For BSD Unix, It's Sayonara

Jason Levitt and Evan Schuman
Open Systems Today

June 22, 1992

Berkeley, Calif. -- The end of a major computing era will come next month when the final release of BSD Unix gets shipped, marking the University of California at Berkeley's withdrawal from the operating system business.

A lack of funding and the increasing sophistication of commercial Unix implementations have combined to halt Unix development at the university, meaning that 4.4BSD Unix-set to go Alpha by next month, with general availability by December-will be Berkeley's last OS.

BSD has been popular in the academic community for many reasons, mostly because-since 1988-it has not required a license fee and often features cutting-edge technology years before commercial computer vendors' versions.

"With any commercial version of Unix, you usually have to sign your life away," said Evi Nemeth, an associate professor of computer science at the University of Colorado at Boulder, adding that this no-license feature made BSD ideal for teaching students about operating systems and networking.

BSD Unix was responsible for much of the Unix innovation during the 1980s and came into being around 1978 when Berkeley graduate student Bill Joy added virtual memory capability to a VAX 11/780 port of Unix. The resulting system was named 3BSD (4BSD in 1980) and became the dominant Unix version at universities.

Since that time, nearly all of the important Unix enhancements have come from the BSD releases of Unix. Collectively, these enhancements are often referred to as the Berkeley "extensions" and are now part of System V Release 4. The programming API for OSF/1 is based on 4.3BSD libraries and system calls.

Originally, BSD was funded by the government through DARPA. But that funding was cut off in 1988 when the vendor community started to show a strong interest in doing its own development.

Since 1988, BSD research and staffing has been supplied through vendor grants, mostly from Hewlett-Packard, the Open Software Foundation and Cray Research.

Kirk McKusick, the Berkeley research computer scientist who is in charge of the BSD project, added that NASA Ames also has been a major sponsor.

"It's becoming increasingly difficult to get funding from the private sector," he said. "With the recession, these companies are obviously looking at the bottom line and external research is an item that can be easily reduced."

While fund raising has gotten more difficult, today's technology has forced the operating system to grow substantially more complicated, with the 100 different commands from a few years ago now at 500 commands, McKusick said.

"The complexity has grown almost exponentially," he said, adding that his staff of four would need be doubled to "continue the level of quality."

Also, McKusick said, the university itself "is not as interested in having us around as they did 10 years ago" because vendors today-such as Cisco Systems-have packaged some of the networking technology that made his team so indispensable before for systems administration.

"We are so good at getting our stuff disseminated that we are no longer needed," he said.

Another factor contributing to the OS' demise, said Keith Bostic, fellow Berkeley programmer and the project's second-in-command, was the increasing sophistication of vendors in working with Unix.

In the early 1980s, Bostic said, just about the entire Unix community was either using a form of BSD or AT&T's System V. "If you bought {Unix} from DEC, you were running BSD. If you bought {Unix} from Sun, you were running BSD," he said. "In today's climate, it is not the same as 1980."

Bostic said that today's Unix vendors are more sophisticated and are making many more changes to whatever version of Unix they are starting with.

BSD 4.3 "was the last version that major vendors shipped right out of the box. Vendors now tend to pick and choose," he said. "The vendors are basically going to lose a research group."

There are still options to secure BSD code, with one company, Berkeley Software Design (Falls Church, Va.), a company employing former Berkeley programmer Mike Karels, planning to offer a commercial version of Unix for SPARC systems based on the 4.4BSD code but free of AT&T source licensing requirements. It currently offers BSD/386, a version of Unix for 386 machines based on the Berkeley NET2 release.

And there are several outfits that will distribute the code freely, often directly on the net.

The University of Colorado's Nemeth, regarded as an expert on Unix security and perhaps best known for having co-authored the Unix System Administration Handbook, said last week that she would like to have her university conduct some of the operating system research that Berkeley is now abandoning.

One of the most popular accomplishments of the BSD team, said Nemeth and others, was its sifting through mountains of software out in the industry, finding the best examples and incorporating them into a BSD version. McKusick said it was often a matter of convincing contributors "that they would rather have fame than fortune."

Nemeth cited a more colorful-and oft-cited-description: "It was a matter of their taking it in and peeing on it until it smelled like Berkeley."

"This is a valuable function and the community needs the service," she said, adding that she is trying to convince DARPA to resume funding BSD research at her university with it paying for three to four full-time staff, two or three graduate students and two or three undergraduate students. Nemeth estimated that she would need about $700,000.

But she said last week that she now is leaning toward securing interim funding from industry, in an attempt to demonstrate to DARPA that her people can do the job effectively.

Nemeth said that she is hesitant, though, to have vendor support made permanent. "I don't see having to constantly hustle for money."


Copyright 1992 CMP Publications, Inc. All rights reserved.