[oi-dev] OpenIndiana and illumos

[Alasdair Lumsden]

On 18 Nov 2010, at 01:34, Garrett D'Amore wrote:

> So, I have been having "reservations" about OpenIndiana and illumos
> interaction.
> 
> OI is still working on a b148 release, based on Oracle's code base.  Yet
> SX11 based on b151 is already out.  (I'd argue that OI is "too little
> too late".)

148 was pretty much complete months ago but unfortunately nobody on the team (*ahem* Albert Lee *ahem*) has actually performed the release itself.

g11n was holding the release up, however when the question of 148 was raised on IRC I stated "Just release it the binary g11n as we did for 147". However it still didn't happen. I delegated release engineering to Albert and I expected him to take responsibility for the releases. I have deliberately *not* pestered or badgered him because I cannot badger and pester someone every month for releases, and I was hoping the pressure would mount to the point something would happen.

However we're all still waiting.

Unfortunately I have also been absent myself for the past month due to relentless work commitments and an enormous change in personal circumstances which have left me with virtually no time to work on OI. I knew I had buckets of free time during the OI bootstrap phase and I did my best to locate developers and put in places the infrastructure and people needed for the project to reach critical mass and "take off", and for release engineering to be handled by someone else, because I knew I wouldn't have time later on.

But quite frankly, OI has not yet reached that critical mass, and is right now dead in the water unless we can recruit committed community volunteers, or find a commercial sponsor with deep pockets.

I will hold my hand up and take responsibility for this failure, because I effectively started the project. If I need to be replaced for the project to move on, then so be it. However not a single other person on the project has expressed interest in managing things, and it has been incredibly hard to find developers and people interested in contributing. Heck, even the existing contributors were highly reluctant to put their names down for positions on our "Human Resources" page and I had to bully the ones we do have to do so. Hence, the HR page is laughably incomplete:

http://wiki.openindiana.org/oi/Human+Resources

It has been an uphill struggle all the way.

> illumos needs changes to SFW, and other supporting consolidations.  It
> needs a cooperative relationship with a primary distribution.

We want that too! Why have none of your team built and delivered an Illumos repo to us to use? Why has nobody tried to help us with getting Illumos into OI? You think we have time? We barely have enough resources to build something that was designed to all build together, let alone integrate something that has had its major subsystems reworked to hell and back.

> However, because of the insistence of continuing to "follow" Oracle
> rather to proactively lead, it seems that progress towards any of
> illumos' goals is not something I can reasonably expect from OI in the
> near, and possibly even mid-term, future.

Completely incorrect and off the mark.

We want OI to be a leading distribution based on the Illumos kernel. That has always been the goal. However if we fork all of the userland software right now, we'll fall even further behind Oracle, and then the project is dead. Oracle has hundreds of staff members working on pkg/JDS/SFW/XNV/etc - we have what, 6, maybe 8 part time developers volunteering?

Until the project grows and attracts more developers, we have to build on top of what Oracle chucks over the fence. This is a "commercial reality". It makes sense. We already have plans on how to extend and enhance what Oracle provide, Guido laid it all out here:

http://wiki.openindiana.org/oi/OpenIndiana+Addon+Consolidations

This lets us obsolete, replace and extend what Oracle provide, so that we will "lead" rather than "follow". However to ignore/fork away from the code Oracle are continuing to contribute would be suicide.


> OI itself is understaffed.  I can help with that.  But I refuse to
> contribute resources towards a solution that doesn't at least indirectly
> lead towards my goals of a usable illumos-based distribution.  Burning
> my resources on a Solaris 11 clone that doesn't at least support my own
> kernel bits is not economically sound.

Garrett, where are you getting the idea from that we don't want to use Illumos? This is just insane!

We haven't integrated Illumos because nobody on the team knows how or has had time to investigate. What have you done to help? Nobody on your team has even delivered a tar of a depot with the latest build of Illumos to us yet, despite building the thing all the god damn time. How hard would that have been?

Rather than complaining, contribute! You even know we are understaffed! You said it right there!

We intend to integrate Illumos at our next build, b150. This intention was set in stone months ago. 148 was meant to be out months ago too though.

> In my opinion, its time to just abandon the OI 148 "effort".  It will be
> too little, too late, following on the heels of SX11.  There is
> absolutely no point in a "new" release of an OpenSolaris clone that is
> older than what Oracle itself ships.

148 is pretty much all built anyway, and the main point of it was to address some of the main bugs that hit oi_147, for example the branding errors and some release issues.

So there is still a point in releasing it. It's pretty much completely done, someone just has to assemble the bits with distro_import and then distro_const:

http://wiki.openindiana.org/oi/oi_148

The person that did it last for us was Albert Lee and he has the expertise in doing this. Nobody else on the team does.

> The *right* solution going forward is to move efforts towards a
> self-hosting distribution based on illumos, working on the *future*.
> Perhaps taking a number of months to arrive at a complete product, but
> doing so in a way that is collaborative with the illumos developers.
> (Right now there is a huge disconnect with the illumos-gate team and the
> OI team.)

We would love that - how would you suggest we begin the process?

> So, my question to the OI team leads: what are you going to do?  Are you
> going to try to continue to follow Oracle's lead?
> 
> Or are you interesting in *innovating* and joining a team that is intent
> on standing on its own and cutting the dependencies from Oracle?
> 
> If you're not going to do the latter, then I need to know -- now, so
> that I can start staffing my own effort, starting first with proper
> "forks" of the other key consolidations.

The intention has always been to innovate, to use Illumos, to collaborate with your team and to produce an amazing compelling operating system with free security and bug fixes on a stable branch, with a regularly updated development branch. We would love to end our dependence upon Oracle.

But as you have observed, we're chronically under-staffed!

The fact is that none of this is about software engineering. Building a successful operating system distribution is about building a community. Fostering an environment which attracts developers who want to contribute their time to the project. Inflammatory mailing list posts like this one from you, which attacks the project and is full of language that suggests you're bitter and angry, just annoys people and puts people off. You are a brilliant engineer, but you are not going to foster collaboration and innovation shouting and ranting at a bunch of volunteers.

Anyway, moving forward, we want to work with you. We just need to on-board more developers and solve the release engineering issue, and onboard some "project leads" who can spend time locating developers, helping on-board people, etc. Because I unfortunately don't have time to do this.

The biggest hurdle right now is release engineering. Just yesterday I made an informal job offer to a member of the OI team whom I am hoping can spend at least some of his time at work doing OI release engineering to keep things going. This will cost my business money but I am happy to contribute because OI must succeed. If I had more money I would donate more resources but the fact is we're in a shitty recession right now and money doesn't grow on trees, and neither do developers.

> If you do choose the latter, then I can immediately start helping you
> with some of your challenges surrounding improved integration of g11n
> and sfw, and other bits.  There are many other opportunities once we
> agree to have reasonably synchronized/coordinated code bases between the
> underlying illumos foundation and the other consolidations.

That would be awesome. We want very much to work with you. The Belenix guys have also reached out to us and are interested in contributing and working with us to keep the userland software rolling.

If we can all work together then the future should be very bright. We just need more developers!

> And to those of you running OI "in production", you're nuts.  OI has had
> exactly "one" release (of dubious quality), and has a sum total of a
> half dozen engineers behind it; and I'm not sure all of them are
> actually software engineers and not just integrators.  I sure as heck
> don't advocate illumos for production use *yet*, even though my own
> desktop runs it.  (Although I'm working hard to get it into a condition
> when I will be happy to do so.)

We're a rabble of inexperienced volunteers who stepped up to do this because nobody else was going to. We all want OI to succeed. We do need real, hardened software engineers with experience and time to contribute and join the project, to lead and to teach others. People *are* slowly joining the project, we just need to get the momentum to pick up a bit.

My absence hasn't helped, and I apologise for that. But I have a mortgage to pay, a business to run, etc etc. Volunteering some staff time should hopefully help a lot, but we need more developers developers developers. Oh dear, I'm getting visions of Steve Balmer bouncing up and down on stage with sweat pouring off him.

We would love your help and we are more than happy to do what you suggest as long as it is sane and reasonable, and so far nothing you've said is unreasonable.

Alasdair.