Sunday Feb 17, 2008

A Chuckle?

I left a comment on Micheal Dolan's blog -- Emily Ratliff covers Roy Fielding’s Departure from OpenSolaris -- the other day, but it has not made it out of moderation yet. Or, the comment just didn't take. Not sure. I didn't see one of those warnings saying my comment is being moderated, but when I tried again it said I had already submitted that comment. But no comment. So, I gave up. I'll just post it here instead.

Hello, Michael.

It's sad that you get a "chuckle" about some of this stuff. Many people here are trying hard to make the OpenSolaris project work (under difficult circumstances sometimes), and they don't deserve your sarcasm.

The request-sponsor program you laugh at was never intended to scale as a general integration program. Everyone involved in OpenSolaris knows that. It was always designed as a temporary solution, but since we are clearly behind in  moving the main ON gate and some of the integration tools outside, the request-sponsor program has had to fill the void. And yes, it's quite backed up. Everyone knows that, too. Emily's quote you cite is somewhat misleading, though, because it assumes that the Linux and OpenSolaris processes are the same (or at least similar enough to compare). They are not. Linux is an open development community that has been open for a very long time. It started small and grew large in the open over many years. OpenSolaris, however, is still very much emerging and trying to open from the inside of a company. We started big (in terms of size of code base and number of engineers and processes, etc), and are trying to open in stages in a responsible way since Solaris is one of the core products of the company. Also, it has proved to be a complex engineering, legal, business, customer, marketing, and cultural task of moving such a large operation outside and teasing apart the development process from the productization process while we simultaneously build, ship, and support a product. Linux never had to do what we are doing, so the issues are extremely different. A better way to look at it would be to compare the opening of Solaris with, say, the (potential) opening of HP-UX and/or AIX if those systems are ever opened. I can only imagine that the issues would be pretty similar given the size and age of those code bases, the complexity of their development methodologies, and how important they are to their respective companies. But regardless, we are certainly far behind where we'd like to be at this point. If you'd like to follow the SCM migration project, you can check in with one of my colleagues here: Mike has some pointers in that blog that will take you to the project pages on

Also please keep in mind that the request-sponsor program is only one part of the OpenSolaris project, and it's only one way people are contributing (and it's only about code, whereas people contribute in non-code ways as well). And it largely represents integrations to the ON consolidation, not all other consolidations and development projects on the site. If you wanted to really compare OpenSolaris and Linux from an engineering perspective in terms of the flow of code being integrated into the kernel, you'd have to look inside Sun for that since the ON consolidation gate is not opened yet. Then the data points would be quite different, and the "teeth brushing" analogy may not hold up so well. But one really can't compare the two systems that way, of course, since our development process are still largely behind the firewall. That will change over time, though. Currently, we are more of an open source project, and we are slowly (very slowly) moving to an open development project as well. In fact, some consolidation gates are already outside, and there are already external contributors.

And there is no need to be "irked" about the user group example you cite, either. I left a comment on Emily's blog about this. There are no NDAs on OpenSolaris. The issue there was the building security procedures since the meeting was at a Sun site. Sometimes it's cumbersome holding community meetings at corporate facilities for just this reason. Many groups that are run by Sun employees deal with this ok now, but others choose other facilities, and other groups are led by non-Sun employees at universities or other companies. It's not a big deal, and it in no way reflects poorly on those people involved. If you feel we are "messed up" in our approach to building a community and feel you can help, we'd be more than happy for you to get involved. We are still learning and trying new things all the time. However, uninformed opinions expressed from the outside without any direct participation have relatively little value. So, I encourage you to subscribe to advocacy-discuss here since that is the Community Group that is home to the OpenSolaris User Groups. That's a good place to start since we are kicking around some ideas for UG community-building in two recent threads here and here so you may want to participate in those discussions.

Finally, when you say that you should have applied for a job at Sun to fix this "mess" that, too, is hurtful to the people working on the project, most of who are honorable people just trying to do a good job. We are a small and young community and we are doing our best given the circumstances in which we find ourselves. All projects in all industries in all regions experience rough patches in their life cycle. That is clearly the case with us in some areas. However, there is a great deal that is going extremely well on OpenSolaris, too, and we can't lost sight of that. That's what I keep telling people in the community, and many people agree and are willing to do the work necessary to improve things. And in terms of a job, you'd have to check with Sun HR on that. But you don't need a job at Sun to participate in OpenSolaris, though. Just jump in. But if you think you can "fix it myself" I'd humbly suggest that you will have to talk to many thousands of developers, many of who are Sun's top engineers, to earn consensus for your ideas. OpenSolaris has a distributed leadership model, so no one individual gets to run things. We certainly have a long way to go, though, so please feel free to contribute. All the lists are here and that's the fastest way to talk directly with the kernel engineers.



« July 2016

No bookmarks in folder