The Open Software Foundation
and the Open Source Movement

James Loveluck

Last updated 1 Feb 2005

A Little History

In this note I present a model of software development which was adopted by the Open Software Foundation (OSF) and which shares certain characteristics with the Open Source movement, to the extent that OSF offerings, including source code and specifications, were available under equitable licensing terms.

The OSF was founded in 1988 by seven of the largest computer manufacturers of that period: IBM, Digital, Apollo, Hewlett Packard, Groupe Bull, Nixdorf and Siemens, following the “Unix wars” which opposed these companies to AT&T and Sun Microsystems, suspected of attempting to take control of the evolution of Unix. Later, other companies joined OSF (Fujitsu, Hitachi, and even AT&T and Sun), while others disappeared following mergers (Siemens-Nixdorf, HP-Apollo).

In contrast to other industrial organisations (such as X/Open, for example) the goals of OSF included not only the elaboration of common specifications, taking into account relevant industry standards, but also the development of reference implementations of the corresponding technologies. Both specifications and reference implementations were supposed to be developed according to an open process including the following stages: Request for Technology (RFT) requirements in a specific area were defined in a process which involved consultation of OSF members, after which any organisation could respond by submitting relevant technology; an open evaluation of the submissions was undertaken, involving external consultants to guarantee neutrality; after selection, rationale was published in order to guarantee the impartiality of the result. The subsequent stages involved integration and/or development of the necessary technologies, after which the specifications and software were made available under open and equitable conditions. In practice, this open process was restricted to OSF members - over 700 at the zenith of the organisation. Although the OSF was not formed with the goal of distributing “free software” it was nevertheless committed to this open process for the selection of technologies and definition of their specifications. In fact,  as I shall describe below, the OSF went much further than this in certain areas and did in fact distribute a significant quantity of software under conditions similar to open source ones.

Implicit in this model of operation was the idea of sharing of resources among the founding members of the OSF in order to develop common technologies and to guarantee their interoperability. The target technologies were restricted to those considered to be part of a common infrastructure, which would not inhibit competition between vendors in other areas.

In 1996 the OSF and X/Open merged to form The Open Group (TOG), and the initial goal of developing reference implementations of the open technologies did not survive much longer. Today TOG is focused on the definition of IT industry specifications and conformance tests for these specifications.

The OSF Technologies

In this section we describe the technologies developed by OSF and examine the extent to which they were successful. Without doubt the most well-known of these technologies is OSF/Motif, the multi-window system for Unix, based on X11. OSF/Motif, and its successor the Common Desktop Environment, was adopted by almost all the Unix vendors and is still part of the Unix offering of Sun, HP, IBM and others. In fact OSF/Motif has almost become “Open Source Software” as since May 2000 a version Open Motif  exists for which the source code is distributed without charge, the only condition being that it must be used exclusively with an open source operating system, such as Linux.

The other technologies developed by OSF include:

• OSF/1: a Unix-like operating system with POSIX-conformant interfaces and libraries
• DCE (Distributed Computing Environment): an infrastructure to support distributed applications including components such as a Remote Procedure Call (RPC), security services based on Kerberos, a distributed file system (DFS) and a directory service.
• ANDF (Application Neutral Distribution Format):  a compiler which generates a platform-independent intermediate language, thus leading to application portability.
• DME (Distributed Management Environment): an infrastructure and set for system and network management

These technologies did not enjoy the same success as OSF/Motif. For example, the complete OSF/1 system was adopted only by Digital, although several other vendors integrated components (such as the libraries) in their Unix offerings, and provided system and library interfaces conforming to the OSF/1 specifications (the Common Application Environment); some of the DCE components were fairly widely adopted, such as the RPC mechanism (even Microsoft implemented an RPC conforming to the DCE specification), but the complete DCE infrastructure was less widespread.

Although it was not among the initial objectives of the OSF, a significant number of software components of OSF technologies were eventually distributed under conditions similar to those for Open Source software. Software distributed via the “OSF Software Mall” included a restricted version of OSF DCE and a version of ANDF corresponding to the GNU C compiler (gcc). In certain cases the licence terms for these software components excluded commercial use, but this was not always the case.

The OSF Research Institute

The OSF Research Institute (OSF RI) was part of the initial vision of the founding members, and was seen as a shared facility for advanced development in key areas of information technology of interest to the OSF sponsors. There were two OSF RI sites, one at the OSF headquarters in Cambridge, Mass., right next to MIT, the other close to the Grenoble university campus.  Initial research was focussed on operating system development but research activities were later extended to cover other areas, including: Internet security (in particular, secure access to Web servers using POSIX-style fine-grained access control); other Web software, including the first WebMail application; and Java technology (ports of the JDK to a number of platforms and a Java compiler). The OSF RI was not subject to the same commercial constraints as the other departments of the OSF, and a substantial number of software components developed by the RI were distributed free of charge.

The Grenoble OSF RI site was heavily involved in many of the projects mentioned above, and in particular it led the Java compiler development as well as a number of aspects of the work on operating systems. The following paragraphs provide more detail on the latter subject, which was a good example of the Open Source approach.

Operating System development by the OSF RI was based on the Mach microkernel from Carnegie Mellon University (CMU). Versions of Mach for a number of platforms were distributed as open source software by CMU, and the OSF RI followed this tradition by distributing a version of Mach with a number of extensions (in particular for multiprocessor and massively parallel machines). It is interesting to note that the Free Software Foundation (FSF) also worked on an operating system (Hurd) based on the Mach microkernel, and in the early days of the OSF RI several members of the FSF team participated in the OSF RI workshops on microkernel technology. The goal of the FSF team was to develop an operating system “from scratch” but as far as I am aware a development of Hurd was never completed.

A version of OSF/1 based on the Mach microkernel was developed, but was not distributed under open source licence conditions because of the dependence on Unix and OSF/1 licences. However, at a later stage the Grenoble OSF RI developed MkLinux, a version of Linux based on Mach. MkLinux was adapted to a number of hardware platforms, and in particular the Apple PowerMac, with support from Apple, and over 100,000 copies of MkLinux for PowerMac were distributed in 1996, well before the existence of other ports of Linux to Apple hardware. To this day the source code for Darwin, the open source version of Mac OS X, includes a substantial amount of microkernel code adapted to the Apple hardware by the OSF RI. Versions of MLinux were also developed for the Intel x86 and PA-RISC hardware platforms, the latter with support from HP.

Conclusions

The OSF was not the first organisation to distribute “free software”, and at its inception this was not one of its goals. The Free Software Foundation was founded in 1985, and at the time the OSF was created it was already distributing software under the GNU licence, including GNU Emacs and the gcc compiler. Nevertheless, the OSF introduced the idea of open source software to the industrial world and pioneered the concept of an open process for the definition and evaluation of technology components, which were later integrated and/or developed in a partnership between the development teams of the OSF and those of the industrial sponsors.

The OSF technologies did not enjoy the success expected at its inception and among the reasons for this are:

• The technology evaluation process was quite lengthy and that of technology integration even longer, and this was out of line with the timescales required by the industry
• Even after the technology components had been integrated by OSF, the vendors had to invest almost the same amount of time and effort to integrate the offerings with their own environment and to assist customers in migration to the new technology
• Although the development of OSF technologies was financed in large part by the sponsors, they did not always adopt them and thus did not benefit from their investment
• The OSF business model was not very favourable to the sponsors: in addition to the costs described above they also had to pay licence fees for the technologies which had been developed thanks to the funding they provided!

These problems in the OSF process led to results which did not meet the initial expectations of the founding members, who envisaged that sharing of resources would lead to economies, and that the common development process would lead to compatible, interoperable technologies.

In spite of these difficulties, the early years of the OSF were an exciting period to live through. There was a very real feeling that a new way of building computer systems was being created, embodying a common process for technology development which could facilitate interoperability, resulting in significant advantages to end users. It was not yet the open source development model, with its sense of a community of developers contributing to freely available source code, but the OSF did manage to carry the computer industry a significant distance in this direction, while maintaining the support of its industrial sponsors.

References

Although the OSF Web pages disappeared long ago, many of them can still be found using the “WayBack Machine” of the Internet Archive project:

http://web.archive.org/web/*/http://www.osf.org/

and for the OSF RI :

http://web.archive.org/web/*/http://www.osf.org/RI/

in particular, the Grenoble site:

http://web.archive.org/web/*/http://www.gr.osf.org/

The archives do not extend further back than 1997, after the merger of OSF and X/Open, but the refocusing of activities by TOG was far from complete at this moment, and the Web pages indicated are essentially those of the OSF.

Philippe Bernadat also pointed out to me the following Web page with a brief summary of OSF, X/Open and TOG:

http://kb.indiana.edu/data/agoq.html?cust=765842.75392.30

 Latest News