EuroBSDcon 2011 conferences

Hans van der Looy did the opening keynote, discussing the sad sad state of the Internet PKI setup, from the not-so-optimal architecture (what exactly is a certificate supposed to guarantee?) to the mindset of the CAs (and their behaviour in the event of a breach, as was seen in the Comodo or the Diginotar cases). He also mentionned possible alternatives to the current scheme.

Martin Husemann started the show with the NetBSD automated testing framework. The NetBSD project set up a testing suite with the desired properties of being automatic, easy to run, able to flag quickly new issues, without being intrusive. Martin explores the components of the testing framework, and then explains the factors that made this testing experiment a success.

Bjoern Zeeb reported on the state of IPv6 in FreeBSD. He covered both was is already available and what is planned for 9.0 and beyond. He also talked about the advocacy done by FreeBSD during the World IPv6 Day and the motivations behind all this hard work. It is now possible to have a FreeBSD system that speaks only ip6; this has been a good exercise for testing the quality of some network software.

Zoltan Arnold Nagy's presentation of NPF: a new packet filter was a big success of the week-end. He described the reasons behind the birth of NPF, discussed the design of NPF and his own contributions, and finished the session with a very nice exchange with the audience. NPF will be the default firewall of NetBSD 6.0; it went from scratch to functional in about a year.

Paul Irofti described OpenBSD's New Suspend and Resume Framework. The most challenging aspects of the work around powersaving in OpenBSD were the complexity of ACPI (especially in comparison with the APM framework), and the buggy, sometimes plain stupid ACPI code in vendors' products. Nevertheless, the OpenBSD managed to produce a reliable ACPI implementation, that uses neither Intel's (which is used by Linux) nor Microsoft's (obviously). Note that the document linked above is from Paul; other OpenBSD-related talks from EuroBSDcon 2011 and other conferences can be found in the OpenBSD presentations page.

Jean-Yves Migeon did a very technical roundup of the state of Xen under the *BSD systems; in particular the NetBSD support in both dom0 and domU roles. He covered quite a lot of ground: the architecture of Xen, the hard-to-follow pace of development, the sysadmin's view… The talk drained so many attendees that, like Zoltan, Jean-Yves had to go to the main room in order to welcome everybody. Unfortunately, the technical nature of the talk did not lend itself to note-taking.

Marshall Kirk McKusick closed the day with an History of BSD. The subject is amply covered on the Internet, but Marshall's tone and passion made a very nice session for closing the day on a less technical note. « Atomic rename(2). I rewrote that for FreeBSD a couple of months ago. That's only the tenth time or so! »

Herbert Bos discussed the architecture of the Minix 3 OS, which is currently under heavily development. Minix uses the mini-kernel approach, with many servers running in non-privileged mode. This allows for increased reliability, as large parts of the OS can be brought back to a functionning state after a crash.

Claudio Jeker presented the MPLS framework in OpenBSD. His talk started with a presentation of the protocol itself, for the members of the audience who were not familiar with it. He then mentionned the implementation itself; OpenBSD is proud to be one of the first (the first?) OS to have out-of-the-box MPLS support. Claudio also did a live demo between three hosts (two end nodes playing the role of PE routers, and one P router). Claudio insisted that OpenBSD's MPLS implementation was not on par with hardware-based ones (think Cisco or Juniper gear); however, OpenBSD could have a nice role to play in an ISP setup as a PE router for not too bandwidth-hungry customers.

Michael Dexter surveyed the virtualization strategies in BSD, from chroot(8) to BHyVe. He approched the problem from an interesting point: virtualisation systems reject the traditional singletons of the Unix model. For instance, chroot(8) challenges the unicity of /; Xen disproves the unicity of the kernel; Apache's virtual host support recuses the unicity of an ip address on a host; FreeBSD's jail(8) rejects the unicity of the root user… Michael also presented the goals and the current state of BHyVe, a new type 2 hyvervisor for FreeBSD. In particular, the project needs more testers.

Eric Allman looked back on 30 years of sendmail, the history of the project, what allowed such a longevity (with BSD being one of the only codebase that has been used for longer), what was done right since the beginning and what the stupid decisions were.

Pawel Jakub Dawidek presented Highly Available Storage for FreeBSD, a new feature of FreeBSD already available. He described the working of HAST, but mostly from the point of vue of a sysadmin. He also did a live demo, and showed benchmarks of HAST.

Mark Kettenis recalled his work on porting OpenBSD to Sun's UltraSPARC T1 and T2 processors. He was the speaker for the last session of the week-end. His talk was quite technical, and gave a very interesting insight into the porting effort to a new architecture. In particular, this porting effort had several aspects, as UltraSPARC provides very nice hardware virtualisation features. The OpenBSD port is now mature, and OpenBSD can be used as both a guest and the control domain. Some configuration tasks related to hardware partitioning still require booting to Solaris, though.