Linked by Eugenia Loli on Wed 31st Dec 2003 18:41 UTC
Gifts, Contests, Easter Eggs Here is a short list of the things I totally loved this year. Use the comment section to tell us about your favorites too!
Permalink for comment
To read all comments associated with this story, please click here.
Let me clarify my position
by Kingston on Thu 1st Jan 2004 13:51 UTC

Let me clarify my position regarding the "OS with most overall potential." Also note that it is based on my current (probably flawed) understanding of the technology.

Matt lost his commit bit in FreeBSD land in early 2003 (late january of early february, I can't remember). DragonFlyBSD was first announced on 16 July 2003 by Matt Dillon om the FreeBSD mailing lists. It was derived from FreeBSD 4.8 which was released on 3 Apr 2003 by the FreeBSD project.

After "destressing" for a few months, he says that he decided to implement one major subsystem 'goal' before announcing it. This initial subsystem was the Light weight kernel threading subsystem. This means that in the space of a few months, he nearly single-handedly wrote a brand new completely mutexless threading system for the kernel, as well as the beginings of the light weight ports/messaging system to go along with it. A number of former kernel processes (such as the pageout daemon) have been reimplemented entirely as threads, and under the LWKT system are free from both any process context, as well as from the Giant kernel lock.

The ports/messaging system will allow DragonFly to natively support clustering, without the need for a third party message passing system. The fact that each CPU in a system has it's own self-comtained LWKT scheduler, and the fact that the threads are per-CPU, and are not preemptively moved from one processor to another except through the "asyncronous IPI messaging interface," should allow the DragonFly kernel to scale much better than other OSs, as there is less "guessing" involved in thread migration operations among other things. I am told that the message passing system also (nearly) completely removes many situations where deadlocks could occur in various opperations.

Since becoming a public project, a few interesting new features (new to BSD if not to computer science in general) like Application Checkpointing, and Variant Symlinks, have been written, and ACPI, PAE and the ATAng frameworks have been imported from FreeBSD. There is a hack that allows the use of the FreeBSD ports collection until DragonFly gets it's own based on either the Variant Symlinks or the forthcoming message based VFS framework. The nVidia drivers have been ported to the kernel. The conversion of numerous subsystems in the kernel to the new ports/messaging system is also seeing tremendous advancement.

This is a hell of a lot of progress for such a new project, to have occured in so short a time, and with so few people. Currently the project includes seven committers and a handful of people submitting patches.

DragonFly is still a development OS and is not claimed to be ready for production use, although in my own experiences, and in those of others, it seems that it is. It is fast (I'd REALLY like to perform the tests on DragonFly ASAP!), and it seems every bit as stable as FreeBSD 4.8.

If indeed it proves to be as scalable, debuggable, maintainable and extensible as Matt Dillon believes that it will be (in the first few releases), than I can see DragonFly BSD giving both Linux and FreeBSD (as well as a number of other OSs for that matter) a run for their money, in other words, it is IMO the OS with the most potential.