Welcome to Chapter II. In short order I learn that Microsoft is way behind in the products I work on. BillG is really concerned about Steve Jobs NeXT computer. I move offices and find out I was part of a reorg.

This is a quick setup for a chapter about the creation of Microsoft Visual C++ 1.0.

Back to 007. Windows 3.0 Buzz

1990 to 1992: The big bet on Windows begins to pay off, but Microsoft struggles to win over developers who were enamored with object-oriented programming and C++ to more easily build GUI programs. Within Microsoft, differing cultures emerge between groups, and eventually define the company.

Microsoft was starting to lose its edge, even with Windows taking off.

The company got its start with languages and developer tools, but by the early 1990s tech enthusiasts and hobbyists were moving away from BASIC toward more advanced or professional languages and tools. Developers were being wooed by an exciting upstart, Borland International. Led by an energetic Frenchman, Philippe Kahn, Borland captured the hearts and minds of enthusiasts with a line of TurboTools that integrated a compiler, code editor, and debugger into one slick package. It was fast, really fast—fast the way that got under the skin of BillG. With support for both Pascal and C, and priced favorably, the products and the company became a favorite among independent developers. It also didn’t hurt that Borland expanded its product base to include a killer spreadsheet in Quattro Pro (get it, it came after Lotus 1-2-3) and Paradox, an industrial-strength database for MS-DOS.

Microsoft secured the professional end, particularly with Microsoft C 5.1, the product we were using for our ET++ experiments. The successor product, Microsoft C 6, re-upped the professional standard but was late (as everything was) and lacked the pizzazz of Borland. Microsoft responded to Borland with Quick C, a similarly integrated all-in-one product for MS-DOS and with C 6, timed with a Windows version as well. Quick C was viewed as a defensive move. It was. We were focused on the high-end commercial developers, not the hobbyists or solo developers.

Being squeezed from the low end was one thing, but the high end was becoming problematic for Microsoft. Not only was the C 6 product late, but it was C and neither C++ nor object-oriented. This challenged Microsoft’s perceived leadership. In addition, Borland’s products performed better than the anticipated C 6. Internally, the teams, particularly Excel, began testing whether they could move to the Borland tools. This was especially noteworthy because it was coupled with a move away from Microsoft’s proprietary C-like language CSL.

As if this wasn’t enough, the growing importance of the graphical user interface would soon require a wholesale reinvention of tooling. Microsoft was way behind. Windows was our platform, but we lacked convincing and competitive tools. Squarely between Windows and OS/2, the Tools teams all but sat out developing new tools for GUI programming and focused simply on the programming language. The additional tools for developing the interface of menus and dialogs, as well as the complexities of making a GUI program, were left to the respective platform teams. These teams shipped tools in a software development kit (SDK), which was complete but not as polished a product as Borland sold.

This was my first experience with disruption, though that word was years away from business vocabulary. In 1990, we just called it competition. And losing.

While our marketing team talked all about revenue and market share, like any good business, in reality Tools was not really a business. An important lesson about Tools and platforms that I was learning in real-time was that a robust platform company invests (that is spends money on) Tools at an irrational level to support the platform. The reason Borland could have a Tools business was because it was spending much less than Microsoft, and so could be profitable. Microsoft was spending more money and making a worse product. In other words, I was part of an irrational investment that wasn’t paying off. We were a poor business, and we were failing at building tools professionals wanted to use. Losing control of the Tools was akin to losing control of the platform.

Apple invested heavily in tools for the Macintosh, just like a winning platform should. There was also a vibrant market for many different languages and tools for creating apps for the unique graphical platform. This was well known within Microsoft because so many of the Apps engineers got their start writing Mac software in college, including me. There was always a sense of envy regarding the elegance of building GUI applications with a GUI toolset—bootstrapping was when programming tools created themselves. Historically, programmers viewed bootstrapping as an important, if mostly symbolic, step in developing a platform. Microsoft was far from bootstrapped, as it was still using Xenix and OS/2 to develop for MS-DOS and Windows. Worse, most developers at Microsoft thought this was a superior way to build software. All that marketing we were doing explaining that a GUI was easier to use had the effect of telling programmers that GUI was how lesser programmers worked. Worse, that is what our own marketing team was telling us about our own customers.

Besides, even with those great tools and a lead of several years, Apple was still far behind in PC sales. Steve Jobs was no longer at Apple and his new company, NeXT, was top of mind for BillG and the industry. As successful as Windows was for Microsoft, there was a strong belief that no single system would dominate. NeXT was new. NeXT was led by Steve Jobs. And most of all, the product was clearly innovative and setting the bar by which BillG would judge our product and technology.

Steve Jobs at an event in San Francisco launching updated NeXT computers and tools. A key part of this presentation that caught Microsoft’s attention was the presence of Lotus CEO Jim Manzi describing how they built a unique new spreadsheet product on NeXT, called Improv. “We would not have been able to invent such a revolutionary new product on any other platform” definitely gets your attention.

Microsoft had to do something, something about Borland and NeXT.

Moving offices at Microsoft was a constant. It was time for our first move, to new buildings that were even bigger than the big double-X layouts. These were the new buildings for Apps: 16, 17, and 18. Rather than low slung grey these new buildings were 3 stories of glass and brick and featured their own courtyard and fountain, which would be the site of our future ship parties. The buildings had huge atriums of open space and skylights while still maintaining the sacred 9x12 foot single office with a door. The three buildings were connected by an enclosed tube system that looked like a Habitrail. The uniqueness of these tubes meant that they were frequently used for photo shoots and videos.

The night before a move (my first of a dozen I would experience), MSMOVE dropped off a stack of boxes, a roll of tape, and a move form. There was a move form to draw placement of the huge oak desk, return, guest chair, and developer folding table at the next home. The movers showed up at the end of the day, unplugged phones and computers, and put everything on carts and trucked them to the new office. Unpacking could happen 12 hours later. The MSPHONE person eventually showed up to make sure the newly hooked up phone worked with your assigned number.

This was going on every night in every building at Microsoft. It was like a giant nine-square puzzle, where at any given time an entire group existed in the moving trucks in the parking lot because there was never enough space for everyone, even with the new buildings.

My new office was right around the corner from a connecting tube and for the first few months I could hear the door slamming every time someone entered the tube. There was a design defect that turned the tube into a wind tunnel, making the door difficult to open while also forcefully closing it with high-speed wind. Eventually some vents were added solving this problem.

As with every hallway of developers, there was a deep sense of personalization that came with the private office space we were each allotted. Since most people were not yet grown-ups, there were no family photos, or traditional memorabilia. Rather, offices were filled with some form of personal collection representing youth. I saw a collection of vintage car license plates, a wall of album covers, and a beer can collection sitting on a custom shelf high up on the wall. And most every office featured a pyramid of used beverage containers, usually Mountain Dew or the ever-present Washington apple juice, and the occasional chocolate milk containers (the ones from elementary school).

Next to each door, there was a full-length window (a relight), about one foot wide, made with that type of institutional glass from high school with wires inside that gave a sense of involuntary confinement. While designed to let light in form outside, it often served as an outward-facing sign of personal expression. Usually the first thing decorated in a new office, relights were covered in stickers, signs, news articles, Dilbert comics, jokes, and bad code examples or bug burn down charts, all expected to be updated with some frequency.

Like the first day of high school, people wanted to stand out, but they didn’t want to stand out. I nervously attached a few items to my glass, mostly American tourist-trap kitsch from my recent cross-country drive: a Wall Drug sticker, a Corn Palace postcard from South Dakota, and a copy of Elvis Presley’s death certificate I had acquired in Memphis. Road and street signs were popular, and my “Slow Children At Play” road sign also followed me around for a decade.

I didn’t realize it but our ET++ team had just been reorganized. Not only did I move offices, but I found myself on a new and bigger team with a clarified mission.

On to 009: Password is ‘NeXTStep’