« Back to Thesis Home

---

Chapter 7 — The Future of the Web
(Alternative Edition)

---

7.1   Introduction

In Chapter 2 the author described the World Wide Web as being "only in its infancy" — barely two years old as a mass phenomenon, and growing at a rate of several hundred per cent each year. The aim of this chapter is to consider, with as much honesty as the author can muster, what happens when an infant of that size grows up.

It is tempting, when writing about a technology this new, to confine oneself to the optimistic. The Web is exciting; the author is enthusiastic; the literature is broadly hopeful. However, the author has come to believe over the course of this report that the most important question is not what the Web could be, but what its incentives will quietly push it to become. The two are not the same.

7.2   The Browser Wars Will End, but Not Cleanly

In Chapter 5 the author evaluated Netscape Navigator 2.01 against NCSA Mosaic 2.0 beta 1. Forrester's research suggests Netscape currently holds 74 per cent of the browser market. It is the author's view that this dominance will not last.

Microsoft has bundled Internet Explorer with Windows 95 at no additional cost. This is not a feature; it is a strategy. When a browser ships with the operating system, the cost of switching becomes higher than the cost of accepting what is already installed, and the user — who in Chapter 6 was shown to prefer the path of least resistance in almost every measurable dimension — will mostly accept the default. Netscape sells its browser; Microsoft does not need to. This asymmetry is decisive.

The author predicts that within five years Netscape will be a curiosity, and within ten its remains will have been reabsorbed into an open-source project of some kind. NCSA Mosaic will be discontinued sooner still — likely within the next eighteen months. The lesson here is not about either product specifically, but about a pattern that will recur: dominant Web applications will be displaced not by better applications, but by applications bundled with the platform beneath them.

7.3   The Web Will Become the Platform

In Section 5.13 the author noted Sun Microsystems' Java applets, and the in-line scripting language Netscape calls "JavaScript". Both are presented in the trade press as curiosities — toys for animating banner adverts and displaying stock tickers.

The author believes JavaScript in particular has been underestimated. Java will struggle, because it requires a separate runtime and the user, again, will not download things they do not have to. JavaScript is already in every Netscape browser. Whatever runs everywhere, wins. The author predicts that within fifteen years JavaScript will not be a toy for ticker tape; it will be the language in which the majority of new software is written, including software with no apparent connection to the Web at all.

This will produce a strange inversion. The Web began as a way of viewing documents. It will become a way of running programs. The browser, which in this report has been treated as an application for accessing the Internet, will itself become an operating system — the operating system most people use most of the time, even if they do not realise it.

7.4   On Cookies, and What Comes After Them

In Section 5.17 the author noted that future versions of Netscape Navigator will retain a feature called "cookies", which allows Web sites to store information about a user's visits and retrieve it on subsequent occasions. Mills (1996), writing in InfoWorld, observes mildly that this allows electronic merchants to learn about the user's habits and preferences and to tailor a page for the user.

The author wishes to register, in this chapter, a more pointed view. A cookie, as described, is a small text file stored on the user's own computer at the request of a Web site the user has visited. This is not in itself remarkable. What is remarkable is the asymmetry: the Web site knows what it wrote; the user, in almost every case, does not. The user is being asked to host, on their own machine, a record of themselves that is legible to others and illegible to themselves.

If this practice spreads — and the author can see no reason why it would not, given the commercial pressure to spread it — then within ten years it will be normal for an average user to be tracked, in a manner they would not consent to if they understood it, across dozens of separate Web sites at once. The mechanism described by Mills as "tailoring a page for the user" is the same mechanism, viewed from the other side, as building a dossier on the user. The two descriptions are not in conflict. They describe the same act.

Johnson (1995), whose worst-case scenario the author quoted earlier, has been treated in the trade press as alarmist. The author respectfully disagrees. Johnson's scenario — citizens treated as consumers to be targeted rather than as citizens to be connected, with most of what they do recorded and analysed for use against them later, in advertising, insurance, and employment decisions — is not a worst case. It is the most likely case, because it is the case that the commercial structure of the Web rewards.

7.5   The Web Will Eat the High Street

At the time of writing, online commerce is a curiosity. The author has personally purchased one item via the Web (a book, from a small American company called Amazon.com, which arrived three weeks later, slightly damaged). Most retail analysts treat Web commerce as a niche.

The author believes this is wrong. The reason high-street retail has historically been protected is that physical shops offer three things mail order cannot: immediacy, examinability, and trust. The Web is already eroding the first via overnight delivery; the second via richer media (Section 5.12 noted Netscape's plug-in support for video and audio); and the third, eventually, via reputation systems and payment guarantees that are at this point only theoretical.

Once all three are eroded, the high street has no defensible moat. The author predicts that within twenty years, every category of retail not requiring genuine physical presence — books, music, film, software, clothing, electronics, groceries — will have shifted decisively online. The shops that survive will sell experiences (coffee, food, haircuts, gymnasiums) or genuine immediacy (pharmacies, corner shops). The middle, where the author currently buys most things, will hollow out.

This is not, in the author's view, an unambiguous good. The high street is not merely a place of commerce; it is a place where people who would not otherwise meet are brought into casual contact. Its replacement with a Web of one-to-one transactions, each one optimised in isolation, will produce towns that are more efficient and lonelier.

7.6   The Mobile Question

In Section 6.4 the author noted with some surprise that eleven of the twenty-one respondents to his usability survey expressed enthusiasm for the idea of touch-screen access to the World Wide Web. At the time of writing, no such device exists in any usable form. The author treated the response as aspirational rather than predictive.

On reflection, the respondents may have been seeing something the author was not. The mobile telephone is now in roughly twelve per cent of UK households, and growing rapidly. Within ten years, the author predicts, it will be in almost every pocket. Within fifteen, the distinction between a "mobile telephone" and a "small computer with Web access" will collapse, and the latter will be how most people in the developed world access the Internet for the first time.

This has profound implications for everything in this report. The HCI principles drawn from Shneiderman and Preece, the icon-based GUI conventions inherited from the Xerox Star, the keyboard-and-mouse paradigm assumed throughout Chapter 5 — all of these assume a user seated at a desk, in front of a CRT monitor, with both hands free. When the primary access device is held in one hand, viewed at arm's length, and operated by the thumb of the hand holding it, almost every design principle in the established literature requires re-evaluation.

The author would, frankly, like to write that thesis instead of this one. It will be necessary within the decade.

7.7   The Question of Truth

In Section 1.6 the author noted, with what he now sees as understatement, that the Web is "not always an authority on facts" because "there is no one actually monitoring subjects". The author treated this as a citation-quality problem — something to be solved by careful cross-referencing with paper sources.

It is a larger problem than that. Paper publishing has historically included a number of slow checks — editors, publishers, distributors, libraries — each of whom had to be persuaded that a piece of writing was worth handling before it reached the reader. These checks have been criticised, fairly, as gatekeeping. They were also, less obviously, a quality filter. The Web removes all of them at once.

In the short term, this will produce an explosion of useful information, written by people who would never have had access to a printing press. In the long term, the author suspects, it will also produce an explosion of plausible-looking falsehood, written by people whose interests are served by its being believed. The two will not be easily distinguished by the casual reader. The Web will not become less truthful than the printed page; it will become more variable, with a higher ceiling and a much lower floor.

Section 4.6 noted Sun Microsystems' finding that "the more well-organised a page is, the more faith I will have in the info". This was reported as a usability insight. It is also, viewed less charitably, a vulnerability. A well-organised page is not necessarily a truthful page. The user who confuses the two — and the author's survey suggests most users will — is in trouble.

7.8   A Note on Artificial Intelligence

The author has avoided the topic of artificial intelligence in this report, on the grounds that the AI winter of the late 1980s has left the field in poor health, and that the available systems are not yet capable of the tasks claimed for them. This judgement is correct for the present moment. The author wishes to register, however, that he expects it to be wrong within twenty-five years.

The reason is not any specific breakthrough, but the conjunction of three trends already visible: rapidly falling computation costs, rapidly growing digitised corpora of human writing (the Web itself being the largest of these), and steady if unspectacular progress in connectionist models of the kind dismissed during the 1980s. By approximately 2020, the author predicts, it will be possible to build a system that converses in natural language at a level indistinguishable, in most practical respects, from a competent human. This system will not be intelligent in the philosophical sense. It will be useful in the commercial sense, which is the sense that determines what gets built.

The implications for HCI are considerable. The icon-and-menu paradigm evaluated throughout this report assumes that the user knows roughly what they want and needs help locating it. A conversational interface assumes the user can describe what they want and needs help refining it. The two are not the same problem. Much of the careful design work documented in Chapters 3 and 4 will, in this future, need to be re-done from first principles for an interaction model that does not yet exist.

7.9   What This Means for the Designer

The author has, throughout this report, drawn on Shneiderman (1992) and Preece et al. (1994) for principles of good interface design. The author wishes, in this closing section, to suggest two additional principles that the canonical literature does not yet contain, but which the projections above suggest will become important.

First, designers should treat their users' attention as a finite, exhaustible resource which it is morally serious to consume. The Web will become, in the author's view, the most powerful attention-capture medium ever built. The temptation to design for engagement — measured in clicks, minutes, return visits — will be enormous, and the commercial reward for doing so will be enormous. The user will not consent to this in any meaningful sense, because the costs of capture are paid in fragments and the benefits are paid in cash. The good designer, in this environment, will need to actively design against the incentives of the medium they work in.

Second, designers should treat the information they collect about users as a liability, not an asset. Each datum stored about a user is a datum that can be lost, stolen, subpoenaed, sold, or used against them by parties not yet known to the designer. The cookie described in Section 5.17 is the beginning of a long road. The designer who collects only what is necessary, and discards it as soon as possible, will be doing their users a service that the users themselves cannot easily see and will not thank them for. The designer who collects everything will be hailed, in the short term, as "data-driven", and will produce, in the long term, considerable misery.

Neither of these principles will be popular. The author submits them anyway.

7.10   Conclusion

The author began this chapter by suggesting that the most important question about the Web is not what it could be, but what its incentives will push it to become. The predictions above are the author's best attempt to follow those incentives forward in time.

They are, taken together, a more pessimistic forecast than the trade press currently offers. The author is aware of this and wishes to note that he does not enjoy the role. The Web, evaluated as a piece of technology, is the most exciting thing the author has encountered in three years of studying computing. The Web, evaluated as a piece of commerce, will be one of the most consequential developments of his lifetime. The two evaluations are not in conflict.

It remains entirely possible that the worst outcomes described above will be avoided, through the combined effort of careful designers, sensible regulation, and an unusually alert citizenry. The author hopes so. He does not, on the evidence currently before him, expect so.

Either way, the people who design the interfaces between human beings and this new medium will, for the next thirty years at least, be doing some of the most important work in the discipline of human-computer interaction. It is the author's hope that this report, in its small way, contributes to a tradition of taking that work seriously.

---

RECONSTRUCTED MAY 2026