BSDNexus forums has an interview with DragonFly lead developer Matthew Dillon. They discuss a wide array of topics ranging from recent VFS work to SSI clustering.
If there’s one thing next to LWKT that DragonFly is doing right, it’s reimplementing the VFS interface to use object lists rather than strings. The traditional Unix VFS interface is just so curmudgeoned and crufty to incorporate new ideas into, and I wholeheartedly agree with moving as much complexity as possible out of the kernel and into userspace.
Once they’ve got further along with their admirable work – I think a lot of people would be interested in a textbook on the matter. Something that explains not just BSD, the issues thet Dragonfly felt needed addressing, and how they addressed them.
There’s enough technical meat there for a book or two.
It is going to be interesting to see how FreeBSD develops and contrasts with DragonFlyBSD. The approaches definitely diverge and I seem them growing into two very distinct trees of code (they already are.) Since 5.3, it seems that FreeBSD 5.x is starting to stabilize. DragonFlyBSD is also making steady progress towards their ambitious goals and remains relatively stable (considering the huge amount of redesign that is occuring in the kernel.) Perhaps DragonFlyBSD will get better traction in the scientific computing circles where the clustering capabilities and scalability will be a huge boon to them.
FreeBSD 5.3 still has bugs (like usb printing recenty pointed out in the lists) that have been there for 8+ months. I still believe they lack the needed resources to finish the fine grained locking system. There’s also the problem of the system being inefficient on UP hardware (95% of the users), while DragonFly has no problems scaling up and down.
If there’s one thing next to LWKT that DragonFly is doing right, it’s reimplementing the VFS interface to use object lists rather than strings. The traditional Unix VFS interface is just so curmudgeoned and crufty to incorporate new ideas into, and I wholeheartedly agree with moving as much complexity as possible out of the kernel and into userspace.
Once they’ve got further along with their admirable work – I think a lot of people would be interested in a textbook on the matter. Something that explains not just BSD, the issues thet Dragonfly felt needed addressing, and how they addressed them.
There’s enough technical meat there for a book or two.
It is going to be interesting to see how FreeBSD develops and contrasts with DragonFlyBSD. The approaches definitely diverge and I seem them growing into two very distinct trees of code (they already are.) Since 5.3, it seems that FreeBSD 5.x is starting to stabilize. DragonFlyBSD is also making steady progress towards their ambitious goals and remains relatively stable (considering the huge amount of redesign that is occuring in the kernel.) Perhaps DragonFlyBSD will get better traction in the scientific computing circles where the clustering capabilities and scalability will be a huge boon to them.
I tried the 1.0a release, but it wouldn’t load the driver for my Via Rhine II NIC. Sadly that kinda killed my first go around with it.
I am curious, and I am looking forward to trying the next release once it is announced.
FreeBSD 5.3 still has bugs (like usb printing recenty pointed out in the lists) that have been there for 8+ months. I still believe they lack the needed resources to finish the fine grained locking system. There’s also the problem of the system being inefficient on UP hardware (95% of the users), while DragonFly has no problems scaling up and down.
Wrong Matt Dillon.
You mean the Sheriff from Gunsmoke