Programming Windows: Civil War (Premium)

In early 1997, Microsoft was preparing to release Internet Explorer 4.0, through which it would fully integrate the web browser and its Internet capabilities into the Windows shell. But rather than waiting to deliver this functionality in Windows 98, the next version of DOS-based Windows, the IE team was giving it away for free to users of Windows 95. This would undermine the Windows 98 upgrade, some key Microsoft executives believed, not to mention the next version of NT, setting off a civil war inside the company that would derail strategies, careers, and Microsoft’s booming Internet product division.

The seeds of this discontent were sown a year earlier, when Microsoft hosted a Professional Developers Conference (PDC) in San Francisco. Microsoft had introduced the world to Win32, its 32-bit Windows APIs, and Windows NT at a PDC in 1993. So, it used its March 1996 PDC—known as the “Internet PDC”—to signal the seriousness of its shift to the Internet.

“For software developers, Microsoft’s developers’ conferences were the equivalent of papal encyclicals,” Wall Street Journal reporter David Bank writes in Breaking Windows: How Bill Gates Fumbled the Future of Microsoft. “Thousands of programmers gathered to find out in what direction the overlords from Redmond would take them.”

That direction was firmly rooted in Internet technologies. But there was a battle raging internally for the soul of Microsoft, a battle that none of the several thousand developers in attendance at PDC 1996 could detect from Bill Gates’ steady recitation of Microsoft’s strategy. On one side was Brad Silverberg, the man who had most recently led the development of the highly successful Windows 95. And on the other was Jim Allchin, who had been leading the development of Cairo, a planned second major release of Windows NT that never materialized.

Since shipping Windows 95, the two men’s roles had changed. Silverberg had shifted over to lead Microsoft’s new Internet Platform and Tools Division, taking development of Internet Explorer with him. And Allchin now led all Windows development—both Windows 95 and Windows NT—and planned a unification of the two product lines.

They did not agree on Microsoft’s direction at all.

Silverberg was pushing for Microsoft to move past Windows and create an “Internet platform” that would form the basis of the next decade of advances. As Banks describes in Breaking Windows, the word platform was “explosive” within Microsoft, as it indicated internally and externally that this thing could replace Windows. And as Microsoft swiftly and completely embraced the Internet, it seemed that it would do just that.

But Allchin saw Microsoft’s focus on the Internet as an overreaction that threatened his unification plans. Windows was Microsoft’s platform, he argued, and Silverberg’s Internet offerings should be described less loftily as frameworks—frameworks that ran on Windows—and not as a platform.

Allchin was right to be worried about Silverberg’s plans: Silverberg literally wanted to turn the tables on Netscape and its goals of creating a web-based platform of its own by embracing, extending, and extinguishing the competition. In Silverberg’s view, it should be Microsoft that created the pervasive Internet platform that obsoleted Windows, not Netscape.

Indeed, in his first official appearance as the leader of Microsoft’s Internet Platform and Tools Division internally, the first slide in Silverberg’s presentation was titled “Let’s obsolete Windows with Windows + ActiveX.” For the next year and a half, Silverberg’s plan became Microsoft’s strategy, and the Internet Platform and Tools Division set out to “reposition developer priority from Win32 to ActiveX.”

As PDC 1996 unfolded, Microsoft had racked up some impressive wins. The newly componentized Internet Explorer 3.0 had hit 40 percent usage share to Netscape Navigator’s 60 percent, and it was on track to surpass Navigator by the end of the year. Microsoft had licensed Java from Sun Microsystems. It had convinced arch-enemy Steve Jobs, now running NeXT, to demonstrate his firm’s WebObjects technologies at the show. And one day after AOL, the leading Internet company, had announced a deal to bundle Navigator with its service, Microsoft had convinced the company to reverse course and bundle IE instead. (As it turns out, IE’s componentization made it technically superior to Navigator and a better fit for AOL’s needs.)

Microsoft, it seems, could not be stopped.

And so it is somewhat fascinating to realize that, in this moment of time, Microsoft wasn’t just talking about the Internet, wasn’t just releasing a few products here and there; it was feverishly racing forward embracing the Internet to such a degree that it was ready to sacrifice its core product, the foundation of its very success. Bill Gates was so thoroughly consumed by the Internet, by beating Netscape—and, eventually by beating AOL as well—that he, and thus Microsoft, was ready to turn its back on Windows.

This was not lip service. The Microsoft of 1996 and early 1997 was implementing real change, was moving down a path it would not again travel until Satya Nadella took over as CEO in 2014, almost 20 years later.

Microsoft was “reinventing the company,” Banks wrote, “crafting a new positive image, and establishing a new set of working relationships with the rest of the industry.” Microsoft would “make the transition from defensive adolescent to confident, respected adult. Silverberg was advocating a new way of gaining a competitive advantage,” of “embracing and extending Internet standards and working in an open, cross-platform way … Silverberg’s group pledged to support the Internet’s open standards and to submit specifications for its new features to industry standards bodies.”

“Netscape said they were moving on Internet time,” Bill Gates later observed. “And they were quite a bit ahead of us. So, we must have been moving on two-times Internet time. Otherwise, how do you catch someone moving on Internet time?”

Standing at the podium addressing PDC 1996, Gates presented his vision of the future. Microsoft would evolve Internet standards with the industry. It would bring “active pages” to the Internet that could include electronic commerce with encryption and other security controls, 3D and multimedia, and quality of service. To many in the audience, Gates’s shift to the Internet must have seemed familiar: Microsoft was, to date, infamous for delivering software like Windows and NT and ran poorly on the hardware of the day but would come into its own as hardware performance caught up. Likewise, the Internet that Gates described in 1996 was nothing like the reality of slow dial-up speeds and spotty reliability.

But it was a compelling vision. And it involved evolving Windows such that it would integrate with this wave of Internet technologies, switching the way that developers interacted with the OS away from Win32.

“The idea is that when you’re using local data or using remote data [in Windows], it should [work] the same,” Gates said. “What we want to do is have everything that comes up on the screen use the browser. The browser will be at the center of the system. What used to be called User, the central part of Windows that does dialogs, that will be our HTML engine. Extended HTML will be everywhere: Forms, packages, dialogs; our Help system won’t be a separate EXE now, it will simply use the Extended HTML. The editor we will have built-in to Windows will let you compose HTML forms; that’s the successor to WordPad which we have today. By doing that, the browser is always in the working set.”

The slide Gates was speaking to noted that Windows would provide one interface for files, messages, pages, and documents, and that interface would be HTML-based. There would be one standard for graphics and multimedia, and it would come from the Internet. There would be quality of service integration in Windows, because content would come from the Internet and it needed to be reliable.

This unification, this “synthesis,” would also help with ease of use, Gates said. That was a common understanding of the day, that the double-click method of opening files in Windows would give away to the single-click method used on the web. This notion was based only on supposition and would later be proven incorrect. And it wasn’t the first time, nor the last, that Microsoft would seize on some notion because it seemed to make sense, and not because it did any testing to determine whether its hypothesis was correct.

(Another classic example of this thinking is Microsoft making the toolbars in each Office application as similar as possible; the theory was that a user familiar with Word would be able to quickly transition to an unfamiliar application like Excel because of the similarity of their toolbars. This was later proven to be wrong.)

Microsoft’s embrace of Internet technologies and standards was laudable, and its desire to move the platform forward—to evolve or “extend” it—was likewise ambitious and surprisingly forward-leaning. There were real-world benefits to this shift, not just on the client, but also on the server and in businesses, where new Internet-based internal networks called intranets were starting to spring up everywhere.

But the messaging rang false in some quarters. Most notably inside of Microsoft.

“It was such a disaster, such a disaster,” NT lead Jim Allchin said later, recalling PDC 1996. “Sure, it looked good. But what was it? What pieces of code were included? Can you get me a programming manual for this? I want to check out what kind of code I can write here.”

Taken in the context of Allchin’s internal debates with Silverberg, his complaints could be seen as self-serving. But with Win32, at least, Microsoft had a clear and concise message for developers. Here is the API you should be focused on. With the Internet, especially in early 1996, it was all so vague.

Worse, because Silverberg and his Internet Platform and Tools Division owned the browser, that also meant that they owned the Windows shell. And that meant that Allchin and others who were plotting a way forward for Windows couldn’t control their own destiny. As bad, the code that they did get from the IE team was unreliable and poorly written. Compared to the computer scientists working on NT, the IE guys seemed like children. And they weren’t just going to ruin Allchin’s baby. They were going to destroy Microsoft’s most important and foundational business.

Starting with IE 3.0, the Internet Platform and Tools Division also greatly expanded the selection of applications it provided. In addition to the browser, Silverberg’s team delivered programs that might otherwise have been included in Windows and thus formed part of each release’s value proposition: Outlook Express, NetShow, NetMeeting, FrontPage Express, and other applications didn’t just ship outside of Windows, they were given away for free.

“Grabbing more and more of Windows and shipping that everywhere for free, that’s a way to destroy a business,” Allchin complained. “The team was just grabbing the source code out [of Windows] and then shipping, using it any way they could to serve their mission. That means that the next version of Windows [Windows 98] doesn’t have any of those features [uniquely], because they were already given away for free. Absolutely, Windows 98 was hurt.”

Based on Gates’ instructions, all that mattered was browser share, and Silverberg argued that revenues were “not the right metric” for Microsoft’s success in this market. This was insanity to Allchin. And as PDC 1996 concluded, he set about to regain control of the Windows user interface. And to get Microsoft back on track with Windows as the center of its strategy for developers and customers.

The resulting memos, and there were many, factored prominently in Microsoft’s coming antitrust trial. Allchin, who only wanted what he felt was right for Windows and thus for Microsoft, was later cast by DOJ lawyers as a dastardly villain for wanting Microsoft to “leverage” Windows, and for using Windows marketshare, and not IE usage share, as the company’s key metric. But to a computer scientist like Allchin, it was not illegal for one to leverages one’s strengths, it was only logical.

“The platform is Windows,” he argued, again and again and again.

Finally, Allchin took his case directly to Bill Gates, noting that Microsoft was about to lose its most important franchise if the firm didn’t step back from the Internet cliff. He noted that Silverberg’s team was now talking about bringing the IE integration pieces to non-Windows OSes as well, a move he called “a disease within Microsoft.”

Gates, who had expressed his own suspicions and fears of Microsoft’s Internet embrace, finally agreed with Allchin. Windows should be Microsoft’s “single point of integration.”

“I agree that making sure applications are primarily on Windows is something that we have lost [sight] of,” Gates wrote to Allchin, adding that the notion that Microsoft would advance features for Windows competitors was the result of the “free lunch syndrome we have allowed to develop [internally].”

Microsoft in 1995 had “turned on a dime” in embracing the Internet. Now, just two years later, it was putting on the brakes. He bluntly told his business leaders that Microsoft’s strategic advantages were more important than its customers. He instructed them to lock users into “Office/Windows/Microsoft” specific Internet functionality, to halt work on industry-standard HTML support in Office, and to contort Microsoft’s Java so that it included Windows-specific functionality. The old Microsoft was back. The old Windows was back.

As for IE 4.0 and Windows 98, it was too late to roll back the clock on what Microsoft had given away for free. But the vaunted Internet integration in the file system and elsewhere was dramatically scaled back. The one-click file selection method was disabled by default, because of performance problems and user confusion. And instead of pushing a content-forward model that would benefit any web user, Microsoft instead created a partner-driven “channels” model by which well-heeled content producers could pay the firm for exposure in Windows 98.

By the time PDC 1998 rolled around, everything was different.

Silverberg’s Internet Platform and Tools division had been dismantled, with IE and shell development handed back to Allchin and his Windows teams. Silverberg was placed in charge of Office, a product line for which he harbored no affection, and he eventually went on sabbatical and then, after a short period of consulting, left Microsoft for good.

Allchin, meanwhile, continued with his plans to integrate the Windows 9x and NT lines into a single family of products. The original goal was for this to happen in 1999 with Windows NT 5.0, but that was eventually pushed back to a future release.

“The right thing happened,” Allchin later said in an interview with David Bank. “A whole bunch of confluences happened … Confusion from developers was certainly one. Confusion internally was another. I don’t want to be remembered as the guy who destroyed one of the most amazing businesses in history … I know that when the day is over, I will be judged as ‘am I the guy that kept the company funded?’ There is tremendous responsibility on me to not screw up … I’m on watch.”

Microsoft would stumble badly in its coming antitrust trial, and Allchin’s strategy and poor performance in court played a key role in its guilty verdict. Worse, under Allchin’s watch, Microsoft would stumble even more badly as he tried again to make his Cairo dreams come true. But we must at least give Allchin credit for one thing: He did successfully unify Windows 9x and NT, finally, in 2001.

But that, of course, is a story for another day.