Open Source OS/2: The Impossible Dream

UPDATED: For nearly three years, a group of OS/2 supporters has been asking IBM to offer up the source to the long-abandoned operating system. For three years, IBM has said no.

Is it finally time to relegate the petition to the same retirement patch as OS/2 itself?

The latest rejection from IBM’s legal department came just two weeks ago. The letter from Yvonne Perkins in IBM’s Enterprise Platform Software group read, in part, “We have considered the positioning of OS/2 and open source several times in the past, and for a variety of business, technical, and legal reasons we have decided to not pursue any OS/2 open source projects.”

As a project abandoned in 1997, OS/2 doesn’t seem to have much value left in it, especially when comparing its kernel with Linux and Mac OS X today. But some of the underlying object-oriented (OO) technologies in its desktop could have some value, since other than OS X, there really isn’t an OO desktop on the market.

With some of the old Workplace Shell technologies, it would be possible to build componentized apps like Lego bricks that inherited capabilities from other apps and shared services. Linux doesn’t have this with KDE and GNU Network Object Model Environment (GNOME) , its two competing interfaces.

OS/2 has been off the retail market since 1997 and off the market entirely since 2005. Serenity Station handles some sales and support, but IBM itself is trying to migrate customers off the platform.

There have been a few snide reactions, but the bulk of open source advocates have an inkling that IBM is simply unable to offer up OS/2 source code.

The situation with OS/2 is actually worse than most realized. If the state of OS/2 is any indication, then don’t hold your breath for any other older, multipartner projects to ever see the light of day.

David Moskowitz, president of the consultancy Productivity Solutions and a contributor of a few lines of code to OS/2 in his own right, said the Free Software Foundation is not new, but the concept of open source is relatively recent.

“From 1987 to 1997, the concept of open source as we know it didn’t exist,” he said. “The code simply wasn’t maintained for this kind of release.”

Moskowitz and others said there’s simply no way OS/2 could be released because it was not maintained in a manner compatible with open source development.

“Community people seem to think it’s a matter of Jonathan Schwartz saying ‘it’s open source’ and magically the next day there’s a full copy of the code in a Subversion project ready for download,” said Simon Phipps, chief open source office for Sun Microsystems and a former IBM staffer. “It’s not like that.”

Specifically, a company’s internal source code is not kept in a state for public consumption. It has to be put into a distributable package, and all of the headers need to be changed — which means testing to make sure all the files still work together. And that’s assuming the code passes all legal checks.

Sun started discussing open sourcing Java all the way back in 2001, according to Phipps. From 2001 to 2003, it did some investigative work to find out how hard it was going to be.

“Once they decided to go for it, it took us a year to get the code in a state where it could be put under an open source license,” Phipps said.

And that was Java, a programming language — not an entire OS. Java is a project under current development at Sun, which created more than 95 percent of the source in house, and the project’s leader, James Gosling, is still with the company.

In the case of OS/2, development ceased more than a decade ago, and a large amount of the code belongs to Microsoft, which is not known for releasing its code.

Even worse, it had been developed in IBM’s Boca Raton facility, which closed in 1996. Today, no one knows where all the code actually is: The staff and everything held at Boca has been scattered to the wind, according to Moskowitz.

Assuming IBM even had all the code in one place, it would have to go through all of it, line by line, and find out who wrote what — an IBM staffer, a Microsoft programmer or a third party.

“Without having all the code, all the contracts and potentially all the access to people … it’s extraordinarily difficult for IBM to determine what is absolutely releaseable unfettered,” Moskowitz said.

The code is such a mix of sources that he said he had no idea if something even remotely buildable could be cobbled together.

Moskowitz gave an example of how daunting a prospect assessing OS/2 could be: He had once been charged with clearing an application for release. Going through 100,000 lines of code cost $250,000 and took a team of eight nearly four months to track down all of the contracts and people involved.

Too often, the team would identify a staffer believed to have written a single line of code — only to discover that they merely rewrote the code, which had initially been authored by another staffer. Then, the task became chasing down the original coder. In some instances, four or five people ultimately may have been responsible for just one line of code.

And that was a simple app. OS/2, until version 3, was a complex, joint Microsoft/IBM development. Originally known as OS/2 NT, it was going to be Microsoft’s high-end operating system until the two companies had their highly publicized split in 1990.

At that time, David Cutler, a programmer extraordinaire, defected from DEC to Microsoft and wrote what would be the kernel of Windows NT 3.1.

Prior to that, both companies had programmers working at each other’s facilities. Unless every line of code in OS/2 is signed, people vetting the code today will have no way of knowing who wrote a particular bit of code, and when.

Sun has also released the Solaris operating system as an open source project, but was only able to do so because it bought an outright perpetual license in 1994 from Novell, which owned the Unix System V source on which Solaris was based.* In the case of Solaris, Phipps said Sun needed to do four years of due diligence to prepare it for release.

Phipps said such situations often prove just too expensive, because the effort requires nearly as many lawyers as programmers. And without a clear business case, the undertaking may simply prove unjustifiable.

“The investment in patent searches, code scrubbing and due diligence is huge,” Phipps said. “It involves a great deal of engineering and a great deal of legal work.”

“I can see IBM looking at that and deciding they don’t have any money to waste on OS/2, which they won’t make any money from anyway,” he added.

*Corrected to clarify that Novell owned Unix System V source, from which Solaris was derived.

News Around the Web