OpenSolaris Governing Board resigns [LWN.net]

As expected, due to Oracle's unwillingness to appoint a liaison to work with the board, the OpenSolaris Governing Board (OGB) has resigned. From the resolution: "[...] Whereas, without the continued support and participation of Oracle in the open development of OpenSolaris, the OGB and the community Sun/Oracle created to support the open Solaris development partnership have no meaning, and
Whereas the desire and enthusiasm for continuing open development of the OpenSolaris code base has clearly passed out of Oracle's (and thus this community's) hands into other communities [...]"

to post comments

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 1:26 UTC (Tue) by martinez.javier (guest, #59933) [Link] (34 responses)

It's a shame that a company like Oracle doesn't understand the power of the open source development model.

The reason why IBM, HP, SCO and others don't have a propietary Unix OS anymore is because is cheaper and better for them to have a team of kernel developers that work against Linux. Clearly doesn't exist a company that could afford the cost of the whole Linux ecosystem, it just doesn't make any sense. This applies to Solaris/OpenSolaris as well.

If Oracle pretends to continue developing OpenSolaris behind closed doors, the only thing they will do is killing the OpenSolaris project and the Solaris product as well.

It's strange that Oracle has a totally different approach with Linux, where the company have engineers working upstream and making Linux a better OS.

When the same company have different behaviors, I just dont get it. Maybe that is why I will never understand the corporate world.

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 2:02 UTC (Tue) by vv (guest, #69729) [Link] (24 responses)

Hate to burst your bubble, but both IBM and HP still have proprietary Unix offerings (namely AIX and HP-UX). They might not be as relevant now as they were, but they still exist and are still developed.

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 2:53 UTC (Tue) by lakeland (guest, #1157) [Link] (23 responses)

This is technically true, but I don't think it is relevant.

I do not know of any new sales of either HPUX or AIX. In my opinion, both companies continue development in order to support legacy customers rather than to offer a viable alternative to Linux.

It's almost like saying IRIX was still in development 5-10 years ago - it was, but customers were being gently and smoothly migrated to Linux.

I see this resignation as a very significant step, showing that Oracle was unable to react quickly enough to placate the open source community. I had expected Oracle to attempt to promote two visions simultaneously - at least for a while.

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 7:19 UTC (Tue) by danielpf (guest, #4723) [Link] (4 responses)

Perhaps Oracle sees no future in Solaris and wants to maintain it for legacy customers just like IBM and HP do with their respective Unix?

For Oracle, as well as for the open source community, it would make sense to focus on Linux on the longer term. There is already enough fragmentation, cutting from time to time the dead branches is healthy for the whole tree.

Solaris going forward

Posted Aug 24, 2010 12:15 UTC (Tue) by markhb (guest, #1003) [Link] (2 responses)

My reading of the supposedly leaked memo from last week didn't leave me with the idea that Oracle plans to let Solaris go gently into that good night. On the contrary, it looked more like Oracle's plan is to be able to go to a CIO and say "We'll sell you Oracle, Siebel, Peoplesoft, etc., running on an OS we've specifically optimized for them, on our hardware, which gets you one support number to call for everything with no fingerpointing between e.g. us, Red Hat and your blade vendor." That's going to be a pretty compelling proposition for a significant number of enterprise decision makers (I've spoken with at least one CIO years ago who specifically said he'd run Oracle-everything to get rid of the support fingerpointing), and Oracle doesn't have an antitrust consent decree hanging over it like IBM* and Microsoft did/do to raise red flags re: bundling.

* - I'm genuinely not sure if the old antitrust rulings against IBM from the 50's and 60's are still in effect or if they've been lifted.

Solaris going forward

Posted Aug 24, 2010 15:32 UTC (Tue) by epa (subscriber, #39769) [Link]

That's certainly an appealing proposition. However, with a bit of clue and organizational flexibility, there's no reason other vendors couldn't offer the same thing. If you run Red Hat on Dell hardware, and you report a bug to Dell's helpdesk which their techs believe is Red Hat's fault, it should be their job to take it up with the OS vendor and sort it out between them.

Solaris going forward

Posted Aug 24, 2010 23:34 UTC (Tue) by lakeland (guest, #1157) [Link]

Yes, that was exactly my reading too

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 14:38 UTC (Tue) by zotz (guest, #26117) [Link]

If that was the case, they should GPL or AGPL the lot.

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 7:45 UTC (Tue) by trasz (guest, #45786) [Link] (12 responses)

I do know. And the development continues not to offer an "alternative", but to offer a functionality that is not available with Linux. Oracle is doing the same thing.

As for IRIX - yeah, 5-10 years ago the situation looked exactly opposite, with major Unix vendors trying to migrate to Linux.

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 23:26 UTC (Tue) by Cyberax (✭ supporter ✭, #52523) [Link] (11 responses)

"I do know. And the development continues not to offer an "alternative", but to offer a functionality that is not available with Linux. Oracle is doing the same thing."

Yes, there's still areas not adequately represented in Linux, but they are shrinking fast. Few more years, and Linux will catch up in these areas. Another couple of years and Linux will surpass commercial offerings.

Winning on full-stack integration? That's more like it.

OpenSolaris Governing Board resigns

Posted Aug 25, 2010 7:20 UTC (Wed) by trasz (guest, #45786) [Link] (10 responses)

You're assuming that Linux is catching up, which is not quite true.

OpenSolaris Governing Board resigns

Posted Aug 25, 2010 14:55 UTC (Wed) by Cyberax (✭ supporter ✭, #52523) [Link] (9 responses)

>You're assuming that Linux is catching up, which is not quite true.

In the last flamewar about Solaris we've identified three its advantages:
1) Zones (lightweight virtualization).
2) ZFS.
3) DTrace.

Btrfs is clearly going to be at least as powerful as ZFS. Linux Containers already provide most of Zones' functionality, though LXC is not yet polished enough. For DTrace replacement, something like SystemTap is going evolve, sooner or later.

And what's more important, Linux is not stagnating. It's moving forward, albeit somewhat erratically (see 'uprobes').

BTW, systemd uses Linux Containers - it's one of the best uses for container infrastructure in Linux that I've seen.

OpenSolaris Governing Board resigns

Posted Aug 25, 2010 17:07 UTC (Wed) by clump (subscriber, #27801) [Link]

Solaris's full virtualization solution? Oops. Watch for a change of subject.

OpenSolaris Governing Board resigns

Posted Aug 25, 2010 23:56 UTC (Wed) by davecb (subscriber, #1574) [Link] (2 responses)

Don't forget scalability. which Sun spent considerable money (IMHO, the
majority of their investment) on. Solaris happily supports 128 threads
on a 4-socket machine, and for an older chip, 64 sockets on a dual
backplane.

Both are significant for a hardware vendor specializing in large
conventional databases, and 32*64 sounds like a really pleasant number
no matter what kind of large program you're trying to support (:-))

--dave

OpenSolaris Governing Board resigns

Posted Aug 26, 2010 1:07 UTC (Thu) by Cyberax (✭ supporter ✭, #52523) [Link] (1 responses)

I wouldn't worry much about it given that Linux totally dominates the HPC and supercomupter markets.

There's even XKCD for that: http://xkcd.com/619/ :)

OpenSolaris Governing Board resigns

Posted Aug 26, 2010 10:40 UTC (Thu) by trasz (guest, #45786) [Link]

Except that HPC workloads are totally different, scalability-wise, from normal server workloads. This, again, is the reason why e.g. IBM uses AIX instead of Linux in scalability application benchmarks.

OpenSolaris Governing Board resigns

Posted Aug 26, 2010 10:47 UTC (Thu) by trasz (guest, #45786) [Link] (4 responses)

Btrfs is alpha-quality and it will take few years to fix that, let alone complicated subjects such as deduplication. Also, even the _wishlist_ (as published on Wikipedia) misses some important ZFS features, such as log and cache devices. Features neccessary for good SMB serving are also missing. And ZFS is still under active development and gaining new features.

Pretty much the same thing with LXC and SystemTap - it will take a few years. And other operating systems are moving forward too.

What does systemd use Containers for, btw?

OpenSolaris Governing Board resigns

Posted Aug 26, 2010 11:25 UTC (Thu) by rahulsundaram (subscriber, #21946) [Link]

Btrfs works just fine and Meego is using it as their default filesytem already. deduplication is a "nice to have" feature in many cases and not a blocker. For systemd and cgroups, refer to

http://0pointer.de/blog/projects/systemd.html

OpenSolaris Governing Board resigns

Posted Aug 26, 2010 12:17 UTC (Thu) by patrick_g (subscriber, #44470) [Link] (2 responses)

>>> Btrfs is alpha-quality and it will take few years to fix that

According to this phoronix benchmark btrfs is already faster than zfs.

OpenSolaris Governing Board resigns

Posted Aug 27, 2010 17:55 UTC (Fri) by dlang (guest, #313) [Link]

perfomance has little to do with the quality that we being referred to.

In addition, I would expect that the performance will probably drop in at least the near term as additional code needs to be put in place to fix corner cases that aren't handled properly now (and no, I'm not saying that I know of any, I'm saying that for any new filesystem there are always corner cases that reveal bugs and or pathalogical performance problems that require special-case code to fix initially)

OpenSolaris Governing Board resigns

Posted Aug 29, 2010 9:43 UTC (Sun) by trasz (guest, #45786) [Link]

Phoronix "benchmarks" are fundamentally flawed (for example, one of the "I/O tests" doesn't benchmark any disk I/O at all). Sometimes the results are really hilarious; for example, in one of the benchmarks they "measured" that SQLite on FreeBSD was an order of magnitude faster than on some Linux. There is no point in even looking at their stuff - it's a garbage.

Also, take a look at performance features that ZFS has, and which are not in the Btrfs wishlist yet, such as log devices and cache devices - so called "hybrid pools" in general. Using these properly you can get a change in performance in the order of magnitude. Meanwhile, Btrfs is still missing basic features such as support for higher RAID levels. Fundamental design flaws (http://lkml.org/lkml/2010/6/3/313) don't help either.

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 13:33 UTC (Tue) by Nelson (subscriber, #21712) [Link] (1 responses)

I do not know of any new sales of either HPUX or AIX. In my opinion, both companies continue development in order to support legacy customers rather than to offer a viable alternative to Linux.

I don't think it's quite like that. Both companies will do what it takes to win a deal and if that's an AIX or HPUX solution over a Linux solution then that's what it is. Some of the virtualization work in AIX is unlike anything that really exists in Linux yet, they are still rolling new versions of AIX with new features in addition to new hardware support. If you buy HP Itanium hardware, it's HPUX. I'm certain there are new big deals being won on those platforms, in addition to supporting old customers.

Parts of those companies, I'm sure, would love a Linux-only world. Other parts, are just fine with supporting AIX and HPUX too. What both of those companies do much better at is playing down their loyalties and the PR, they want to make money more than kowtow a specific technology. Sun never could understand that or the opensource models and Oracle has inherited it. I would have thought Oracle would have been a bit better at the PR side of things though.

Am I the only person that thinks this stuff about OpenSolaris is a little bit heavy on the PR and light on substance? I've never seen it used seriously, I'm unaware of any great community support, the few people that do seem to use it have a filesystem fetish. It'd be like close sources Ingres again, nobody would really notice beyond the press release. I get the emotional attachment but this isn't news or a surprise.

OpenSolaris Governing Board resigns

Posted Aug 26, 2010 10:17 UTC (Thu) by buchanmilne (guest, #42315) [Link]

Some of the virtualization work in AIX is unlike anything that really exists in Linux yet

s/AIX/PowerVM/

Linux can run as a Logical Partition (LPAR) on top of only the firmware-based PowerVM hypervisor on pSeries, just as well as AIX can, and is just as supported (e.g. SLES and RHEL are certified, Debian, Fedora are options), even as the VIO server (LPAR).

There may be other pSeries-specific hardware features where Linux is still catching up, but then again, good is AIX's support for x86-specific features?

OpenSolaris Governing Board resigns

Posted Aug 25, 2010 10:17 UTC (Wed) by dbht (guest, #69760) [Link] (2 responses)

I'm afraid there is a misunderstanding here. I can speak by AIX. AIX is under heavy development (AIX 7 is almost released) and I've turned my eyes to it a few months latter. Today I'm a Linux and AIX administrator and I must say AIX is better than Linux on HA tools and LVM. It's seen now with EXT4 Linux can fight at same level JFS2 in stability and administration, real meanings for people how wants to pay for a expensive hardware and software.

I believe Oracle is strategic killing Sun's open source projects to turn them in commercial available products. Take a picture: Oracle has EMC, Solaris, Hardware SUN, Database and ZFS (from Solaris). Its clearly that they are building a solution for cluoud and high volume databases based on these thechnologies. There is no meaning spending time with us if they can play alone.....

I regret Oracle's behave but this is the reality: Oracle is a business not a culture.

Cheers,

Davi

OpenSolaris Governing Board resigns

Posted Aug 25, 2010 11:34 UTC (Wed) by tzafrir (subscriber, #11501) [Link] (1 responses)

I'm trying to follow your logic here.

So, if someone from HP needs to setup a database-backed solution for a customer on a HP system with Linux or HP/UX:

* OracleDB can't be used (Oracle will not get its cut if Oracle Solaris OS and the and Oracle SPARC hardware are not used)
* DB2 can't be used (IBM will not get its cut if the AIX OS and the POWER hardware are not used)
* MS-SQL can't be used (Microsoft will not get its cut if MS-Windows is not used)

So the logical choice is PostgreSQL and the likes. Right?

Please help me understand your business logic.

OpenSolaris Governing Board resigns

Posted Aug 25, 2010 23:59 UTC (Wed) by davecb (subscriber, #1574) [Link]

Oracle can make money if the customer buys an Oracle database to run
on their HP machine. They might wish to make more, but until and
unless they have a monopoly, they will offer their products on HP/UX,
just like IBM will offer DB2

--dave

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 2:49 UTC (Tue) by ringerc (subscriber, #3071) [Link] (6 responses)

With Linux, they don't really have a choice.

They could fork the kernel, of course, and it'd rapidly diverge to the point where it'd become unrealistic to extract useful new code from it and merge it with mainline. Especially if they used the "scatter-it-everywhere" approach to defeating modularity that some corporate in-house kernel devs seem to favour. The downside would be that it'd become rather hard to merge fixes and improvements from mainline Linux into their fork, too. IMO that's Linux's real protection against proprietary forks - not the license, but the fact that technical issues mean that if you fork, it quickly becomes difficult and expensive to get new improvements from mainline, so you find yourself unable to take the work done by others for free while selling your own.

Forking Linux wouldn't do them much good even if it was practical to do, though. If "Oracle Linux" became its own semi-Linux-compatible kernel and OS, it'd be competing against mainline Linux. They'd have to invest a huge amount of development resources to get anywhere, to very little benefit. More importantly, they'd have to convince others (hardware vendors, software developers, etc) to support it.

With Solaris they have an OS with a license and contributor agreement that lets them easily take it in-house again. Nobody can grab and re-use improvements they make unless they say so. Solaris has some existing technical advantages over Linux, and the benefit of significant existing software compatibility, including with their own products. Unlike a fork of Linux they wouldn't have to convince everyone to care; people already support and use Solaris.

In the short term, I suspect they'll make more off it than Sun was going to with their OpenSolaris plan.

That said, I suspect this development will send Solaris back down the path toward irrelevancy that's being followed by AIX and HP-UX, and well traveled by IRIX, OpenServer, UnixWare, and others long forgotten. Sure, AIX and HP-UX are still used, and still developed, but they have rather niche markets these days and aren't exactly the giant cash cows they used to be for HP and IBM, despite the eye-bleeding license fees they charge to people unlucky enough to need them or the associated hardware.

Given that OpenSolaris wasn't exactly a runaway success, maybe they're making the right commercial choice even if it *does* send Solaris toward the niche/legacy area in the end. They'll make a bucket of cash out of it in the mean time. It may not become irrelevant anyway; there are enough people leery of Linux's fast-moving nature and often rather marginal or buggy support for things like NFS, multipath IO, production system tracing and diagnostics, etc. If Oracle can keep Solaris's kernel ahead of Linux's ongoing development, this may well be a good commercial decision.

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 5:16 UTC (Tue) by cmot (guest, #53097) [Link] (5 responses)

IMO that's Linux's real protection against proprietary forks - not the license, but the fact that technical issues mean that if you fork, it quickly becomes difficult and expensive to get new improvements from mainline, so you find yourself unable to take the work done by others for free while selling your own.

I'm curious to see where Android is going: it more and more looks like a fork to me, and there may be enough vendors behind it to keep it going, and the mobile hardware space might be just different enough from what most mainline Linux users are working against.

Android fork?

Posted Aug 24, 2010 8:49 UTC (Tue) by tialaramex (subscriber, #21167) [Link]

If it's a fork Google sure are paying engineers a lot of money to chase mainline for nothing. Gingerbread appears to be 2.6.33 or so, while Cupcake was based on 2.6.27.

Now, evidently Google are not just going to roll over when it comes to features they feel they need and which aren't in mainline. But that's true for Red Hat or IBM too. If you have the team to support it, a big invasive change can live outside the tree for a long time before either getting merged piecemeal or becoming obsolete.

We also know that the "separate platform" thing turns out to be bullshit. A bunch of the tricks Linux is using on my desktop PC come from big iron, and another bunch come from embedded systems. Code gets pasted into Linux from some million dollar supercomputer project, and before you know it that code is needed to make your laptop perform well. The idea that Linux is de-optimised by its broad target is popular, but so is belief in alien abduction. Both lack convincing evidence.

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 9:04 UTC (Tue) by cate (subscriber, #1359) [Link] (3 responses)

I don't think it is a real fork. I see more as quick and stupid solutions for specific usage of android and stricter deadlines, thus IMHO not usable on future hardwares (more desktop like).

But I'm curious on Google moves: they have to support three linux kernel flavours: the nearly standard one (optimized and used for infrastructure, IIRC) and the two specialized kernels: android and chrome.

Small correction...

Posted Aug 24, 2010 13:32 UTC (Tue) by khim (subscriber, #9252) [Link] (2 responses)

s/the nearly standard one/the most customized one/

It was discussed already. It's not like Google is doing something different from other vendors: most vendors heavily customize kernel. It was also discussed already (not that Andrew is also Google employee).

I think the logic here is simple: if you platform is dead because of repeatedly missed deadlines then it does not matter what happens with it long-term. If it's alive and you have time then you can think about upstream.

RedHat used this approach for years: only when they had enough clout to make sure features will be merged fast they started to work directly with upstream. Looks like Google will need to do that too... but they can decide that it's cheaper to support forks...

Small correction...

Posted Aug 24, 2010 13:44 UTC (Tue) by arjan (subscriber, #36785) [Link] (1 responses)

> RedHat used this approach for years: only when they had enough clout to make sure features
> will be merged fast they started to work directly with upstream

as someone who worked at Red Hat for years on kernels, I would like to call your remark utterly false and misleading. You can like or not like Red Hat for many reasons, but they have been contributing very very actively to the Linux kernel since day one, and as someone who worked on many of their kernels, all their kernel package work was also aimed at upstream.

Small correction...

Posted Aug 24, 2010 15:20 UTC (Tue) by tialaramex (subscriber, #21167) [Link]

I almost agree. I think "all" is stretching it. We have a sysctl set on lots of customer RHEL boxes which appears to say "The VM is kinda broken in this kernel, let's bodge it enough to avoid churning the disks pointlessly". It works very nicely, but I can't imagine anyone even posted it to the LKML, let alone intended it to attain Linus' seal of approval. This work was not, then, "aimed at upstream".

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 4:26 UTC (Tue) by nicooo (guest, #69134) [Link]

> It's a shame that a company like Oracle doesn't understand the power of the open source development model.

It's Sun that didn't understand open source.

> It's strange that Oracle has a totally different approach with Linux, where the company have engineers working upstream and making Linux a better OS.

They aren't losing money with Linux.

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 7:39 UTC (Tue) by trasz (guest, #45786) [Link]

Actually, IBM and HP do have their own Unices, and IBM (no idea how HP-UX looks these days) is investing serious money into developing it and actively pushes it to the clients. What you're describing is a situation from half a decade ago, when IBM claimed that Linux was to replace AIX. Now they are back to AIX again.

OpenSolaris Governing Board resigns

Posted Aug 24, 2010 5:42 UTC (Tue) by alison (subscriber, #63752) [Link]

I'm with cmot: I fear that Android's dominance in the ever more important mobile space means that it may come to compete with Linux as an alternative, non-GPLed offering. As Matthew Garrett's talk at LinuxCon illustrated, hard feelings between the mainline kernel community and Android developers now run fairly deep and the separation may become a divorce, I mean fork. As one tired business model represented by Solaris/HP-UX/AIX peters out, a determined new rival is coming forth.

OpenSolaris Governing Board resigns

Posted Aug 25, 2010 10:27 UTC (Wed) by dbht (guest, #69760) [Link] (1 responses)

Hey guys: How much from the OS market share OpenSolaris has? 1%? 3%? Less than 1%? - This is clearly the reason that Oracle is passing away OpenSolaris. Probably some peaces of the SO will be maintained live but the whole idea about OpenSolaris is not important to Oracle any more... I've never used Solaris/OpenSolaris before, but I`m symphatized with you OpenSolaris.

Davi

OpenSolaris Governing Board resigns

Posted Aug 27, 2010 20:26 UTC (Fri) by ESRI (guest, #52806) [Link]

Oh, far less than 1% I'm sure.

I think the large majority of OpenSolaris contributors were Sun employees anyways...

The real "win" having OpenSolaris around was the work some of the VAR's (maybe that name doesn't really do them justice) like Nexenta were doing on top of Solaris...

But Oracle probably saw that as competition. If there was some huge community and culture outside of paid Sun developers that had grown around OpenSolaris they might have thought differently about things. However, it ended up being more or less Sun developing a lot of cool code that everyone else got to use for free...

Maybe not entirely accurate, but I bet that's how Oracle saw things.