DragonFly At DragonFly 1.0-CURRENT

CoolVibe writes "For months, the DragonflyBSD fork of FreeBSD was maintaining compatibility with the existing FreeBSD-STABLE branch by using the 'FreeBSD 4.8-STABLE' name internally. In a few commits, Matt Dillon changed all the names, and DragonFly is finally sailing under its own banner. Things that DragonFlyBSD already has that FreeBSD-STABLE doesn't are, among others, application checkpointing, variant symlinks (not unlike Domain OS), Light-weight kernel threads, a more efficient slab-allocator, a multithreaded network stack, and the rcNG system."

Comments

I wonder

[Leroy_Brown242]

One has to wonder if DFBSD will die due to lack of following, or if it will be the next awesome BSD. I am currently running FreeBSD 4.8 Release on my workstation. It might be worth it to grab a spare machine and install it to see what's up. Only if the distributed.net client will work on it though. ]:3}>

Re:I wonder

[CoolVibe]

It's already very good (and stable). Well allright, sometimes it breaks (that's development for ya), but wait a few hours, and the system is stabilized again (patches are backed out, or bugfixes get in). It's not like CURRENT, where you can hose your system beyond recognition if your not careful.

Your Distributed.net client will Just Work Fine(tm). I'm still running that KDE 3.1.4 on my laptop that I compiled under 4.9-RCmumble, and that's still working quite spiffy.

Dragonfly is almost a drop-in replacement. You can just pull it over your existing 4.x-STABLE box, and all your apps should work fine (except for kernel modules). Oh, if you have an NVIDIA gfx card, I ported the binary kernel module to DFBSD, and it's sitting smugly in the override ports. (read dfly's UPDATING about dfports)

DragonFly as of current perfectly fills that niche where people want the latest and greatest, but don't want to run FreeBSD-CURRENT just yet. It's mostly production-safe. You can always revert back to your old FreeBSD-STABLE without much hassle.

So, if you're really curious, just give it a go!

Re:I wonder

[merdark]

I think this is one of the most promising new OS (or OS varient if you want to be picky). Personally, I can't wait for the package management system to get set up. Eventually, you will be able to run DBSD with binary only upgrades and installs. Should be cool.

Ports are great, but damn does my p2 400 dislike hours and hours of compiling. :)

Variant symlinks are really cool

[Euphonious Coward]

Variant symlinks as found in Domain OS (nee Domainix, nee Apollo Aegis) are symlinks that refer to environment variables, e.g.

ln -s '/etc-$(HOSTNAME)' /etc

to help enable sharing the root file system. (I don't know the variable-reference syntax used in Dragonfly). This was one of the really cool things about Aegis, which was based on Multics, not Unix. Unix/Linux/BSD have still not caught up to the networking capabilities of Aegis, and what they do have is usually clunkier than the way it was done in Aegis.

I thought about implementing variant symlinks on Linux. Probably it would need a new system call to tell the kernel where the process keeps its environment variables, to be run at each program startup, and a new process table entry field.

Re:Variant symlinks are really cool

[CoolVibe]

Variant symlinks as found in Domain OS (nee Domainix, nee Apollo Aegis) are symlinks that refer to environment variables, e.g.

ln -s '/etc-$(HOSTNAME)' /etc

to help enable sharing the root file system. (I don't know the variable-reference syntax used in Dragonfly).

It's exactly the same. The variables for symlinks can be set with the varsym(1) tool, and of course the vfs.varsym_enable sysctl has to be set to '1'. You can set symlink variables in global, user and process context.

Re:Variant symlinks are really cool

[CoolVibe]

You are right of course. I don't use varsyms yet, but I had a look at the code and that confirms what you said. Not much documentation yet, but the varsym tool has a manual page. It works like you expect. One sets a variable with varsym, and then you can use that variable (with ${} notation) in symlinks. Example:

> cat > file-file1
this is file 1
^D
> cat > file-file2
this is file 2
^D
> ln -s 'file-${foo}' test
> varsym foo=file1
> cat test
this is file 1
> varsym foo=file2
> cat test
this is file 2

Of course, varsym defaults to user context, which means that the var is only 'visible' to the user. With the -s flag, the var becomes system wide, and you can also restrict the view to the process.

There isn't much more to it. It's easy, clean and simple :)

Re:Variant symlinks are really cool

[CoolVibe]

They will be used, but not gratuitously. It's mostly to get our packaging system to go where we want to, without having to worry about the new VFS stuff yet. Variant symlinks solve about 90% of the problems at 10% of the cost, so that's why they were implemented. Also, it kinda depends on what the person that installed DFly on his machine is doing :)

When DFly's VFS subsystem is getting into shape, the need for varsyms will be less and less.

Hmm, well..

[JVStalin]

Things that DragonFlyBSD already has that FreeBSD-STABLE doesn't are, among others, application checkpointing, variant symlinks (not unlike Domain OS), Light-weight kernel threads, a more efficient slab-allocator, a multithreaded network stack, and the rcNG system." Oh, boy! Let's look at 5.1-RELEASE's features: rcNG, KSE, Mandatory Access Control framework, better SMP, fast ATA drivers. I hate to say it, but, DragonFlyBSD is all most as silly as that xMach project. :-) It's about arrogance, not software.

Re:Hmm, well..

[Anonymous Coward]

You might want to take a look at the DragonFly design page. FreeBSD-CURRENT's SMP design is based
on the traditional "lock everything down" model, which has scaling problems with current hardware trends. In addition, their particular implementation will be particularly hard to maintain in the long run. Really, the only thing you mention that is
likely to be considered for incorporation in DragonFly is tht ATAng.

You sound like one of the BSD developers who dislikes Matt on principle. So you may be right in your case about the arrogance vs. software.

I really wish I had a spare box right now...

[acidtripp101]

The ONLY thing that was keeping me from using dfbsd was the fact that I had to
a) install freebsd-stable
b) cvsup the dfbsd sources
c) recompile everthing
d) then have my system

Now that dfbsd has it's own ISO, I might have to find an old junk box somewhere to install it on (I actually like freebsd 4.x more than the 5.x series so far... MUCH faster, but I'm sure that'll change when it goes stable (no more debugging symbols, etc.)

Re:What kind of chip are you running?!?!

[merdark]

Go touch the core of your P2 when it's working and it'll talk through you. Its a very intense experience. ;)

MIDI support?

[thanjee]

Doe DFBSD have *WORKING* Kernel MIDI support?

The first BSD to have that will become my favourite. So far FreeBSD is in the lead :)

Re:Server Names

[CoolVibe]

Might I refer you to RFC 2100? It explains why some hosts have such weird names sometimes. :)