1990, 1994, 1998
I will buckle down to work as soon as I finish reading the Internet. — Stewart Brand
Two billion people will use the Internet today1. The Internet matters because so many people use it, and so many people use it because it matters. The first side of that virtuous circle makes the bigger difference. When a new person joins a network, the value of the network increases not by their individual contribution in isolation, but by the value of all the new connections they form with others already on the network2. Just as today’s page connectors are built on top of Von Neumann computers resembling the first bespoke contraptions, so today’s Internet as experienced by ordinary people is built upon code runner networking machinery closely resembling the whole of the Internet back in 1990. That Internet had begun by connecting machines at universities to each other so that computer engineers could learn about long distance networking. Like Online System, the early Internet was “about itself”. By 1990, about 10 million people were using page machines, but only about 1 million were using the Internet3. Online System had brought page machinery and online presence into a wonderful whole for the few dozen inventors who ever got to use it. Two decades later, millions of people had one or the other, but not both working together.
The Internet’s skeleton and nerves are its address system. Like mathematics and weather forecasts, address systems consist of narrowing layers4. To send a postcard to the head of the United Nations, you might address it as follows:
Secretary Ban Ki-moon
405 E 42nd Street
Reading from bottom to top, you see layers that steadily narrow the focus from the whole world to one person at one desk, in only five lines. The first line, “USA”, tells you which of the world’s 200 countries he’s in. The second line, “NY”, tells you which of 50 states he’s in. With only five letters we have made the problem (200 x 50 =) 10,000 times smaller. The third line specifies which city, bring us down to about 500 square kilometers on earth’s surface. The fourth line, the street address, shrinks the area down to a single building5. And finally from the thousands of people working in that big building we pick just one to get our postcard.
Before cellular networking came along, phone numbers worked the same way. Here is the number of a telephone at the United Nations:
From top to bottom this time, the “1” specifies the USA. “212” does the work that “New York City, NY” did, identifying one city in one state. “963” specifies a neighborhood within the city6, and “4475” goes to a specific telephone line. With just four layers we have addressed exactly one of the world’s billions of telephones. The specific, ephemeral, and useful (“Call Secretary Ki-moon right now!”) rest on general, durable foundations such as country and area codes.
The Internet has several address systems. The first once worked very much like phone numbers before mobile: it was a set of numbers specifying physical locations at which machines could be reached7. It connected places at great expense rather than people at modest expense. It served institutions that relied on code runners. The institutions hired specialists to manage their Internet-connected code runners for them. One such institution was CERN, a cross-governmental organization for nuclear research. A CERN engineer, Tim Berners-Lee, found himself grappling with a problem very much like Vannevar Bush’s: how do you coordinate scientists who are continually burying their old work under their new work?
CERN is a wonderful organisation [of] several thousand people, many of them very creative, all working toward common goals. … [but] information is constantly being lost. … Often, the information has been recorded, it just cannot be found. … CERN meets now some problems which the rest of the world will have to face soon.8
Berners-Lee tackled the problem of organizing new knowledge from a slightly different angle than Bush had. Bush was a director of research who imagined a new tool for individuals. Berners-Lee was an individual employee and a writer of code who imagined an information machine that could bring order to an organization’s output9. For years Berners-Lee had written his own information machines to track the links between documents, concepts, and people: his own versions of Memex’s associative indexing10. At CERN he built an information machine to manage the links. Like Macintosh, Berners-Lee’s work centered on the notion of a page, but where Macintosh provided reminding pictures for the creation of printed pages and Memex before it snapshots of printed books, Berners-Lee’s pages resembled what Online System had produced: writing that held links to other writing. The relationship to the physical world dropped out. He named his system for the links, not the pages, calling it Web. Just as Bush had called “the tying of two items together” the essence of the need his tool served, Berners-Lee said:
[Because] the actual observed working structure of the organisation is a multiply connected “web” whose interconnections evolve with time … we should work toward a universal linked information system.11
Just as Hopper had created codes to wrap machine language in something easier to use, Berners-Lee created Hypertext Markup Language12 to help people publish and link pages. Information machines running at lower layers translated the human commands “publish” and “link” into the specific steps that moved pages between distant code runners. The Web brought the Internet’s global address space for institutional machines and computer operators to individual pieces of work and general researchers. It brought Bush’s focus on associative indexing, on the connections between pages rather than the pages themselves, beyond tool makers to the larger audience of tool users.
PARC had taken Engelbart’s brilliant, exclusive vision for making digital pages and made it practical for a larger population of professionals. Berners-Lee did the same for the linking of those pages across world-spanning networks. Like PARC, Berners-Lee took an important aspect of Engelbart’s expert-only vision and made it practical for a much larger population of professionals. Like PARC, he did not quite reach a mass audience. The Web as Berners-Lee first built it required required training in code runner skills. Macintosh had previously shown how to make page machines which millions of people could use without computer-specific training. Macintosh and the early Web embodied two strands from which Engelbart had woven Online System: crafting and storing pages in the first case, linking them on a network in the second. The two tools had separated out distinct parts of their common heritage, and independently determined how to make what remained usable by many more people than Online System. But the separation that made improvement possible fractured the power of the original design. Pages crafted on a Macintosh could reach readers only in the form of paper copies carried to them one at a time. Pages linked on the Web could by read at any time by anyone in the world — anyone, that is, belonging to the tiny minority with access to obscure code runner skills. Neither tool fulfilled Bush’s vision, by then half a century old, of unifying the literate world.
Building on Macintosh and the early Web, graduate students made Memex real in six short years. First among them was Marc Andreessen, who led a small team of computer scientists and investors in the creation Netscape Navigator, the first popular web browser for page machines13. Navigator brought new people to the Web even more rapidly than Macintosh had brought new people to page machines14. By 1996 millions of literate people were using the Web. Memex had shown that an information machine could enormously improve how we worked with words and audiovisual information, but it was merely an idea. Online System had shown how to bring readers and publishers together between cities, but only for its dozen-odd builders. At the same time, Online System showed how computers could power Bush’s vision of pages and connection. Macintosh and the Internet had each brought half of Online Systems’ potential to a broad audience, at the cost of discarding the other half. The Web and Navigator brought the two halves together, and by making them available to millions of people increased their value much more than a million times. Personal computers do more than run web browsers, but web browsing has become their most important use. That is why we should call them
Berners-Lee built an information machine to solve one company’s problems on top of an address system that spanned the world. Andreessen promptly brought the world to it. Within a few years, a tool that worked very well at the scale of thousands of people was managing the reading and publishing of millions. By 1998 the Web looked like CERN in 1990 or the desks of Bush’s employees in 1945: there was so much wonderful work arriving so fast, new thoughts were forgotten before they were read, because no one could keep track of them. Berners-Lee had managed to solve a medium sized problem by making it enormous.
Like Bush and Berners-Lee, Larry Page thought about the problem of organizing scientific research. Where Bush had been a prominent leader, Berners-Lee was a regular staff member. Page ranked even lower: like Andreessen, he was a student. But he tackled a larger problem: not organizing the work of one scientist (as Bush had) or of one company (as Berners-Lee had) but organizing the whole Web even as it grew more enormous by the day. Page borrowed an approach for ranking scientific papers called citation analysis15. A new scientific paper will often point at, or cite, previous papers. Those cited papers are probably more important than similar papers which do not earn pointers. In turn, a pointer from an important paper (defined as one with a lot of pointers to it!) is a better signal of importance than a pointer from an ignored paper. Notice that a citation does exactly the same work as an associative index or a web link: from within one piece of knowledge it points to another. Page weighed all of the citations pointing to a paper, and all of the citations to the citations, and so on, until all the references in a body of papers together were given a numeric score that ranked them relative to each other. You can turn that idea into a mathematical formula, which means you can turn it into a logic recipe. Page crafted such a recipe, but for Web pages instead of research papers. He named his invention PageRank, after the Web and himself.
Memex and HTML are tools operated by individuals to make connections at the speed of one mind. Because it was a logic recipe, Page’s tool could run on a code runner — in fact, on as many code runners as he could get. He used dozens, then hundreds, then thousands of code runners to operate PageRank, building a company around them to keep them working. He called the result Google, meaning an enormous number. Google users discovered that PageRank achieved its goal: for the first time since the Web had grown to world scale you could find what you wanted on it. Researchers can publish their work on the Internet because Berners-Lee invented the Web. A broad audience will read it because Andreessen made Navigator. The audience can find what they are interested in because Larry Page invented PageRank and built Google. Bush focused on the problem of recording one mind’s attention as it moved associatively from page to page. Berners-Lee focused on the problem of recording an institution’s associations as they moved along with the changing work and staff. Page used math to make an information machine that could track the associative attention of the whole world. Google is so central to working with information today that it has become a verb: to google a topic means to search for what is known about it16.
PageRank made the Web work by treating links as citations. Echoing Bush 70 years later, Kathleen Fitzpatrick argues that the connections between works define scholarly communication:
Citations are the highway markers of an ongoing conversation, one that does not end with the text presently being written, but that has the potential to stretch both forward in time and outward in unexpected directions. Any given scholarly exchange could result not just in the rebuttal of prior arguments but in those arguments’ potential recirculation and reinterpretation in the context of another scholar’s work.17
The Web is woven for knowledge work.
Bush’s problem has not stopped expanding. The most sophisticated and efficient users of Google today find that “trying to navigate the growing deluge of data [just within their scientific field] has become a second job.”18 The United States’ library of scientific papers for health and medicine adds new ones at a rate of two a minute, year round19. Like Memex and the Web, Google has solved the problem of published work burying itself unread by pushing it out to a vastly wider layer, one we will perhaps never manage to tame. We have not escaped Bush’s quandary:
Those who conscientiously attempt to keep abreast of current thought [find that] truly significant attainments become lost in the mass of the inconsequential. … publication has been extended far beyond our present ability to make real use of the record.20
Google culminated the augmentation of human knowledge begun by Bush, making the Internet a tool for knowledge work. Today the scholarly Internet has been subsumed in systems that connect almost everyone.
2. This assertion, often attributed to networking pioneer Robert Metcalfe, is called “Metcalfe’s Law.” Though helpful, it is no more a law than “Moore’s Law” or “Godwin’s Law.” While we are fixing names, perhaps we might agree to switch from “Law” to “Epigram”, promoting to “Insight” the ones that hold up particularly well. Of course, downgrades to “Bloviation” will also prove warranted at times.
4. This characteristic of information pops up regularly in information machines. Transitions in computing, whether historical (Online System to Macintosh) or social (government weather forecast released at 5am becomes a rain alert on my iPhone at 8am) often involve paired opposing shifts in scale. Online System was powerful, but only a few hundred people used it. Macintosh was much more limited a “knowledge augmentor”, but millions owned one. Going the other way, the weather forecast for a populous region matters to millions of people, but the weather app on my phone shifts with my location. Once you start looking for this structural dynamic, you’ll see it everywhere. For example, a tweet by Benedict Evans glosses the history of computers from Hopper to today as such a paired shift: “Punchcards: data is in your hand, but content is massively abstracted. Smartphone+cloud: data is in another country but you touch the content”.
6. The useful geographical specificity of prefixes is being abandoned as mobile telephones replace fixed-line phones. Consider the parallel with previous footnote.
7. Zack Bloom, “The History of the URL: Domain, Protocol, and Port”.
9. Bush’s organization in 1945 and CERN in 1990 were about the same size, a few thousand people. Separated by a half century and very different technical building blocks, the two inventors settled on the same behavior — the manual creation of links — as the key element in their systems. In Weaving the Web, Berners-Lee cites the many predecessors who built working hypertext systems that did not become popular. We might ask ourselves to what extent the false starts stalled because they didn’t happen to set their first fragile roots in a pot of just the right size.
10. Tim Berners-Lee, Weaving the Web, Chapter 1.
11. “Information Management: a Proposal”
12. Unlike machine languages or Hopper’s codes, Berners-Lee’s HTML cannot express universal logic.
13. Andreessen’s reworking of the code runner Web to a page machine Web was a creative and enormously consequential effort. Zack Bloom makes the strong case for Andreessen as an inventor in “To What Extent Did Marc Andreessen Invent The Internet?”
The idealists got things started. People like Douglas Englebart (who envisioned Hyperlinks) and Tim Berners-Lee (who allowed them to exist on the Internet). [But a]fter crafting the foundation, the idealists … invested their time creating tools which almost no one uses … It was the pragmatists, like Marc Andreessen, who saw the web … as a way people could build powerful things and solve real problems today. In the process, he created many things which don’t really align with the [idealists’ formal] vision … ma[king] it possible to do more on the web in a way that actually got adopted. … The web never would have existed without the idealists, but you wouldn’t be reading this on the web without the pragmatists.
Bloom’s point is well taken, but he undersells Berners-Lee as a pragmatist and technical politician. See Weaving the Web.
14. About 50 million people after 5 years per “Internet User Forecast by Country”, versus perhaps 1/10th that many Macintoshes sold in its first 5 years.
15. Berners-Lee pointed in this direction in his original design document: “Information Management: a Proposal”: “An intriguing possibility, given a large hypertext database with typed links, is that it allows some degree of automatic analysis. … It is also possible to look at the topology of an organisation or a project, and draw conclusions about how it should be managed, and how it could evolve.”
16. Recall that “google” meant a number, so we are in effect making a number into an instruction, faintly echoing Turing’s insight.
17. Fitzpatrick, Kathleen. “The Future of Academic Style: Why Citations Still Matter in the Age of Google” Los Angeles Review of Books. 29 March 2016. lareviewofbooks.org/article/the-future-of-academic-style-why-citations-still-matter-in-the-age-of-google (I do not dare bypass MLA style when quoting the eminent Ms. Fitzpatrick.) Consider how much denser the footnotes are in this section than in the previous. The Web exemplifies itself.
18. Esther Landuis, “Scientific literature: Information overload”, Nature 535, 457-458 (2016) doi:10.1038/nj7612-457a. For an individual researcher, the phenomenon has the nickname “tab bankruptcy”. Star Simpson offers the Japanese word “Tsundoku” when the reading material is books. The predicament is arguably older than the English language.
19. Op. cit.
20. “As We May Think”