News for hackers. Stuff that matters.
Re:DragonFly BSD 4.6 Released by PeCaN on Wednesday August 03, 2016 @04:59PM (#12221466)
I run Postgres, MonetDB, Redis, and Disque on DragonFlyBSD for various hobby projects. It's been extremely stable, easy to administer¹, and fast².

Two of my favorite things:

1. HAMMER, and the upcoming HAMMER2, are very nice filesystems. HAMMER isn't a CoW fs like ZFS and btrfs; HAMMER2 is however. Also, somehow HAMMER2 is more stable than btrfs despite having about 3 developers.

2. Like Solaris, DragonFly isn't afraid to deviate from and advance traditional *nix design. vkernels, lightweight threads, a super cool message-passing API, NT-esque architecture of a decoupled microkernel-ish design running in one address space, etc.

It also has neat cluster features not unlike VMS and Plan9, but I haven't used them.

--

¹ I use FreeBSD as a desktop OS so my opinion here could be a bit skewed. But I think, like most BSDs, it's a very well-thought-out system.

² Disque³ on my DragonFly VPS is a bit faster than on my similar Arch Linux VPS; otherwise my workloads are too small to make a good comparison.

³ https://github.com/antirez/disque

Re:DragonFly BSD 4.6 Released by anexprogrammer on Wednesday August 03, 2016 @05:08PM (#12221533)
> despite having about 3 developers

If one of those developers is Matt Dillon low numbers are rarely a hinderance. He turns out an awful lot of superb quality code, ever since the days of DICE on the Amiga. That ended up generating faster code than either of the "professional" offerings, including Lattice. Lattice C on MSDOS later became the first two versions of Microsoft C.

One of the people I respect most in this industry.

Re:DragonFly BSD 4.6 Released by PeCaN on Wednesday August 03, 2016 @10:05PM (#12222891)
Yeah, Matt Dillon is up there with John Carmack for me in terms of programming masters. He pumps out a lot of high-quality code.
Re:DragonFly BSD 4.6 Released by olavgg on Thursday August 04, 2016 @05:59AM (#12224494)
I think the same about Poul Henning Kamp. I feel I read his name in every module in FreeBSD. Jails, GEOM and so on. And Varnish is a fantastic well though product from him.

The architectural difference of how data is stored between Redis and Varnish is very interesting to read about. And I think it tells us why that Redis performs better on Linux and Varnish performs better on FreeBSD.

Re:DragonFly BSD 4.6 Released by rodgerd on Wednesday August 03, 2016 @05:49PM (#12221819)
> NT-esque architecture of a decoupled microkernel-ish design running in one address space

Given Mr Dillon's background I'd have called it Amiga-ish, but perhaps that's just nme.

Re:DragonFly BSD 4.6 Released by PeCaN on Wednesday August 03, 2016 @06:07PM (#12221936)
I'm more familiar with NT's internals than Amiga's, but I'm sure Amiga is what influenced the design.

Though Amiga was more of a true microkernel but without memory protection; NT and DragonFly have memory protection but the kernel is a single executable composed of multiple message-passing modules.

Re:DragonFly BSD 4.6 Released by anexprogrammer on Wednesday August 03, 2016 @06:16PM (#12221985)
I'm fairly sure, around the time of the original fork, Matt announced the intention for DragonFly to take inspiration from the Amiga for some aspects of the new OS and kernel. Messaging was one of them. I forget the others.