4,580 Posts served

11,094 Conversations started

How open is OpenSolaris ?

By Ashok Raj (Intel) (5 posts) on January 21, 2009 at 9:25 pm

I have always wondered what makes a successful open source project, especially since i worked extensively on Linux and now working on OpenSolaris. Dave Stewart did a presentation a while back on the same subject in another conference. Iam not an expert about all open source projects in general, but only familier with Linux and Solaris... so we are just speaking kernel only here. Iam not worried about Apache or other application kinds since i really dont have first hand experience, hence i would only speak from my direct experience.

Without a doubt Linux is much more popular and successful, sometimes its a perception but as a developer I beleive that its really true. Its also true that the simple structure and the minimal infrastructure lends itself as a Geeks OS. So here are some of the things that make Linux more appealing to a wide audience.

1. It fits almost any device. From gumstick to really large systems from many OEM's. Linux offers CONFIG_thisthing and CONFIG_thatthing so you can make it really really small to fit any device and memory footprint.
   
2. Linux kernel development is truly done in the open. Very rarely you would see a successful kernel component developed by some body just drop in open and expect it to be included in mainline. I mean the design and development happens in the mailing lists seeking input from community members.
   
3. Remote install is much simpler with Linux, you can just unroll the rpms, or even keep the iso image in ftp host, use remote vnc to complete installs.. you dont need to learn much to get things done.
   

Now iam not complaining that Solaris doesnt have the above features, but its simply hard.

About being open, I dont think we can say for sure about being open the same way Linux kernel developers see the development process. There was an article in CRN trying to access the same topic when OpenSolaris launch was being talked about.

In order for OpenSolaris to be really perceived in a way Linux kernel development is understood, I beleive OpenSolaris is not there completely yet, but making strides in that direction albeit very slow. Opensource is not saying "yes, the code is openly available", but it is really practiced with a true intend to let others participate. So when can we really say OpenSolaris is really open?

1. When you see a project discussion before the one-pager in mailing lists. It appears in many/most cases when you see the one-pager for project proposal, the code is most likely written and just about to be published. This isnt interesting to someone who wants to be fully engaged in the development process to begin with. Now this may not be true of all projects. Some parts of tesla development was atleast discussed and developed actively in open mailing lists such as the PowerTop, Deep C-State management. This surely points to a step in the the right direction.
2. Timeliness in responses, when a project proposal comes in externally, help guide for the project to quickly take off, with the same attension one would give for an internal project. You really want more participation and be engaged, rather than loose the audience.
3. Success is when you see contributors outside of the company being involved heavily in the development process. Well, i dont have any numbers to prove/disprove but certainly one can gather from the commit logs, that there is very little participation in any quotable numbers outside the company.
4. There should many distributions as one could find in the Linux world. There appears a movement happening, not sure if there are just purely pet projects or would eventually become valuable clones such as the Redhat and the Suse's. Its really encouraging to read that there are indeed an attempt to package and distribute opensolaris.

To conclude, I would say more needs to happen to make OpenSolaris really open to attract other developers to contribute to the project.

Should we be more patient? Read Dave's blog here. If it takes too long, people would quit paying attension!

Share the code or Share-The-Control ??? another one from Dave.

Maybe next time i will try to write about something more technical :-)

Categories: Open Source, Uncategorized
Tags: linux, opensource, solaris

For more complete information about compiler optimizations, see our Optimization Notice.

Comments (6)

January 21, 2009 10:23 PM PST Eric Saxe

Nice post. My take (in a nutshell) is that OpenSolaris is more open than it was, and not as open as it will be. Realzing open development means changing the established day-to-day habits (and the mindset) of existing Solaris developers...and from what i've seen, that's slowly but surely been happening.

January 22, 2009 4:03 AM PST Tony Awtrey

Ha! I'm the ISV they quoted in the CRN article you cited. Just surfed in from a Reddit link. Three years on and Sun is still trying to generate interest and support without yielding control. Someday they will figure out you can't "do" Open Source, you have to "be" Open Source. Big difference.

January 22, 2009 4:42 AM PST James Mansion

Well, that's a definition of 'open' that is handy for wannabe OS hackers. But its all but irrelevant for an OS user. The openness in terms of visible roadmap, free access to early builds and the like, is very valuable. An ability to spend a huge amount of time grovelling in kernel code so I can say I can wrote some of it - well, that's not actually very valuable at all and is, if anything, a distraction. Personally I'd rather Sun didn't try to kill itself by being 'a better Linux than Linux'. I'd rather Sun steered the ship and had a planned course that it could execute. If that means that the usual open source 'show me the code' bunfight and random-walk progress can't happen, then that's an advantage as far as I can see. The number of open (by your definition) projects that manage to articulate a roadmap AND deliver to it is small. I suspect its because you get bikeshed discussions on the one hand, and on the other you don't have a tie up of responsibility and authority in project leadership - the authority simply isn't there when you're receiving contributions as charity. Under the cicrumstances I'd rather have the roadmap and give up the 'openness' that you seem to cherish. I know why I want a roadmap. But why do you want your definition of 'open'? What does it give you? (Rather: what does it give you THAT YOU NEED?) I think Sun's problem is that they want the roadmap and the control, but they've (mistakenly, IMO) advertised that they want the community development model. I don't think you can have both.

January 22, 2009 10:51 AM PST Ashok Raj (Intel) Total Points: 430 Registered User

Thanks @Eric Saxe... You are right, its mostly as you pointed out, getting rid of old habits. With Tesla, you are headed in the right direction.. Kudos to you and your team!

January 22, 2009 11:10 AM PST Ashok Raj (Intel) Total Points: 430 Registered User

Good analysis @James Mansion. If there is really no intend to get participation from community, and only for Roadmaps, it would be named OpenRoadmap! My personal thought is that Sun intends to attract real community model, but just unable to shake old habbits. Community development model has some real value, from personal experience, You are right about charitable contributions... and yes the mailing list can be a bit noisy... but thats not what happens always. Linux maintainers are as I call "Benovelant Dictators" - You get more testing for free - Why wouldnt you want some free work from kernel wannabies? Certainly reduces cost of maintaining over the long run. Their wants your gain.. sounds like the adv i see in bad economic times "Their loss your gain" :-) - Increased device driver support, there are more companies contributing directly to Linux kernel. Say for e.g i dont have to hunt one vendor usb->serial to work with. Much simpler when people are directly contributing v.s one company doing it all.

January 23, 2009 1:06 AM PST James Mansion

> Why wouldnt you want some free work from kernel wannabies? Because the cost of managing and QAing the process is high and out-of-the-blue 'contributions' can be disruptive? Because such contributions tend to be made outside of any controlled process and require excessive after-the-fact scrutiny? I don't think the issue is that it is impossible to engage in an OpenSolaris project - just that it IS a project, so it will tend to have a plan and a delivery target and some internal test and documentation resources (I guess - I don't work at Sun) so that the standards are retained - and remember that Sun has a focus on backwards binary compat and de jure standards adherence that Linux et al have never had. This will mean, inevitably, that its going to look a lot more like 'work' and less like 'fun'. Its not a playground, testbed or ego-trip - its work. That is going to be a hurdle for casual hackers. It need not be a hurdle for Intel or AMD - for whom it is also work and it should not be difficult to engage and plan at a level that fits the project integration deadlines. It will be tougher for smaller hardware manufacturers but only in the sense of raising their game. All these participants have financial self-interest at stake and should be prepared to work in a structured way. Ditto large ISVs like Oracle or IBM who want to make the platform a better host for their products. That leaves the knowledgable who scratch their itch with hardware tweaks. Sun really need to have a seperate process for this because many tweaks will regard PCI and USB device IDs and quirk table entries - and there should be a process for accepting those into a fast-track QA since they should be low risk. I wouldn't suggest using a Linux-like dev model at work. Code what I like, perhaps in competition with others. run the gauntlet of egomaniac reviewers and have no guarantee of integration in the end? No thanks. Suggesting it as a methodology in any professional IT organisation would invite ridicule and with good reason. Why should something as technically demanding as an operating system be developed in a shambolic way? Linux makes progress but at huge cost and I think any suggestion that a more measured (dare I say 'professional') approach change to be more like it is deeply suspect. James (Do I use Linux? Oh yes, every day, and I have fince SLS came on a subscription of floppy disks.)

Trackbacks (1)

• Intel Software Network Blogs » Interview: Ashok Raj, Senior Staff Software Engineer for OpenSolaris
  September 17, 2009 9:00 AM PDT

Leave a comment