COMDEX was a major trade show and convention for the computer industry that was started by Sheldon Adelson and Richard Katzeff of The Interface Group. In 1982 at the end of November, Bill Gates went to COMDEX in Las Vegas and he saw Visi On from VisiCorp.
While this doesn't appear to be too advanced, there’s actually quite a bit going on here. Visi On was mouse driven, had a consistent interface design, was multitasking, and applications could share data. The system also had a built-in installer and help. It was limited to 640x200 monochrome CGA, had no icons, and was exceedingly text oriented. The system also lacked scroll bars. To scroll, a user holds the right mouse button and drags. More interestingly, VisiCorp’s GUI was intended to be extremely portable. The system was written in a strict subset of C for a virtual machine called the Visi Machine, and only the Visi Host (the core of Visi On) was sensitive to the underlying hardware and operating system. All applications were written for the Visi Machine and Visi On. This is very interesting, but it was also quite costly. Visi On required a minimum of 512K RAM and 5MB of HDD space. Bill Gates wasn’t pleased. Microsoft, at that time, had nothing like this.
When Gates returned from COMDEX, a graphical user interface to run on IBM PCs became a priority. Microsoft did have efforts to create a shared text-mode UI framework (Interface Manager) and a shared graphics UI framework (GDI) prior to this (starting in 1981), but neither of those were making much headway. In late 1982, the two teams were merged, and they then began work on Microsoft Window Manager.
Phil Lemmons mentions in his December 1983 article in BYTE, Microsoft Windows, that Microsoft intended for Windows to be cheaper in purchase price and in system requirements than any competition. The target was two 360K floppy disks, CGA, and 192K RAM. To sweeten Windows as a target for developers, all development could be done in MS-DOS or Windows itself. This is in contrast to Visi On where a UNIX machine was required for development, or the Apple Lisa where the price started at $9995 (around $31000 in 2023). Additionally, even at this early stage, Windows was an operating system all its own and not merely a DOS shell as is commonly thought. To draw to screen or to a printer, Windows used GDI (the same mentioned earlier). This was the Graphics Device Interface. It provided an abstraction layer for developers, and a driver system for hardware. The result was that developers needn’t worry about the underlying graphics hardware at all, and they could simply target Windows. Also, while applications for Windows were cooperatively multitasked, DOS programs were (by necessity) preemptively multitasked via WINOLDAPP (WINdow OLD APPlication). In the event that a DOS application wrote directly to video memory making any multitasking impossible, Windows would terminate and stay resident allowing the application to do its thing, and then restore the Windows environment when the uncooperative application quit. Importantly, Windows had already been ported to multiple different and incompatible machines at this early time. The BYTE article lists the IBM PC, HP 150, DEC Rainbow 100, Bytec Hyperion, Apple IIe, Eagle PC, Burroughs B20, Compaq Portable, Zenith Z-100, Wang Professional, TI Professional, Honeywell Microsystem 6/10, Computer Devices Dot, and the Columbia Data Products MPC Portable.
I’ve used both the name Window Manager and the name Windows. The name Windows was chosen around the time that BYTE ran their article. This name change was suggested by then head of marketing, Rowland Hanson. His belief was that the name Windows would be more appealing to customers. In hindsight, I believe he was correct. Microsoft WM or MSWM doesn’t have the same feel.
Development was slow and Windows missed its ship date several times. Between the start of 1983 and the end of 1984, Windows had had four project managers, and there were doubts internally about whether or not Windows would ever ship. To make matters worse, IBM had released TopView in August of 1984 for $149 (around $456 in 2023), and Tandy had released DeskMate in November of 1984. Microsoft needed to do something quickly before a competitor became too entrenched.
Tandy Trower had taught himself to program with BASIC on a TRS-80 Model 1. He then switched to software development from semiconductor engineering in 1979. First, his work was in education and curriculum products for schools, and then in April of 1980 he was hired by Atari in the personal computer division. His primary job early on was to evaluate third party titles for publication or purchase. His position shifted later in that year toward product management of educational and entertainment software. In 1981, Trower recommended to Atari’s executive management that they license Microsoft BASIC for the 400 and 800 computers. Atari’s BASIC was not fully compatible with those of the Apple II or the Commodore PET, and it lacked several different commands that competitors had. Bill Gates traveled to Sunnyvale to negotiate the deal himself.
The contact with Microsoft made an impression on Trower. In late 1981, he reached out asking about any open positions. Steve Ballmer offered him a position on the product management team for BASIC. His success in this position expanded his responsibilities to Pascal, COBOL, FORTRAN, C, Flight Simulator, and to the BASIC compiler and interpreter for Macintosh. In late 1984, Trower had a project review by Gates, and Gates wasn’t happy. Turbo Pascal by Borland was gaining traction, and Microsoft’s products weren’t seen as being very competitive. Borland’s products were cheaper, and Pascal was the language computer science students were being taught. As a result of the bad review, Trower was questioning his position in the company. He loved BASIC. It gave him his start, and he thought it was a great product, but he didn’t feel that he was the correct person for the job. So, he spoke with Ballmer.
A few weeks after speaking with Ballmer, he was offered a job managing Windows. Given that Windows was considered vaporware by the industry, doubted internally at Microsoft, and had killed the career of a few managers already, Trower wondered if this was just a ploy to get him fired over the Turbo Pascal debacle. He chose to confront Gates and Ballmer about this. They laughed and then assured him that the offer was real and that they had confidence that Trower could get the job done. In January of 1985, Trower took over the product management of Windows.
His first discovery in his new role was that the development architect and manager, Scott McGregor, had resigned. Ballmer took over the lead development role. Trower then set about figuring out what was complete and what was left to be done. The three core functional components of Windows (kernel, UI, GDI) were in place, but there was still quite a bit to be done and Ballmer was setting a ship date just six months out. For Trower, that wasn’t intimidating. He had a very good record of shipping. What the time crunch did mean was that the tiled windows were staying in place, the system font would remain fixed-width, and the resulting user interface would feel… unrefined. McGregor had come to Microsoft from Xerox PARC where tiling windows and fixed width system fonts were the trend. Shipping by Summer was the goal and therefore Xerox’s trends were Windows’ defaults.
How does one sell an OS that lacks applications, and how does one sell applications for an OS that lacks users, and how does one get users without the applications? This circular dependency was the next hurdle. Even within the walls of Microsoft, enticing developers to write software for Windows was tough. The applications team was focused on the Apple Macintosh where Lotus and WordPerfect were absent. Trower’s former team wouldn’t even build BASIC for Windows. The Windows team itself had already built a few applications: Notepad, Calculator, Reversi, MS-DOS Executive. Lotus and Borland had meanwhile released Metro and SideKick. These were small character-based application suites that could be brought up via keybindings while other applications were in use. These included a text editor, calculator, calendar, and rolodex. The Macintosh bundled a word processor and a drawing application. Trower felt that Windows would need these kinds of applications to achieve any success. They’d need to build the paint program, the word processor, the calendar, and the rolodex. They’d further need to refine the small applications that the Windows team had already built. He took these recommendations to Gates and Ballmer and got them to sign off on the idea. The word processor built some animosity with the applications team as it would compete directly with Microsoft Word, but Gates and Ballmer approved and Windows Write became a Windows mainstay. In doing work on Paint and Write, Microsoft worked with Aldus. Aldus was porting PageMaker to Windows (an early win for Microsoft), and Laser printers from HP were coming out. Both Microsoft and Aldus wished to support these printers, and they worked together toward that end. Ray Ozzie from Lotus who was working on Lotus Notes at the time, gave valuable input to Microsoft regarding the keyboard interface and shortcuts as well as what applications developers needed to transition to Windows.
In the early part of Summer in 1985, it became quite clear that Windows wouldn’t be shipping on time. Ballmer still wanted to release something for application vendors, industry analysts, the press, and other strategic partners. This was named “Premier Edition” and it allowed Microsoft to partially save face. Unfortunately, it wasn’t useful regarding testing as a bug was found in memory management that required a complete rewrite. Gabe Newell (yes, that Gabe Newell) was working late into the night and morning hours, sleeping in the office, and hammering out code to try to get things working and shipping. By November, a release was ready. This was version 1.01 as the Premier Edition was considered 1.0. The release was announced at COMDEX in Las Vegas on the 20th of November in 1985. Over the following months, Microsoft added internationalization and bug fixes which resulted in versions up to 1.04 on the 10th of June in 1987. The end of official support for Windows 1x didn’t come until the 31st of December in 2001.
All of the Windows 1x development was happening while Microsoft was focused on its partnership with IBM. IBM rejected Windows in favor of its own TopView. Later on, Microsoft and IBM were working on OS/2 together and IBM wanted TopView compatibility built in to OS/2. For this Microsoft would acquire Dynamical Research (who’d made a competitor to TopView called Mondrian), and also brought their people to Microsoft. Several of those folks would be crucial to Microsoft’s later success.
Two years after announcement and many people later, Windows made it to market. At this point there were many competitors on the market, and Windows didn’t get a ton of applications. As far as I know, there were: Balance of Power, PageMaker, Soft PC Paintbrush, WinSong Composer, Omnis Quartz, and several applications from Micrografx. Windows’ future was far from certain even within Microsoft. OS/2 was considered the future. Still, the start was promising. Windows had good technology, great people, and a strong company to help it along.
On a side note, I now have readers from many of the companies whose history I cover, and many of you were present for time periods I cover. All corrections to the record are welcome, just leave a comment.
Loved this ones!