Paper - 1999 USENIX Annual Technical Conference, June 6-11, 1999, Monterey, California, USA
1999 FREENIX Track Presentation
Open Software in a Commercial Operating System
History and Topical Focus
Apple Computer has been working for the past couple of years to build a new and modern operating system to advance the Macintosh platform. The Macintosh platform, developed at Apple, while strong in many areas (in particular human interface), lacks the solid and modern foundation best exemplified in current variants of the Unix operating system. Apple acquired NeXT Software, Inc. in early 1997 largely in order to gain NeXT's experience in making an operating system based on a modern foundation which also had made advances in user interfaces, programming models, enterprise software, and other areas in which Apple was not traditionally strong. Thus began the development of Mac OS X Server, Apple's server operating system offering, which is also the basis for parts of Mac OS X, which we will be rolling out later in 1999.
This session will discuss this new foundation: the "Core Operating System", which is largely based on software which was developed at the University of California at Berkeley and Carnegie-Mellon University, and was made freely available with open licensing terms. In addition much of the Core OS utilizes software available from the present-day BSD efforts, the Free Software Foundation, M.I.T., the Apache Group, and countless other groups and individuals who create and publish software under open licenses.
The Value of Open Software
Creating an operating system is an enormous task. The quantity of code required to create a functional and useful system is quite large, and the nature of operating systems is that the quality of that code must be significantly higher than that of most software. Everything depends on the OS: Too little functionality makes life difficult for the developers who wish to support your system. The smallest of bugs can disable most or all of the software than runs on the system--or worse, can cause crashes or data loss. It is therefore quite difficult indeed to create a whole operating system from scratch, and is a task not often undertaken.
It is fortunate, then, that there is a large body of software freely available which we have been able to use as a starting point. Even better, much open software is good stuff; there have been many years of development behind many of the efforts from which we have been able to draw our code.
The Mac OS X Server Core OS is a 4.4 BSD system, based on the 4.4 Lite2 release from Berkeley and updates with code from FreeBSD, NetBSD, and OpenBSD. BSD has seen many years of active development; it is well designed and has provided the basis for many established standards. It has proven to be a tried-and-true system and remains in the forefront as a platform in which to test and integrate emerging technology and ideas. It has become well accepted that free operating systems are not only leading-edge experimental systems, but highly robust and scalable as well.
A non-trivial advantage of open software is the excellent developer support behind many open software projects. Apple has considerable experience with operating systems (we've shipped Mac OS, A/UX, AIX, and Linux systems, and worked on several others), but the engineering effort required to simply maintain and update in-house software is considerable and expensive.
At Apple, I write and maintain software, manage the source code control for all of Core OS (using open software-CVS), go to lots of meetings, and oversee internal releases. In addition, I personally maintain over 60 open software projects in the Core OS build cycle, several of which are quite large. While these 60 projects do take some effort to manage (porting, updating, sending diffs upstream), were they not also worked on by many other developers (many of who I'll never meet), it would not be possible for me to provide nearly so much functionality to the system myself.
Clearly, then, open software offers a great value to businesses that have a use for it in their products.
Where Apple Adds Value
Perhaps Apple could take NetBSD, clean up the Power Macintosh support, stick an Apple logo somewhere, and call this great thing "Mac OS X Server". But then, if you are looking for a good Unix server, you don't need Apple. You can instead quite readily install a NetBSD system on your Macintosh, or (gasp!) a PC, and carry on. We know, in fact, that this is satisfactory for many users. On the other hand, we also know that many of our users would not consider such a thing.
Apple's interest in BSD, therefore, cannot be as a replacement for Mac OS, but as a foundation for the rest of the system we provide: a polished, well-designed user experience. Apple can provide to its end-users a great deal more than current Unix systems offer. Our users expect a system that is easy to understand and use; one that is friendly; one with some style.
This goes well beyond user interfaces; it includes simple to set up hardware which "just works" with the included software, application tool kits for developers which allows them to provide a consistent user experience and give them the ability to leverage all of the services of the underlying system. It also includes five yummy colors.
Macintosh is about personality; it's about the people who use it, and the experience of using it. Combined with the advanced power and robustness of a new underlying foundation, that experience can be... uninterrupted, and all the more pleasant.
The Impact of Various Licenses
There are many sources of open software and therefore many different licenses attached to open software. One of the compelling reasons behind using BSD as the Core OS is the simple BSD license. While an attached credit requirement takes a bit of work to honor, the license allows for any use of the software so long as the copyright is preserved. Companies looking for source with (almost) no strings attached would do well to use code under a BSD-style license.
A lot of companies would not bother to send their changes upstream, for fear of giving up some valuable information or technology, to send any useful code changes up to the upstream maintainers. Occasionally, these concerns have merit. More often, companies simply want a competitive edge, and aren't willing to share the wealth. One has to wonder whether this is good logic. The fact is that more diffs translates to more work. Eventually, the upstream provider will release a new version, perhaps with some important bug fixes and useful new features. If you have put significant work into improving the software, and kept it to yourself, you will have to merge these changes into the upstream source each time you want to update your code. This will cost you some developer time, perhaps enough that updating is not worthwhile, and your competitive edge can erode rather quickly. On the other hand, if you hadn't put much work in at all, there likely isn't much of an edge to lose.
I have made this argument successfully at Apple, and the result is that Apple started a very beneficial technology relationship with NetBSD, in which we can easily update our code periodically, and NetBSD gets an occasional update from Apple as well. Mac OS X Server shipped with a developer CD including a fair amount of source code from NetBSD, (as well as the required source code for GNU software) so that our developers could help out as well with any problems they experience related to that code.
Another popular open software license is the GNU General Public License, which allows mostly unrestricted use of software so long as source code is available under the terms of the GPL for any changes made to the software. This requires a greater leap of faith for companies, and is often enough to discourage companies from using such software in commercial products. However, once having experienced the benefits of open software, the GPL is, in principle, an acceptable license. Unfortunately, the GPL's potential tendency to "infect" additional code keeps many companies (including Apple) wary of using GPL'ed software.
Being a Contributor to the Community
As explained above, there is little reason for Apple to take open source software from BSD and treat it as proprietary software. Since the BSD subsystem is not the primary selling point of our complete system, nor a particularly unique subsystem given its free availability, Apple can only benefit by keeping our code in sync with the upstream source and thereby helping to improve the technology on which we depend. By contributing bug fixes and incremental improvements back to the open source community, we all win.
Apple's positive experience with open software in the recent past, and a belief on the part of Apple's executive team that the benefits of open software are indeed substantial, we have recently taken steps to contribute more substantial technology to the community by forming the Darwin project. It is our hope that by allowing our users and developers to work with us in improving our software, that Apple and its customers will see a great deal of benefit, and build a better relationship with each other. In addition, we hope that our efforts put us on the right track toward a healthy relationship with the greater open source community.
About the Presenter
Wilfredo Sánchez is a 1995 graduate of the Massachusetts Institute of Technology, after which he co-founded an Internet publishing company, Agora Technology Group, in Cambridge, Massachusetts; he then worked on enabling electronic commerce and dynamic applications via the world wide web at Disney Online in North Hollywood, California. Fred is presently a Senior Software Engineer at Apple Computer in Cupertino, California. He works primarily on the BSD subsystem in Mac OS X Server as a member of the Core Operating Systems group, and as technical lead for the Darwin project. Fred is also a member of the NetBSD Foundation, Inc., and a contributor to the Apache HTTP Server Project.
This paper was originally published in the
Proceedings of the 1999 USENIX Annual Technical Conference, June 6-11, 1999, Monterey, California, USA
Last changed: 1 Mar 2002 ml