Hacker News new | past | comments | ask | show | jobs | submit login
DragonFly BSD 4.2 released (dragonflybsd.org)
96 points by ceratopisan on June 29, 2015 | hide | past | favorite | 24 comments



Good to see progress on HAMMER2. The removal of SCTP is a very interesting decision. reapctl(2) has since been renamed to procctl(2), I believe.

DragonFly BSD is an amazing OS that really pushes the limits of the Unix paradigm. For instance, things like process checkpointing as a single syscall with two flags of CKPT_FREEZE and CKPT_THAW, among many other things [1] [2]. Elegant simplicity and powerful functionality.

Any self-respecting Unix (particularly Linux) developer needs to check it out and preferably steal a feature or two. DF is a middle ground between Unix workstation and cluster OS. Anything you do on that ground is bound to be more interesting than Docker.

I've always wondered if someone could add some polish and marketing to make DragonFly a solid competitor to the likes of CoreOS and Mesosphere. It already has similar features neatly integrated with a low cognitive usage overhead, so it's mostly a matter of being a good salesperson.

[1] http://www.dragonflybsd.org/features/

[2] https://www.dragonflybsd.org/docs/developer/DragonFly_Techno...


I'm in favor of Dragonfly removing SCTP. It was okay on paper but never saw adoption and was basically doomed after most firewalls were configured to block it--this is why people are building their own transports on top of UDP these days. It would almost be impossible to launch a new IP magic protocol number now.

I dig the other stuff that Dragonfly is doing too. I've been a longtime BSD fan and I like that there is a strong-willed counterpoint to FreeBSD with a different architectural philosophy. The pressure pushes FreeBSD forward faster than if Dragonfly wasn't around.


SCTP is very much alive and kicking. In fact I have just been debugging an issue with it before reading this post. It's used between Mobile Network Elements, especially in the 4G Space, as the transport that DIAMETER sit on top of.

So I agree it's niche, but it's far from a failure.


It's cool to hear someone using it, but it's still doomed to be relegated to use inside private networks versus being used as a general use internet transport protocol, which was the point I was making.


Ahh, yes. I agree, there's very little SCTP used in public networks.


NAT being the big reason. Ironically now more than ever before is the time for NAT to disappear as IPv6 is deployed.


I agree with SCTP removal http://blog.ipspace.net/2009/08/what-went-wrong-sctp.html

SCTP it's not so adopted as TCP replacement for a few reasons http://stackoverflow.com/a/20290710/66242


This is so true, BSD's in general are very underrated.


I don't think people under rate it. The issue is that BSD is so much server focused in users and development and it lags behind Linux Desktop (which is still a small percentage). There is just a high amount of programs that are not available in BSD.

I use to have me personal server as BSD and used FreeNAS quite a bit. Due to lack of certain "cool" applications I wanted to run on my machine I just switched to BSD.


From my perspective, the BSDs are focused on what they should be: making a good version of Unix. I feel that the Linux developers are spending too much effort on the "desktop", which is why we have now have GNU/SystemD.


That's true. I hate the great divide amongst Linux & BSDs cause otherwise we could have the best of both worlds.


There is not a "high amount" of programs unavailable in BSD.

There are specialized programs that need some Linux-specific syscalls which won't run on it. That's about it.


I've been keeping track of DFly BSD for a while now, and I really like where it is headed. The network stack alone is polished enough that if I were to, say, start a local ISP, I would probably be using it as the primary backend.

Dillon's decision to support only x86-x64 I think was a good choice because it helped reduced the workload associated with his multiprocessing/threading system, which is another awesome set of features that I have a feeling will end up forcing Linux and Mach to re-evaluate how they do some things.

For me, I think it may be time to move it out of a VM and onto a semi-regular use machine bare metal.

Once Hammer2 is prod ready I have a feeling I may like it better than ZFS or BTRFS, but only time will tell I suppose.


I've recently learnt they are rewriting IPFW (https://www.dragonflybsd.org/docs/ipfw2/) to fully take advantage of DFly's features. Eventhough I am not particularly fond of IPFW's syntax and prefer PF it's nice to see that a good SMP friendly firewall is in the works.


And see this recent patch [1] from an Oracle developer to make PF SMP proof. It's under review.

[1] http://marc.info/?t=143526332700004&r=1&w=2


"Eventhough I am not particularly fond of IPFW's syntax and prefer PF it's nice to see that a good SMP friendly firewall is in the works."

OpenBSD is slowly working to make PF SMP friendly: http://quigon.bsws.de/papers/2015/asiabsdcon/mgp00033.html


So is NetBSD[0]... exciting times in *BSD-land...

[0] http://www.netbsd.org/~rmind/npf/


npf is not pf, if that's what you're implying (can't tell?)


ah no -- I was only talking about multiprocessor-ifying packet management.

edit: s/package management/packet management/


oh, ok, yes -- this is getting traction in almost every project. It's nice to see this won't be a bottleneck.


>GCC 5

A bit surprising for a BSD project, seeing as 4.2.1 was the last GPLv2-licensed release and most of the rest of the BSD communities seem, shall we say, not too enthusiastic about GPLv3.


NetBSD also isn't bothered much the GPLv3 change of gcc, NetBSD 7 will ship with gcc 4.8


Its more a case of wanting to use two compilers and GCC is the other choice after clang.


DragonFly has always had more of an emphasis on user-friendliness than the others.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: