It’s 20 years since Microsoft first launched Internet Explorer, on Aug. 16, 1995, as part of the Microsoft Plus Internet Jumpstart Kit for Windows 95. But rather than a celebration, Microsoft sends out a different message via the new Edge browser in Windows 10: “Thanks for all the great work you’ve done, but I got it from here.”
Twenty years on, Microsoft is finally moving away from Internet Explorer and trying to lose the association with IE6—a browser that went from being ahead of its time to a fossilized platform holding companies back. By IE10 and IE11, the browser you love to hate had become a modern browser that was discriminated against as much for reasons of history and loyalty as for any real technical reasons. Edge keeps only the HTML5 standards support from what would have been IE12, losing the ever-increasing burden of legacy compatibility that ended up cramming three versions of the rendering engine in to one browser in favor of a clean break.
But dismissing IE as a legacy system misses out on key developments in browser history, as well as the scope of Microsoft’s ambitions for the web. The story looks a little different when you realize how much Microsoft had in its grasp in the 1990s and how much of the web platform that Google and Mozilla have brought us since Redmond let slip through its fingers back.
Why were there so many missed opportunities? Why did IE drop the ball, what made Microsoft wake up to the potential of the web—and has it found a way to manage compatibility and still stay modern in the world of living standards that never stop changing?
Internet Explorer was too early, not too late
IE spent years catching up to the other modern browsers, and Microsoft went from seeming to ignore the open web to being so active in the W3C that it participates in around 40 working groups, chairing several of them. That’s not Microsoft having ignored the web until it was too late, though. In the kind of irony that has affected Microsoft disturbingly often during its history, that’s Microsoft having been too early to the party, then changing its mind and going home before the cool kids arrived.
Stranded on the Cornell University campus at the end of a recruiting trip in February 1994 by a snowstorm, Steven Sinofsky—then technical assistant to Bill Gates—wandered through the computing rooms to find students using not Microsoft Office and other desktop software but web browsers. He emailed Gates and his team a warning: “Cornell is WIRED!”
That got more attention than J Allard’s memo the previous month, which he’d titled “Windows: the next killer application for the Internet.” Allard had been trying to get Microsoft to take the Internet seriously since he joined the company in 1991 and created the company’s first Internet server as part of a skunkworks project. He thought Microsoft should build its own browser and tried to convince Russell Siegelman to base Microsoft’s planned MSN service on web technologies rather than a proprietary system.
Rob Glaser, who had successfully got Microsoft into multimedia, was suggesting the same idea. But when Siegelman suffered a serious illness, Glaser decided not to push the MSN team to decide in his absence. That turned out to be an expensive decision in the long run, but the Internet enthusiasts had caught Gates’ attention. At an executive retreat in April 2004, Gates and other key Microsoft executives argued their way through a 300-page briefing put together by Sinofsky, and Gates spent his annual Think Week retreat concentrating on the Internet.
The problem was that Gates didn’t see how you could make money from the Internet. The group at the retreat decided to add TCP/IP support to Windows 95 and give Word the option of saving documents as web pages but nothing more radical. That soon came to look like a mistake as Sun worked on Java, promising a future where the web could deliver programs to any computer, and PDF started to replace Word files on the Internet as a way of distributing documents. Both Windows and Office began to look threatened. Bill Gates’ famous Internet Tidal Wave memo came out in May 1995, a few days after Sun launched Java, and in November 1995, Goldman Sachs took Microsoft off its “buy” list because of the Internet threat.
Well before that, in late summer 1994, Microsoft had started work on its own browser, first trying to buy the BookLink browser Sinofsky spotted at the Comdex show in 1994 (AOL snapped it up for $30 million), then licensing code from Spyglass, which had the rights to NCSA Mosaic—the very first graphical browser, built by graduate students who worked for Larry Smarr.
Then the head of the National Centre for Supercomputing Applications, now running the California Institute for Telecommunications and Information Technology and a pioneer in quantified health, Smarr remembers pitching businesses on the huge potential web browsing might have for them—if only it had a graphical interface.
Everyone has heard of one of Smarr’s students who worked on the Mac version of NCSA Mosaic: Marc Andreessen, who went on to found Netscape with SGI’s ex-leader Jim Clarke. They set up the company the day after the Microsoft executive retreat, originally as Mosaic Communications until lawyers pointed out that Spyglass owned the Mosaic trademark.
Not as many have heard of Chris Wilson, who co-authored the Windows version of Mosaic and went on to work at Spry (the first Mosaic licensee). When Wilson moved to Microsoft in summer 1995, it wasn’t to join the IE team; he worked on web search and was part of the team building Blackbird, a graphical interface for creating content for MSN’s proprietary system, designed to compete with the dial-up AOL service.
But the fact that someone who had already built a web browser was working at Microsoft didn’t go unnoticed, and Wilson (who later became the leader of the W3C’s HTML working group and now works at Google) joined the IE team soon after the first version shipped in the Windows 95 Plus Pack. In 2006, he joked that he’d spent 10 years working on IE and would spend the next 10 years making up for it.
The Internet Explorer team grew and grew, from a half dozen people working on the first version under Ben Slivka to nearly a hundred people working on IE3 under Brad Silverberg, who was fresh from the success of delivering Windows 95 and became head of the new Internet Platform and Tools division in February 1996. By the time IE5 shipped in 1999, there were 1,000 people on the IE team, but it was Silverberg’s team of “superstars” (as he and the other managers of IE still refer to them) who worked day and night to build what they thought was the future of the company.
“The most incredible product team I ever worked with,” Silverberg recalls. He continues:
Such a small team, so many unbelievable superstars all working together as a team in some of the most inspired work of their career, under massive pressure and the highest possible stakes. A bit like the original Mac team, the IE team felt like the vanguard of Microsoft, the vanguard of the industry, fighting for its life. Culminating in IE3, which was a brilliant product and changed the rules of the game, both for the industry and for Microsoft—showing Microsoft could be a leader and a good citizen. It was a reinvention of the Microsoft culture.
“Our work was more than work,” remembers Hadi Partovi, the leader of IE product management until IE5.
It was a passion and life mission. We ate all our meals on the job; we worked very, very late nights. I would often go to sleep under my desk at 6 a.m., only to wake up the next morning at 8 a.m. ready for work. We had this sense that this multi-billion-dollar company was going to lose its future unless we could get ahead of the Internet wave, and that meant having the No. 1 browser on the planet.
Getting Internet Explorer to No. 1
In the early 1990s, Internet Explorer was one of a dozen-plus web browsers on Windows and Mac. Microsoft and Netscape competed to add new features for building more powerful websites in successive versions, as well as competing over the best way to build a browser business. Was it was fair for Microsoft to include a browser free in Windows when Netscape was charging up to $12 for it (even charging magazines to distribute evaluation copies on their cover discs, then charging them again when their readers bought a license)? Did Microsoft buy its market share by pressuring PC makers and bundling IE with Windows or earn it by building a more powerful browser that didn’t crash as often?
Netscape CEO James Barskdale was always ready to mock Microsoft’s ability to create a browser and web server. Silverberg once thanked him for the “trash talk [that] helped us get motivated,” and Paul Maritz, then group vice president for platforms (including the web, which meant he was boss to both Silverberg and the new head of Windows, Jim Allchin) pointed out that “the thing that really motivates [Microsoft] is paranoia and competition.”
The war of words was matched by the battle of leapfrogging improvements—and senior figures on both sides look back today and call their competition worthy opponents. That’s despite the time the 10-foot IE logo created for the IE4 launch party ended up on the lawn at Netscape overnight. Netscape employees retaliated by knocking it to the ground, spray-painting Netscape on the side, planting the Mozilla dinosaur mascot on it, and sticking up a sign saying “Netscape 72, Microsoft 18” to show off their far better market share.
Netscape developed the LiveScript programming language, which it launched as JavaScript (although with only a tenuous connection to the successful Java language). Microsoft came up with Dynamic HTML behaviors and won friends at Internet service providers by giving them tools to add their own branding to IE. But Microsoft also supported JavaScript—and used it to build what was probably the very first web app. To be able to show your Exchange emails in Outlook Web Access in IE5, Microsoft developed XHR, the basis of what we now know as AJAX, a technique on which many web apps are still based.
IE was the first browser to have autocomplete, in the address bar and in forms, and you could argue that IE’s channels for following content were the first version of what became RSS. IE3 had the first commercial implementation of CSS in August 1996; Netscape didn’t add that until Netscape Communicator 4, which shipped almost a year later.
IE3 wasn’t only a browser or even a standards-based browser (a very un-Microsoft approach in those days). Years before the idea of web apps was widespread, IE3 was designed to put the web everywhere. “The vision for IE3 was to change the rules of the game and go all in on the Internet,” Silverberg recalls.
The idea was that the Internet should be part of every app, not just something confined to the walls of a browser window. We componentized IE3 with a very elegant architecture, so that anyone could build a browser, so that anyone could include whatever of the Internet they wanted in their app. We won over AOL to build their Internet client using the IE3 components. I viewed IE3 the browser as just another app using the IE3 components. Any app now could incorporate HTML for example into the app—say, for displaying dialogs or for help.
By IE4, Microsoft’s browser included many of the foundations for the web as we know it now, and IE5 was widely hailed at the time as a better browser than Netscape. Even Firefox VP Johnathan Nightingale agrees: “Whatever else, IE5 was a solid product.”
Partovi jokingly puts some of that down to the “latte challenge” he ran in the IE team; if anyone could find more bugs that crashed the browser than he did in a week, he’d bring them a latte every day for the next week. It was a fun way to motivate developers to focus on making the browser reliable enough to use for real work.
Microsoft’s key decision: Compatibility or standards?
Even with all the improvements, both browsers caused problems for developers. Navigator’s popularity meant that other browsers would try to match its rendering of web pages, even when that didn’t match HTML standards. (Spyglass developers had always found it frustrating and the IE team would sometimes refer to this as being “bug compatible.”)
Developers at Mozilla suggested having two rendering engines: one for Netscape-coded pages and one for more standards-based HTML. Microsoft picked up the suggestion in IE5 for Mac as a way of dealing with the problem, but the decision instead created issues that have dogged it to this day.
“At least half of the quirks were Netscape quirks that IE built in to be backward-compatible,” claims IE6 developer David Aronchick. Jason Farnsworth, who worked on the Trident rendering engine for IE6, agrees: “We were working on a codebase with intentionally broken layout capabilities (to match Netscape rendering bugs), trying our best to make something standards compliant and ‘right’ for the first time.”
There are several reasons that IE’s rendering tended to be different from that of other browsers, especially if developers took advantage of the unique Microsoft features in IE. IE6’s notorious quirks mode was introduced to load pages that were built for IE5.5, but it turned to to have a bug. Unless the document type was in the first line of code, it was never actually used, recalls Chris Jackson, Microsoft’s compatibility specialist.