1000BiT - Gary Kildall

The Gary Kildall Legacy

by Sol Libes

Gary Kildall died in July 1994 at the age of 52. The computer media, with a few small exceptions, ignored his passing. The Circumstances of his death are pretty murky. One report attributed it to a fall from a ladder, another an incident at a bar, and another to a heart attack.

Every PC owner owes Gary a debt of gratitude. Bill Gates and Microsoft owe him more than anyone else. Gary was the first person to interface a disk system to a microcomputer and create an operating system for it. He changed what had previously been a circuit designed for process control applications into a fully functional computer. Microcomputers now did tasks previously done only on minicomputers and mainframes. The world changed dramatically because of his work.

Gary received a Ph.D. in Computer Science from the University of Washington in 1972 and joined the Navy. It is interesting to note that both Gary and Bill Gates were born and raised in the Seattle area. Like Gates, Gary also had a passion for computers. However, unlike Gates he completed his college education. Their paths crossed early on when Gates, a high school student, and Gary, a college student, both worked on the same DEC PDP-10 computer system.

The Navy appointed Gary to be a Computer Science instructor at their Post-Graduate School in Monterey, California. At the school, Gary purchased an Intel 4004 microprocessor chip set for himself and his students to experiment with. The 4004 was Intel's first microprocessor and the first in the world. It was programmable, handled 4-bit words and contained all of 2,250 transistors. Intel, at the time was primarily in the memory IC business, and created the 4004 as a custom project for one customer. When the customer wanted to renegotiate the price Intel asked that they give up their exclusive rights to the device. Intel introduced it in November 1971. Much to Intel's surprise the device was an instant success. Engineers began designing it into a wide variety of applications such as scales, traffic light controls, musical instruments, washing machines, printers, and more Intel soon realized that 4004 system designers needed software development support. Gary was hired as a consultant to create a programming language for the device. Gary created PL/M (Programming Language /Microprocessor) to run on an IBM 360 computer and generate executable binary code that was then burned into the ROM memory of the 4004 system.

Marcian "Ted" Hoff designer of the 4004 (Who also worked later in his career for Atari Corp), quickly followed with the 8008, the first 8-bit microprocessor. It was introduced in March 1972. Gary was again hired to develop PL/M for the device. Intel also designed an 8008-based small computer, called the Intellec-8. About the same size as the IBM PC, it was used for hardware and software development. Gary set one up in a classroom at school for his students. To make it easier to use, Gary developed a simulator/debugger program for the unit. Intel began to see that microprocessors helped sell more memory chips and developed a much more powerful 8-bit microprocessor, the 8080. Gary was again hired to create the development software. He was given an Intellec-80 to use at school.

In 1973 Shugart gave Intel a sample 8" floppy disk. Gary was immediately intrigued by the device and with a friend, John Torode, built a controller interface to an Intellec-80. Gary, and his students, wrote a small control program, which he called CP/M (Control Program/Microcomputer). It enabled him to read and write files to and from the disk. Gary copied the commands and file-naming conventions from the DEC PDP-10 VMS operating system. Gordon Eubanks, one of Gary's students, created a BASIC interpreter for the system. Early versions of CP/M and the BASIC interpreter were in the public domain since it had been created at a publicly funded institution. Copies found their way to some other government contractors and agencies.

In 1976, after his discharge from the Navy, Gary became a full- time consultant, using the name Intergalactic Digital Research. Together with Torode he designed floppy disk systems for several microcomputer manufacturers. At the time, MITS and IMSAI, the two leading 8080 microcomputer system kit makers, announced floppy disk systems. MITS offered a version of BASIC (written by Bill Gates and Paul Allen) that could load and save BASIC programs on disk. MITS contracted with another software developer for a Disk Operating System. When shipped in early 1977, it proved unreliable and had poor performance. MITS also refused to license the DOS to other system makers.

IMSAI, needing to compete with MITS, approached Gary for a non-exclusive CP/M license for a fixed $25,000 fee. Since several other manufacturers also wanted CP/M, Gary rewrote it completely to make it easier to install on different hardware systems. He made it hardware-independent by creating a separate module, which he called the BIOS (Basic Input/Output System). He also added an editor, assembler, debugger, and several utilities. CP/M became a full-blown computer development system Gary, and his former wife, Dorothy McEwen, formed Galactic Digital Research Inc. (Later it was changed to just Digital Research, Inc) to market CP/M-80. They placed a small classified ad in Dr. Dobb's Journal and were surprised by the large number of orders from hobbyists for the $90 software package. By early 1977, several manufacturers were including CP/M with their systems. They provided a ROM-BIOS so that CP/M loaded immediately on power-up.

By 1978 Microsoft BASIC and FORTRAN, UCSD Pascal, MicroPro's WordStar, Ashton-Tate's dBase, and other programs were running on CP/M-based on machines from Apple, Radio Shack, Commodore, Zenith, Sharp, and almost a hundred other manufacturers. In 1980, IBM approached DRI, to license CP/M-86, an 8086 version of CP/M then being developed. Gary had been working on this but delayed finishing it while working on several language projects. Intel had introduced the 8086 16-bit microprocessor in June 1978 and followed it a year later with the 8088, a lower-cost and slower version. IBM decided to use the 8088 for its new PC.

Seattle Computer Products in early 1979 introduced the first 8086 computer kit. Sales languished while SCP waited for DRI to introduce CP/M-86. In desperation SCP hired Tim Paterson to develop a DOS for them. Tim quickly created a simplified 8086 version of CP/M, which he called QDOS (Quick and Dirty Operating System, since it did not implement all of CP/M's features). Microsoft, located nearby, modified BASIC for the system.

IBM met with a cool reception when they approached DRI for a CP/M license. Dorothy McEwen and DRI's attorney refused to sign the IBM non-disclosure agreement (Gary did not attend the meeting), refused to make any modifications to CP/M-86 and insisted on a higher royalty than what IBM proposed. Bill Gates, who had been negotiating a BASIC license with IBM, seized the opportunity and offered to provide a DOS/BASIC package to IBM on favorable terms. Gates licensed SCP-DOS (for $50,000) and hired Tim Paterson to modify it to run on the IBM-PC. Microsoft submitted a copy to IBM for testing, who found over 300 bugs. IBM cleaned up many of the bugs, made a number of improvements and wrote the user manual. DRI released CP/M-86 soon after IBM released DOS Version 1.0. DOS had fewer features and poorer performance. IBM offered both CP/M-86 and DOS. CP/M-86 was offered at $240 versus $60 for DOS. Few PC owners were willing to pay the extra money DRI considered suing Microsoft for copying all the CP/M system calls, program structure, and user interface. However, DRI knew it would also have to sue IBM. It knew it did not have the resources for this and that its chances of success were remote. In 1982, IBM asked Microsoft to develop a hard disk version of DOS. Microsoft used the opportunity to completely rewrite

DOS so that version 2.0 was very different from version 1.0 and DRI's opportunity to sue was gone. DRI continued to improve CP/M-86 introducing multi-tasking and multi-user versions. However, they were not completely compatible with DOS and largely ignored by the marketplace. In

1989 DRI introduced a DOS compatible version (DR-DOS) which was recognized as superior to DOS. However, Microsoft marketing tactics (disclosed in the Justice Department investigation) shut DRI out of the market. Microsoft responded with versions 5.0 and 6.0 incorporating many of DR-DOS's features.

Kildall also pioneered in the development of a GUI (Graphical User Interface) for the PC. Called GEM (Graphical Environment Manager), it was demonstrated at the November 1983 COMDEX and shipped in the spring of 1984. Atari also licensed it for use with their new 520ST computers. GEM presented the user with a screen virtually identical to that of the Macintosh. Apple threatened to sue DRI. DRI responded by making some cosmetic changes to GEM. DRI did not recognize the potential of a GUI interface and did not put any marketing effort behind it. DRI eventually withdrew GEM from the retail market. It continued to market GEM to software developers as a front end for their graphics products. The most well-known product to use the GEM GUI was "Ventura Publisher" from XEROX.

Microsoft finally demonstrated their Windows GUI at the Spring 1985 Comdex, shipping version 1.0 in the fall. Microsoft learned from DRI's experience with Apple and made Windows appear slightly different from the Mac GUI. Version 1.0 proved an embarrassment to Microsoft. It was incredibly slow, unreliable, and lacked the smooth performance of GEM and the Mac. Version 2.0 of Windows did likewise. Windows was completely rewritten for version 3.0 and released in the spring of 1990, with the most expensive software promotional campaign the industry had ever seen coupled with aggressive marketing (initial price was $39 and thousands of copies were given away free). Gates did what neither IBM, DRI, Apple, Xerox, or the other GUI developers were willing to do. Namely, to make a total commitment, risking the entire company on the success of a GUI.

Microsoft sought to gain the largest market share by distributing Windows primarily through OEM channels. System manufacturers were persuaded to offer DOS and Windows preloaded onto hard disks by offering a low OEM price of $35 on average while offering Windows to retailers at $75. Microsoft actually made more money on the OEM version because the manufacturer assumed the cost of printing manuals, providing disk backups, the packaging, and support. Version 3.0 also proved unreliable. Microsoft fixed the bugs, added a few minor features and introduced it as version 3.1. Gates turned a major problem into a marketing success. 3.0 owners paying a second time, in effect paid for the repair of design defects.

Gary was also the first person to work on the development of software for driving CD-ROM interfaces. We will probably never know all of the system software work that he has created.

There is no doubt that Gary Kildall led the way in microcomputer software development. I wonder what Microsoft will do now that they no longer have Gary Kildall to lead the way for them?

Many ACGNJers met and spent time with Gary at the 1979 Trenton Computer Festival. I met with him many times, as a magazine author and editor, and President of ACGNJ. I developed great admiration for his talents, his hard work, and willingness to help others. I will also miss him as a friend.