Mosaic, Netscape, and the commercial inflection
The web that Tim Berners-Lee released from CERN in 1991 was a research artifact. It had one server, a few users, and a browser that ran on NeXT machines. The web that became the substrate for civilization was a commercial product, built on the same protocols but with a different cast of characters, different incentives, different priorities, and a different sense of what the project was for. The transition between the two webs took less than four years, and the principal actors in it were Marc Andreessen, Jim Clark, the team that built NCSA Mosaic and then Netscape Navigator, and the financial markets that turned the Netscape IPO of August 1995 into the most consequential commercial moment in the history of the internet. This chapter is about that transition: how Mosaic happened, how Netscape happened, what the IPO meant, what the resulting Browser Wars produced, and how the trajectory of the web bent during the four-year window in which the commercial inflection occurred. The technical decisions of the previous chapter had given the web its shape; the commercial decisions of this chapter gave the web its direction.
Mosaic at NCSA
The National Center for Supercomputing Applications was founded at the University of Illinois Urbana-Champaign in 1986 as one of the supercomputer centers funded by the National Science Foundation. NCSA had, by the early 1990s, become a hub for scientific computing and visualization work, with a substantial software development operation. One of the projects at NCSA was the Software Development Group, where a handful of undergraduate and graduate-student programmers worked on visualization tools, network software, and various internet utilities. Marc Andreessen was one of the undergraduates in this group. Eric Bina was a staff programmer in the same group. Together, beginning in late 1992, they began work on a web browser for X Window System, intended for the Unix workstations that were the dominant scientific computing platform at the time.
The browser was released on January 23, 1993, as Mosaic for X. It was free, freely redistributable, and almost immediately popular. Mosaic was not the first web browser — Berners-Lee’s original NeXT browser preceded it by two years, ViolaWWW and a handful of other browsers existed before Mosaic — but it was the first browser that was both easy to install on common Unix workstations and graphically rich enough that the pages looked like something. The IMG tag, treated in the previous chapter, was Mosaic’s most consequential addition: inline images, fetched alongside the text and rendered in place, made web pages visual in a way that the previous browsers had not allowed. The combination of accessible installation, graphical presentation, and rapid release cadence put Mosaic ahead of every other browser within a few months of its release.
A small team at NCSA, building on the Andreessen-Bina X version, produced ports of Mosaic to Microsoft Windows (released in August 1993, with Chris Wilson and Jon Mittelhauser as principal authors) and to the Apple Macintosh (released in September 1993, with Aleks Totic and Mike McCool as principal authors). The Windows and Mac ports mattered more than the X version had, because they brought the web to consumer hardware. By the end of 1993, Mosaic was available on every major personal-computing platform, and the number of web servers in the world had grown from about fifty at the start of the year to several hundred by the end. The growth curve traced Mosaic’s availability closely.
The Mosaic team had a reputation, at the time and since, for moving fast. Features were added quickly. The IMG tag, the form input controls that let users submit data, the early bookmarking and history features that made the browser usable for sustained reading — these were added during 1993, in cycles measured in weeks rather than months. The development pace was a real competitive advantage: by the time other browsers were catching up to Mosaic’s December feature set, the Mosaic team had already moved on to the next round. The pace was sustainable because the team was small, intensely focused, and operating in the kind of resource-rich but unsupervised environment that NCSA had become for the small group of people who worked on web software there.
It is worth noting what Mosaic was not. It was not invented in a corporate research lab. It was not the product of years of design and review. It was an undergraduate’s project, with a staff programmer’s support, written in a few months on a tight schedule, in a culture that valued shipping. The technical choices were pragmatic. The design priorities were “what makes users like the browser more, ship it.” The discipline that had characterized hypermedia research — the careful debate over linking models, the formal apparatus of the Dexter Reference Model, the attention to argumentation structures and version control — was largely absent from the Mosaic development process. This is not a criticism. Mosaic worked. The discipline that was absent was discipline appropriate to a different scale of project.
Andreessen leaves; meets Clark
Andreessen graduated from UIUC in December 1993 and left NCSA for Silicon Valley. The immediate reason was that NCSA had begun to assert ownership over Mosaic and was, in his view, marginalizing the team that had built it. The deeper reason was that the commercial opportunity Mosaic represented was, by late 1993, visible to anyone who looked: hundreds of new web servers per month, growing demand for browsers, a software market that did not yet have a dominant commercial product. Andreessen moved to California and took a job at a small startup. He had been there for a few months when he was contacted, in early 1994, by Jim Clark.
Clark was the founder of Silicon Graphics, the visualization-workstation company that had dominated the high-end graphics market through the late 1980s and early 1990s. He had recently left SGI and was looking for the next project. A mutual acquaintance had suggested Andreessen as someone worth talking to. The meeting happened. Clark, who had the capital and the network to put a serious software company together, asked Andreessen what was worth building. Andreessen’s answer was a commercial web browser, with the original Mosaic team recruited away from NCSA, building a product designed for the consumer and commercial markets that were about to emerge.
Clark agreed. The company was founded in April 1994, in Mountain View, California, originally as Mosaic Communications Corporation. The founding team included Andreessen, Bina, Totic, Mittelhauser, McCool, and several others from the NCSA Mosaic project. Clark put up the initial capital and brought on additional investors. NCSA, when it learned the company had taken its name and its team, threatened legal action. The company changed its name to Netscape Communications Corporation in November 1994, settled with the University of Illinois for a modest payment, and released its first product — Netscape Navigator 1.0 — in October 1994 (with the final version shipping in December).
The transition from NCSA Mosaic to Netscape Navigator was, on the surface, the same team producing a successor product. Underneath, it was a transition from research to industry. Netscape was a venture-backed company with a billion-dollar ambition. The team had to ship a product on a quarterly cadence, with a roadmap aligned with the company’s commercial plans, against competitors who would soon include the largest software company in the world. The development culture had to industrialize. The features had to drive adoption that would justify the valuation. The web, which had been a research project, was now a market.
Netscape’s product velocity
Netscape moved quickly. The first commercial version of Navigator shipped in late 1994, free for personal use and paid for commercial use. Within a few months it had become the dominant browser, with market share estimates in the 70-80% range through 1995. The product cycled rapidly: Navigator 2.0 in March 1996, 3.0 in August 1996, 4.0 (Communicator) in June 1997, with each release introducing significant new features and reaching tens of millions of users.
The features Netscape added during 1995 and 1996 shaped the web more than the underlying protocols did. Cookies, designed by Lou Montulli at Netscape and shipped in 1994 in early form, gave the stateless protocol a way to maintain state across requests; the cookie became the substrate for everything from logged-in sessions to advertising tracking. JavaScript, designed by Brendan Eich at Netscape in 1995 (originally called Mocha, then LiveScript, then renamed JavaScript for marketing reasons related to Netscape’s partnership with Sun), gave the browser a programming language for scripting page behavior. SSL, designed by Taher Elgamal’s team at Netscape and shipped as SSL 2.0 in February 1995 and SSL 3.0 in late 1995, gave the web encryption-in-transit and made commerce possible. The PNG image format, frames, table-based layout, plugins, embedded multimedia — each of these was a Netscape addition to the web’s vocabulary, mostly shipped before the standards bodies had a chance to specify them, and most of them subsequently standardized in retroactive fashion.
The pace of feature addition had two effects. The first was that Netscape became, briefly, the de facto standards body for the web: features Netscape shipped were features the web had, regardless of what the W3C or IETF eventually said about them. The second was that the web became more complicated, faster, than its specification process could absorb. The W3C, founded in 1994 (treated in the next chapter), spent much of the next decade catching up to what the browser companies had already shipped, with the standards process producing post-hoc rationalizations of decisions made by Netscape and (later) Microsoft engineers.
The features were also, increasingly, features that pushed the web toward applications rather than documents. Cookies enabled sessions, sessions enabled applications, applications enabled the kind of stateful interaction the original web design had specifically rejected. JavaScript enabled client-side logic, client-side logic enabled real interactivity, interactivity enabled the kinds of UIs that desktop applications had been producing and that the web was about to learn to produce. SSL enabled commerce; commerce required logins and shopping carts and payment processing; payment processing required, in turn, more session state, more JavaScript, more application infrastructure. Each Netscape feature shipped for a defensible reason — users wanted commerce, they wanted dynamic pages, they wanted to be logged in — and the cumulative effect was a web that was less a hypermedia document system and more a runtime environment for distributed applications. The shift was not designed; it was the accumulated weight of individual decisions, each defensible in isolation, that bent the medium’s trajectory.
The IPO
Netscape filed for an initial public offering in the summer of 1995. The offering was priced at $14 per share, then raised to $28 the day before. On August 9, 1995, the stock opened at $71, traded as high as $74.75 during the day, and closed at $58.25 per share. The company, which had been founded sixteen months earlier and had never recorded a profitable quarter, was valued at $2.9 billion. The IPO is the conventional marker for the moment the commercial internet became a mass-market financial phenomenon. The internet became a thing that people invested in, that newspapers covered as a financial story, that brought money flooding into anything that could position itself as related to the web.
The IPO had several specific consequences for the web.
It put a price tag on the browser. Until August 1995, the browser was, depending on which one you used, a research artifact, a freely distributed scientific tool, or a low-cost shareware product. After August 1995, the browser was a strategic asset worth billions of dollars. Microsoft, which had been notably absent from the early web, treated the Netscape valuation as a wake-up call: a competitor was emerging in a market Microsoft considered its own (desktop software), and the competitor was being valued at levels that suggested the threat was serious. Internet Explorer 1.0 shipped on August 16, 1995, a week after the Netscape IPO. The Browser Wars had begun.
It made the web a venture-capital category. Before the IPO, web-related businesses had been small startups with limited funding and modest ambitions. After the IPO, web-related businesses were the hottest investment category in the venture-capital world, and capital flooded into anything that could plausibly claim to be web-related. The eventual peak — the dot-com bubble of 1999-2000 — was the direct continuation of the trajectory the Netscape IPO had set. The bubble produced enormous waste (hundreds of failed startups, billions of dollars lost) and enormous infrastructure (broadband deployment, data centers, the early dominance of Amazon and Google) that the long-term web depended on.
It made advertising the default business model. Netscape’s own revenue came from selling the browser, the server software, and various commercial products. The companies that built businesses on top of Netscape’s browser, however, mostly could not charge for what they did: web users had been trained to expect free, and the free model needed a revenue source. Advertising became, by default, the answer. The first banner ads had appeared on HotWired in October 1994; by late 1995, banner advertising was a serious business; by the late 1990s, web advertising was a major industry. The structural commitment of the consumer web to advertising as the primary revenue mechanism dates from this period, and almost everything about the web’s subsequent evolution — the privacy story, the platform story, the algorithmic-feed story — descends from the choice of advertising as the underlying business model.
It changed who was building the web. Before the IPO, the people building the web were mostly the same people who had been on the internet for years: academic researchers, engineers at internet-aware companies, hobbyists. After the IPO, the people building the web were increasingly people who had just arrived, with capital and ambition and an interest in the commercial opportunity. The culture that had built the early internet — open, technical, oriented toward standards and interoperability — was diluted by the influx of commercial actors with different priorities. This was not entirely bad; the influx brought resources and expertise that the academic-and-research community could not have provided. But it was a change, and the web that the new arrivals were interested in building was a different web from the one Berners-Lee had proposed.
The Browser Wars
Microsoft’s Internet Explorer was the company’s response to Netscape, and the response was substantial. Internet Explorer 1.0 in August 1995 was a weak product. By Internet Explorer 3.0 in August 1996, Microsoft had a competitive browser. By Internet Explorer 4.0 in September 1997, Microsoft had a better browser than Navigator in several respects. By 2000, Microsoft had won.
The mechanics of the win were several. Microsoft bundled Internet Explorer with Windows; users got the browser free with the operating system, with no separate download or installation required. Microsoft invested heavily in the browser’s engineering, with hundreds of engineers working on it; Netscape, a smaller company with less capital, could not keep pace. Microsoft used its operating-system position to ensure that Windows favored its own browser; the Department of Justice antitrust trial of 1998-2001 found that Microsoft had engaged in anticompetitive behavior in this regard. By the time Netscape was acquired by AOL in November 1998 for $4.2 billion in stock, the Browser Wars were effectively over, and Internet Explorer was the dominant browser with market share above 90% by 2002.
The Browser Wars are remembered, in retrospect, as a period of fast feature development and incompatibility between browsers. Both Netscape and Microsoft added features competitively, often without standards-body input, with the result that web pages could behave differently in the two browsers. Web developers spent a substantial fraction of their time writing code that worked around browser incompatibilities. The “browser sniffing” patterns of the late 1990s, in which JavaScript code branched on which browser was running and behaved differently in each, are the residue of this competitive environment. The cost of the wars, in developer time and in user experience, was substantial. The benefit was rapid feature development that the standards-driven process probably would not have produced.
The end of the wars, with Microsoft’s victory, produced a different kind of cost. Internet Explorer’s dominance through the early 2000s, with Microsoft having little competitive pressure to improve it, led to a period of stagnation in browser features. IE6, released in 2001, remained the dominant browser for nearly a decade, and its limitations shaped web development through that period in ways that hold web designers’ breath when they remember them. The eventual recovery — Firefox in 2004, Chrome in 2008, the modern browser ecosystem — required the kind of competitive pressure Microsoft’s dominance had eliminated. The recovery is one of the more cheerful stories the web has to tell.
After Netscape
Netscape’s specific story did not end well. The acquisition by AOL in 1998 was, at the time, a substantial transaction, but AOL’s commercial trajectory peaked shortly afterward; the AOL-Time Warner merger of 2000 has been described, by Time Warner’s own executives in retrospect, as a disaster. Netscape’s browser was reduced to a niche product and was eventually discontinued in 2008. The Mozilla project, into which Netscape had released its browser source code in 1998 (a milestone that anticipated by years the open-sourcing of other major commercial products), continued and produced Firefox, which has been one of the most important browsers in the period since.
Andreessen himself went on to other projects: Loudcloud (a cloud computing company, founded 1999), Opsware (the rename and pivot after the dot-com bust, eventually sold to HP for $1.6 billion in 2007), and Andreessen Horowitz (the venture capital firm he co-founded with Ben Horowitz in 2009, which has become one of the dominant Silicon Valley VC firms of the last fifteen years). His role in the broader trajectory of the commercial internet has continued well past Netscape.
Jim Clark went on to start Healtheon (a health-care internet company, founded 1996, merged with WebMD in 1999) and various other ventures. He has been generally regarded as one of the great entrepreneurs of the period, with multiple multi-billion-dollar companies to his credit.
The Netscape engineering team, after the AOL acquisition and the eventual wind-down, scattered across the industry. Several principal alumni went on to other significant roles: Brendan Eich became CEO of Mozilla, then founded Brave; Jamie Zawinski became one of the more visible commentators on Netscape’s history; Lou Montulli, who had designed cookies, continued as an engineer at various companies; the broader team contributed to the development of much of the modern web infrastructure in the two decades that followed.
What the commercial inflection meant
The Netscape story is, in some ways, ancillary to the broader argument of the book. Netscape did not change the web’s protocols substantially; it did not change the URL or the HTTP method set or the basic page model. But the commercial inflection it represented changed the direction the web took in ways that have proven much more important than the protocol-level decisions of the previous chapter.
The commercial inflection meant that the web’s evolution was driven, primarily, by commercial concerns rather than by the requirements of hypermedia as a research field or the requirements of authorship as a humanistic practice. The browser, which was the user’s primary interface to the web, was a commercial product whose success was measured by user adoption and corporate revenue. The standards that emerged were retrospective rationalizations of decisions made for competitive reasons. The features that shipped were features that drove adoption: visual richness, dynamic interactivity, commerce-friendly state management, advertising-friendly tracking. The features that did not ship were features the commercial logic had no incentive to ship: rich linking models, durable addressing, federation-friendly identity, user-controlled customization.
The web after Netscape was, structurally, a commercial product layered on top of a research substrate. The research substrate’s choices — uncoordinated URLs, stateless protocol, permissive parsing — were appropriate to the research project Berners-Lee had been doing. The commercial layer that the Netscape inflection produced was different in kind, oriented toward different goals, and shaped by different incentives. The two layers have continued to coexist, with the commercial layer growing thicker and the research substrate becoming increasingly invisible underneath it.
This is the structure that the rest of Part IV traces. The next chapter takes the standards bodies that tried, with mixed success, to govern the web’s evolution against the pressure of commercial development. The chapter after that takes the browser’s evolution into the modern application platform. The chapter after that takes the platform decade, in which a small number of companies absorbed most of what the rest of us thought we were building. Each of these stories is, in part, a continuation of the trajectory the Netscape IPO set in motion. The web that Berners-Lee specified in 1991 was a research object. The web that the world uses in 2026 is something else, and the Netscape inflection is one of the moments when the something-else came into focus.
The standards story, taken up next, is the story of how the web’s governance tried to keep up with the commercial pace. The story has its own arc — the early IETF dominance, the founding of the W3C, the WHATWG revolt, the eventual peace — and it produces a different perspective on the same trajectory the Netscape chapter has just sketched. The institutions that govern the web are not the institutions that originally proposed it, and the displacement is part of the story.