Life objects to implement the pieces a computer needs: memory, addressing, control logic and a tape.

The first working design is deliberately small. It has three states and three symbols. One machine cycle takes 11,040 generations of Life.

That is not fast.

Speed is not the point.

Logic made from moving dots

The detailed pages break the machine into components with names that sound half like computer engineering and half like astronomy: finite state machine, signal detector, stack cell, comparator, fanout, glider gun, lightweight spaceship gun and output collator.

The tape is implemented as two stacks. Memory cells store patterns of gliders. Rows and columns address the finite-state-machine matrix. Carefully timed collisions create, destroy or redirect the signals that carry information.

The machine is computation reduced to geometry and timing.

Rendell also made the design expandable. The early version could be enlarged to as many as 16 states and eight symbols, enough capacity for a universal machine. In 2010 he documented a Universal Turing Machine in Life, and in 2011 a fully universal version whose stack could continue growing during operation.

One image on the later site shows the fully universal machine after 149 million generations.

That sentence feels excessive until you remember the computer itself is made out of a mathematical toy whose only rule is whether cells live or die based on their neighbors.

The pages preserve downloadable patterns, diagrams, animations, Java tools and explanations of the individual logic structures. That makes the archive more useful than a video of a giant Life pattern doing something mysterious. A reader can trace the design down into gates and memory cells.

This is exactly the kind of technical Web project that deserves to remain linkable. It sits at the intersection of recreational mathematics, computer architecture and pure stubbornness.

A modern processor hides billions of transistors under packaging.

Rendell’s machine does the opposite. It turns computation into an enormous visible contraption and invites you to inspect every moving part.

Explore Paul Rendell’s Game of Life computer designs

Posted on by

Ancient Web: The First International Collection Packed 3,660 Tongue Twisters Into 118 Languages

The First International Collection of Tongue Twisters is what happens when a simple party trick meets the organizational instincts of the old Web.

Visit the First International Collection of Tongue Twisters

The site reports 3,660 entries in 118 languages. English gets hundreds. German, French, Dutch, Czech and Spanish have large sections. Smaller collections range through Acholi, Avar, Guarani, Kikuyu, Luo, Maori, Sardinian, Wolof, Xhosa and dozens more.

The compilation is credited to “Mr. Twister” and dates from 1996 through 2018.

That twenty-two-year span matters because the collection did not emerge from a modern language database. It grew like a personal Web project: submissions, credits, rough translations, transliterations, language pages and incremental cleanup.

The homepage eventually moved to Unicode, which is itself a small history lesson.

A multilingual site from before Unicode was boring

Displaying 118 languages on the Web used to be an actual technical problem.

The current index announces that all pages use UTF-8 fonts, but older material still reveals the transitional era: some entries include native scripts beside romanized versions, some preserve alternative language names, and the site links to help for users having trouble with Unicode fonts.

The collection is more than a giant list of difficult sentences.

It distinguishes tongue twisters from related forms such as shibboleths and repetitive phrases. Language pages include contributor credits and, where available, rough translations or pronunciation help. Several languages even have audio examples.

The counts are wonderfully uneven because human culture is uneven. English has 593 entries. German has 364. French has 200. Some languages have one.

That is more honest than pretending every language arrived through the same neat data pipeline.

For linguists and language learners, tongue twisters are useful little stress tests. They expose sound combinations, consonant clusters, rhythm, alliteration and distinctions that native speakers take for granted. A phrase designed to make a fluent speaker stumble can be brutal evidence for somebody learning the language.

For digital archaeologists, the collection also preserves an older model of global participation. People emailed material to a webmaster, the webmaster organized it, and over years the page became larger than any one contributor could have planned.

No feed. No engagement loop. No AI pronunciation coach.

Just 3,660 ways to get your own mouth wrong.

That is enough infrastructure for one website.

Explore the full multilingual tongue-twister index

Posted on by

Ancient Web: Rome Art Lover Compares Vasi’s 1747 Porta Pinciana With Modern Rome

Roberto Piperno has found a good use for hypertext: stand where an eighteenth-century engraver stood and keep clicking backward through Rome.

Visit Rome Art Lover’s Porta Pinciana page

The page centers on plate number 2 from Giuseppe Vasi’s Grand View of Rome, published in 1747. Vasi showed Porta Pinciana when it was a relatively minor opening in the Aurelian Walls, used largely for access to suburban villas.

Piperno then puts the old engraving beside the city that followed it.

There is a modern photograph, a small 1748 map, notes on neighboring gates, links into Rome’s historic rioni, architectural details, inscriptions, tombs and sections of the ancient wall. Instead of treating Vasi’s print as a decorative antique, the page uses it as a coordinate system.

One gate, about seventeen centuries of modifications

Porta Pinciana began as a small opening in the walls built under Emperor Aurelian in the third century. The page notes that it was enlarged under Honorius in 402 and strengthened by Belisarius after the Byzantine reconquest of Rome in 536.

By Vasi’s time the gate had a much quieter role.

It was closed in 1808 after the Borghese expanded their estate and created another entrance to Villa Borghese. It reopened in 1887 as the surrounding city changed again.

That pattern continues into the twentieth century. Piperno connects the walls to road construction around the 1960 Rome Olympics, archaeological discoveries made during infrastructure work, and even the conversion of some wall towers into houses or studios.

The page is also full of the small details that make specialist websites useful. A gate associated with Villa Medici carries inscriptions inviting guests into the gardens. A Roman sarcophagus found nearby is discussed through its decoration. A reconstructed tomb uncovered during sewer or road work becomes another layer in the same landscape.

And because this is a personal site rather than a museum kiosk, a red Fiat 500 parked by the wall can earn a sentence too.

That is not a flaw.

It is how walking through a city actually works. Monument, inscription, traffic project, ancient tomb, little red car.

Rome Art Lover turns Vasi’s engravings into a sprawling visual index of change. The old plate is never merely “then,” and the photograph is never merely “now.” Links lead sideways into villas, roads, districts, monuments and other Vasi plates until the city becomes a network rather than a timeline.

The modern Web loves before-and-after sliders.

Piperno built the scholarly version: before, after, map, footnote, detour, another century, another click.

Explore Porta Pinciana at Rome Art Lover

Posted on by

Ancient Web: Web Japan Caught Pokémon Before It Became a Global Empire

It is difficult to remember Pokémon as a new thing. Web Japan still has a page from when that was exactly what it was.

Visit the Kids Web Japan Pokémon archive

Filed in the October-December 1997 section of Kids Web Japan’s What’s Cool in Japan archive, the article explains Pocket Monsters to an overseas audience before the franchise had become permanent global furniture.

The numbers already looked dangerous. The page says the Game Boy title went on sale in February 1996 and sold more than six million copies in its first eighteen months. It describes 151 kinds of monsters, the map-based hunt through towns, forests and oceans, and the ability to capture, raise, trade and battle them.

Most importantly, it notices the link cable.

Pokémon was not merely another cartridge. Kids could connect Game Boys, trade monsters and compete with friends. That social mechanic is now obvious enough to disappear into franchise history, but the contemporary page treats it as one of the reasons children were going wild for the game.

A franchise caught while it was still becoming one

The archive records the speed of the expansion.

Strategy books appeared alongside the games. Manga followed. The television series began in April 1997. Toys, stuffed animals, stationery and food packaging were already carrying the characters.

In August 1997, East Japan Railway held a Pokémon stamp rally across 30 major stations. Kids traveled from checkpoint to checkpoint filling booklets with character stamps. Web Japan reports that more than 100,000 fans participated.

Then the page captures a much darker piece of contemporary history. It notes that television broadcasts were temporarily suspended in December 1997 after several hundred children experienced convulsions and dizziness following an episode with intense flashing effects.

That incident is now part of Pokémon lore. Here it appears not as retrospective trivia but as current news sitting beside a radio show and plans for a movie.

That is why surviving contemporary web pages matter.

A modern history of Pokémon knows what happened next: Pikachu became internationally recognizable, the franchise expanded across generations of Nintendo hardware, and the original 151 became only the first layer of an enormous catalog.

The old page does not know any of that yet.

It is standing inside the explosion with a Game Boy, a cable and a train-station stamp booklet.

For CacheRat’s Ancient Web project, contemporary pages like this are unusually valuable. Later histories explain significance. Old pages preserve uncertainty, terminology and scale before everybody knew how the story ended.

Pokémon eventually became history.

This page caught it while it was still news.

Read the surviving Kids Web Japan Pokémon article

Posted on by

Ancient Web: The Encyclopedia of Electronic Music Kept Growing from 2002 into the 2020s

The Encyclopedia of Electronic Music is the kind of website that answers the question, “What if one person simply kept adding obscure synthesizer records for twenty years?”

Visit the Encyclopedia of Electronic Music

Its introduction is dated July 5, 2002, with revisions noted in 2007, 2014 and 2016. The stated purpose is straightforward: help people get acquainted with serious non-academic electronic music through a database that keeps growing.

The site’s definition of the territory is revealing. It focuses on what it calls electronic prog: Classic Electronic, Cosmic, Berlin School, Rhythmic, Melodic, Ambient, Space and Experimental music.

That taxonomy immediately dates the archive to a Web culture where genre maps were often made by enthusiasts rather than recommendation algorithms.

The alphabet is the interface

Artist pages are arranged in giant alphabetical lists.

An entry might be three lines long: artist, country, releases and a compact description. Another may branch into related artists, side projects, reviews or subgenre links. The pages mix familiar names with projects obscure enough that the encyclopedia itself may be one of the easiest surviving ways to discover them.

The underlying HTML has barely concealed its age. Some pages still expose Microsoft FrontPage metadata. Background images, old font tags, named anchors and long vertical documents do most of the structural work.

And yet the content has not frozen in 2002.

Current alphabet pages contain releases from the 2020s, including entries dated 2023, 2024 and 2025. The old interface is still being used to describe new music.

That is much more interesting than a retro redesign.

The site also preserves a reviewer-driven vocabulary that streaming platforms tend to flatten. One project might be described as dark space ambient, another as melodic sequencer music, another as synth-heavy experimental work with guitar, noise or tribal elements.

These are subjective judgments, but they are useful subjective judgments because they come from a consistent human listener trying to tell another person what something sounds like.

The review pages extend that approach into longer criticism. They retain release context, stylistic comparisons and personal reactions instead of reducing everything to a score or popularity metric.

For researchers of independent music culture, sites like this preserve more than discographies. They preserve the language fans used to divide scenes, the artists they considered related and the records that were visible enough to enter one person’s map of the genre.

The Web is good at replacing old interfaces.

The Encyclopedia of Electronic Music demonstrates the opposite possibility: keep the interface, keep adding knowledge.

Explore the Encyclopedia of Electronic Music