Submitted by A recent vague announcement on osol-announce hints that something big is rumbling for OpenSolaris: “A number of the community leaders from the OpenSolaris community have
been working quietly together on a new effort called Illumos, and we’re
just about ready to fully disclose our work to, and invite the general
participation of, the general public.” They have a website, and they’re going to be hosting a conference call on August 3.
If it is a fork, it will be interesting.
But I don’t think there are a viable amount of volunteers to keep up with all the code in the OS.
I hope it can be done. The more diversity in operating systems, the better… and it’s sad an operating system with such a history and being a true free UNIX has to go.
To a point I agree. It doesn’t help much if the total pool of developers are spread so thin that development of each individual OS is slower than it need be.
Diversity is fine so long as it doesn’t distract.
As far as I am concerned, FreeBSD and other BSD’s are, at least, as much UNIX as Solaris is.
Indeed. But the OP didn’t claim *BSD wasn’t Unix. The OP just stated it would be sad to see a (not the) free Unix disappearing.
How? By not being POSIX and SUS compliant?
How old are you? Because either you weren’t around or you must not have known that POSIX & SUS don’t determine what’s Unix & what isn’t. They’re just standards that allow all of the various Unix derivatives to coexist & work together by supporting the same API’s. There was Unix long before there was POSIX & SUS.
Brush up on your history!
The bottom line is, that Open Group is the owner of the Unix trade mark and they get to decide about what the definition of Unix will be. I’m not saying that FreeBSD is bad, you just can’t call an OS Unix if it isn’t certified and doesn’t even comply to required specifications.
Actually the *BSD’s are Unix. They are just not Certified UNIX. That does not mean they don’t comply to required specifications (which one btw.? UNIX 93/95/98 or one of those from this millenium?) or aren’t POSIX compatible.
OS X is Certified UNIX but is definitely no Unix. OpenBSD is definitely Unix but not Certified UNIX. The *BSD’s can claim to be Unix since they are derivatives of BSD Unix. Not paying for certification doesn’t mean they disqualify as Unix. They just can’t use the “UNIX” trademark.
Based on your OSN-past you seem to carry an animosity towards GNU/Linux, *BSD and even OpenSolaris (at least in the past).
Some people obivously don’t think so:
http://en.wikipedia.org/wiki/POSIX#Mostly_POSIX-compliant
This sentence contradicts itself.
Maybe I don’t recall correctly, but AFAIK the Unix code was removed from BSD at one point.
All in all, It’s hard to reason with these kind of folks who approach technology like it’s some kind of religion. You can argue about history all you want, but that’s not really relevant to the topic. By such logic, you could start calling Windows OpenVMS. Or OpenvMS.
Being a sysadmin I certainly don’t like how Linux is aggressively pushed to replace systems, to which it is inferior in the server space, but hey, I’m a Linux desktop user for almost 10 years, so there’s hardly an animosity.
I actually like the BSDs, I just don’t consider them Unix. I don’t remember posting anything negative towards BSDs.
What??
You don’t happen to be a Linux user, do you?
Go where? I suppose you mean “go forward”, right? Because I don’t see it going anywhere else.
Don’t even start on the whole “gazillions of Linux distros” thing. Linux is still Linux, no matter what the distro; it’s just the kernel. And yes, I do “run Linux” in the form of a certain complete distribution based on it.
Are you out of the loop or something? With Oracle at the helm, it doesn’t appear to be going anywhere soon. Hell, even Opera discontinued Solaris support following the Oracle-Sun buyout. And so far its future has continued to look grim.
It’s not only about the kernel, different distros have different filesystem layouts, init systems, package managers, boot loaders, window managers, graphic libraries, etc… Many developers said in interviews and blogs, that they are simply unable to support Linux because of this. I really hope that (Open)Solaris won’t get to this state.
So do I, so I know what I’m talking about.
I don’t see how.
This really made me laugh. Who cares about Opera?
It’s not who cares about Opera, it’s that Opera, who cared for Solaris for a LONG LONG time has stopped doing so.
I believe that they are way too fast to announce something. They didn’t make a poll to ask for ideas.
I agree but they can re use code. Fiasco.OC or Genode OS can re-use code at the kernel level. They can reimplement Opensolaris on top of them. L4Env is agood candidate and they can re-use LINUX-DDEkit or improve FREEBSD-DDEkit. They could also create their own OSOL-DDEkit. My take is to use directly IOKit and port all the drivers there but
1. I cannot estimate porting time since it uses mach and not L4 (possibly not that difficult)
2. Licencing is another setback.
You minimize the volunteers needed in such ways. This is what is traditionally done at the applications level.
Edited 2010-08-02 07:21 UTC
Dude, L4 is microkernel, Genode is something completely different…
Solaris’ sunos kernel is monolithic kernel and btw it is pretty mature. Putting up L4 dancing baloney on top and experimenting with it would throw all that maturity down the drain. Not to mention that by the time such integration can even begin to look like it’s working, everybody will forget that OpenSolaris even existed. And then there is non-fixable and perpetual licensing issue wrt CDDL.
Only reasonable strategy is to take drivers code and libc from FreeBSD and mesh it with OpenSolaris carcass in order to get something workable as fast is possible. That is only technological and legally compatible solution. It would throw down any binary compatibility with Oracle Solaris, but it could get the project going. But then again, porting sunos kernel to BSD libc is a daunting task.
I hope it is not too difficult to just compile the current OpenSolaris source code and swap the proprietary encumbered parts to open source. If they succeed with swapping, then in principle it is just a recompilation every time when Oracle checks in OpenSolaris source code. Which would be often, because development of the OpenSolaris source code is very rapid.
I don’t think that can be done. Remember Mainframe port of OpenSolaris? Oracle killed it just by firing the guy who was assigned to compile blobs on the mainframe. If Oracle stops compiling those blobs for every release, it is as good as dead. It can’t run without libc.
That’s exactly what illumos is working on – i18n is the first (and probably most important) step.
And Garrett actually tries to achieve binary compatibility.
You mean they are rewriting i18n ?
If yes, that’s what I said it should be done. Either rewrite the blob or replace it with FreeBSD libc.
If they are just trying to make a interface to bolt the blob on top of it, then that won’t work. Oracle will screw the easily that way. If nothing else, Oracle can revoke license for the binary and assert copyright infringement. After all, blobs are proprietary software.
I know …
I cannot see your point.
Point is: experimenting with microkernel requires time and people willing to bother with it. They have neither. They need something quick if they want to stay relevant and attract developers. Otherwise why bother at all?
The problem is that currently they’re still hugely dependent on Oracle for i18n in libc and they’ve yet to create a compatible replacement. Until there is full emancipation or at least some sort of time table so that a full distribution can be created external of Oracle – it’ll be a difficult thing to get the traction required. I also wonder that give Oracle’s willingness to hold back contributions will result in a second class distribution – and will that also impact on contributors who work for Oracle? will Oracle clamp down on what programmers can work on?
There is also a string of issues that need addressing; HAL being removed, the role of upower/udisk, the lack of external contributors today (or have they been put off by the difficulty of getting code included?), the replacement compiler – LLVM be an option or sticking with GCC? Is there any corporate backing? How much communication has there been with Oracle regarding this fork? are they killing off the OpenSolaris distribution in favour of letting a more community controlled one take shape which won’t get hamstrung by internal Oracle politics?
Hopefully when there is the conference 3 August that some of these issues will be addressed because such not only affect developers but also end users as well – end users want to be assured that the time and effort they spent training, using, installing and advocating isn’t a waste.
These are my thoughts exactly. The benefit of Opensolaris was that it was/is a pre-preview version of the upcoming Solaris, but backed by the company. You could be assured that new features would eventually make it into the preview version of Solaris and quite possibly,the general OS release. Will contributions from this go into Opensolaris if it continues or pre-Solaris? The relationship will be the most interesting to me.
True, there was always the ability to say in response to “is it corporate backed” to respond saying, “yes it is corporate backed’ when advocating it. Solaris ‘classic’ might be great for the high end of town but isn’t that simply ignoring small to medium businesses – those who make up the majority of economic activity? it seems that like Sun, Oracle is ignoring the small to medium business at their own peril.
Looks like Garrett is working on it. I wouldn’t be surprised if they will pull in stuff like this via *BSD.
http://hg.illumos.org/illumos/rev/c8da1d642945
As for the project. If Garrett is the lead then I think it should be very interesting. He is definitely one of the main kernel hackers doing code commits to OpenSolaris over the last 2-3years. He designed and implemented “Boomer” which is the new sound system which exposes an OSSv4 API. He’s also done alot of driver work.
There was a move to port the BSD implementation over to OpenSolaris:
http://i18n-freedom.blogspot.com/
But it pretty much died around a year ago with no code every moving from simply some proof of concepts into reality. Garrett has done a heck of a lot of work so far – hopefully we’ll see more people jump on the bandwagon and maybe provide a donation link as well because I know I’ll be more than happy to donate $500 once I get working full time (once I graduate university).
I wonder whether there will be a move to get behind LLVM as the default compiler for illumos given how much backing LLVM is getting especially from Apple – although from what I understand its lacking SPARC support. With that being said I’d say the majority of illumos users will end up being x86 users anyway.
The LLVM sounds interesting but we should wait for some more releases until LLDB is ready. I wish OSOL could reuse the publicly released Darwin bits. However to my understanding APSL forbids it.
Going by the recording of the meeting. It would seem they would ideally like to be toolchain neutral. At the moment they basically have to depend on SunStudio which though free is a proprietary compiler suite. Garrett basically says he sees this dependence as a bug. I think the way LLVM/Clang are going especially with it’s inclusion in FreeBSD base system that there will no doubt be work done bringing it over at some stage.
Recording of call:
http://www.illumos.org/attachments/download/4/illumos_recording.mp3
Presentation pdf:
http://www.illumos.org/attachments/download/3/illumos.pdf
“The problem is that currently they’re still hugely dependent on Oracle for i18n in libc and they’ve yet to create a compatible replacement.”
Exactly. And, that will Not be easy to pull off.
Good luck to them if they do try it.
Steven
True, good luck – unless illumos are happy to break compatibility as so far as binary compatibility, it’ll end up being a ‘highway to hell” chasing something that’ll making it even more difficult.
A guy on Linked in found a hg Repository:
http://hg.illumos.org/illumos/
Recent commits include:
“/etc/motd needs to say Illumos.org, not Oracle.”
By the looks of that repo, they seem to consider OpenSolaris as an upstream, so it would be a similair relationship to Ubuntu where Debian is the upstream.
I do not think it is safe to conclude anything from the repo at this stage. But you do have a point. Since Ubuntu is a somewhat friendly fork of Debian, so might Illumos be (of OpenSolaris*). After all it is Nexanta System Inc. who is behind the effort, so comparison with Ubuntu is not entirely implausible.
EDIT: Clarification added *
Edited 2010-08-02 14:26 UTC
Sadly, I think OpenSolaris is ZombieSolaris. It’s dead, but continues to walk around.
If Sun/Oracle abandons OpenSolaris (which it looks like they may have already done defacto), there’s not enough independent kernel developers to do anything but keep the OpenSolaris frozen in time. And an OS frozen in time isn’t much of an OS with a future.
I really liked Solaris, I cut my teeth on Solaris. But I think OpenSolaris is a zombie, and I think Solaris is pretty marginalized right now to a few specific cases (and I think Oracle is fine with that).
There is a difference between OpenSolaris the project and the “Opensolaris Distro” (what used to be called Project Indiana). There was a huge amount of controversy in the community when the “Indiana distro” was rebranded as OpenSolaris. As it caused confusion. The dev repo is currently stuck at b134 which was suppose to be used for “OpenSolaris 2010.03” distro. However they have continued to do drops of Opensolaris codebase. For example you can download the code for B145 and compile it yourself. The project is still running at full steam internally from what I can see. Given that it will produce Nevada (Solaris 11) they just don’t seem to be releasing pre-compiled code for the Indiana distro.
http://dlc.sun.com/osol/on/downloads/b145/
At the moment it looks like Oracle are pursing a Redhat like strategy. They make the code available but don’t release a free binary distro. Instead you will end up paying for Solaris 11 like you have to do for RHEL. What we need is a Centos of the Solaris ecosystem. Illumos appears to fit this niche.
As you can see looking at the build flagdays there is alot going on when it comes to code commits.
http://static.opensolaris.org/on/flagdays/146-150.html
Since nexenta is involved, it might be that they just grab the kernel and use Linux userspace. Which is the only way it’ll make any sense.
Be it as it may, people on Solaris / OpenSolaris have probably been browsing the red hat website since Oracle purchase.
Illumos can’t be nothing like CentOS becasue Solaris 10 is closed soruce. RHEL is open source. CentOS is a carbon copy of RHEL, just without trademarks.
You just can’t do that with Solaris. Illumos is new name for OpenSolaris, because Oracle owns trademark. But it is not just trademark what it’s about, because there are some big binary-only parts in opensolaris codebase, and they need to replace that if they want to make completely open source distro.
Hmmm… While I hope you are wrong I will say this:
Zombie would be a MUCH cooler project name than Illumos.
http://www.theregister.co.uk/2010/08/03/illumos_opensolaris_spork/