OpenBSD

Decision Points

OpenBSD is a Unix-like computer operating system descended from Berkeley Software Distribution (BSD), a Unix derivative developed at the University of California, Berkeley. It was forked from NetBSD by project leader Theo de Raadt in late 1995. The project is widely known for the developers' insistence on open source code and quality documentation, uncompromising position on software licensing, and focus on security and code correctness. The project is coordinated from de Raadt's home in Calgary, Alberta, Canada. Its logo and mascot is a pufferfish named Puffy.

OpenBSD includes a number of security features absent or optional in other operating systems, and has a tradition in which developers audit the source code for software bugs and security problems. The project maintains strict policies on licensing and prefers the open-source BSD licence and its variants—in the past this has led to a comprehensive licence audit and moves to remove or replace code under licences found less acceptable.

As with most other BSD-based operating systems, the OpenBSD kernel and userland programs, such as the shell and common tools like cat and ps, are developed together in one source code repository. Third-party software is available as binary packages or may be built from source using the ports tree.

The OpenBSD project maintains ports for 17 different hardware platforms, including the DEC Alpha, Intel i386, Hewlett-Packard PA-RISC, AMD AMD64 and Motorola 68000 processors, Apple's PowerPC machines, Sun SPARC and SPARC64-based computers, the VAX and the Sharp Zaurus.

When OpenBSD was created, Theo de Raadt decided that the source should be easily available for anyone to read at any time, so, with the assistance of Chuck Cranor, he set up a public, anonymous CVS server. This was the first of its kind in the software development world: at the time, the tradition was for only a small team of developers to have access to a project's source repository. This practice "runs counter to the open source philosophy" and is inconvenient to contributors. de Raadt's decision allowed "users to take a more active role", and signaled the project's belief in open and public access to source code.

A revealing incident regarding open documentation occurred in March 2005, when de Raadt posted a message to the openbsd-misc mailing list. He announced that after four months of discussion, Adaptec had not disclosed the documentation required to improve the OpenBSD drivers for its AAC RAID controllers. As in similar circumstances in the past, he encouraged the OpenBSD community to become involved and express their opinion to Adaptec. Shortly after this, FreeBSD committer, former Adaptec employee and author of the FreeBSD AAC RAID support Scott Long, castigated de Raadt on the OSNews website for not contacting him directly regarding the issues with Adaptec. This caused the discussion to spill over onto the freebsd-questions mailing list, where the OpenBSD project leader countered by claiming that he had received no previous offer of help from Scott Long nor been referred to him by Adaptec. The debate was amplified by disagreements between members of the two camps regarding the use of binary blob drivers and non-disclosure agreements (NDAs): OpenBSD developers do not permit the inclusion of closed source binary drivers in the source tree and are reluctant to sign NDAs. However, the FreeBSD project has a different policy and much of the Adaptec RAID management code Scott Long proposed as assistance for OpenBSD was closed source or written under an NDA. As no documentation was forthcoming before the deadline for the release of OpenBSD 3.7, support for Adaptec AAC RAID controllers was removed from the standard OpenBSD kernel.

The OpenBSD policy on openness extends to hardware documentation: in the slides for a December 2006 presentation, de Raadt explained that without it "developers often make mistakes writing drivers", and pointed out that "the [oh my god, I got it to work] rush is harder to achieve, and some developers just give up" He went on to say that vendor binary drivers are unacceptable to OpenBSD, that they have "no trust of vendor binaries running in our kernel" and that there is "no way to fix [them] ... when they break".