On Wed, Apr 12, 2006 at 08:03:54PM -0400, Bill Rushmore wrote: > I see no reason why the community needs yet another project. We have two > good places to get what we need what could a new community here provide > besides usurping what has been done already? Again, I am not saying this > to be arguementive, I am just still not getting why we need this.
No one is 'usurping' anything. This project team is not demanding that Blastwave be disbanded. Nor are we intending to go off and reimplement exactly what others have done and call it the Official OpenSolaris Software Distribution And All Others Suck And Must Not Be Used Because Sun Says So. That name would be way too long and does not have a convenient abbreviation (you can laugh here; that's not the real reason). The Companion has a few advantages over Blastwave. One of the most important is that we do not duplicate software which is already in current builds of Solaris (put another way, it's in another OpenSolaris consolidation). This comes at a price - we don't support older releases. Neither Blastwave nor Sunfreeware has solved this problem, and it's quite a serious one, especially where libraries are concerned. James touched on this issue in one of his earlier messages. One of the things we'd like to do with the Companion is find a really good way to gracefully retire components once they're integrated into another consolidation, while still providing the necessary functionality to users of older OS releases. Another advantage, while not a technical one, is that some Solaris customers are more comfortable using the Companion than software obtained elsewhere. I'll be the first to admit that there's no good reason - Sun provides no support whatever for this product. Nevertheless, whatever we produce will see the widest usage if it's considered "Sun-blessed." The Companion also offers us a way to start testing some of the processes and tools needed by the rest of OpenSolaris. Because the process around the Companion is much lighter-weight than that used by most other consolidations, we'll be able to start simple and add more tools and process steps when other consolidations are ready to use them. Ultimately, the processes and tools we'll need for ON and other large consolidations will have to scale far better, and offer more stringent quality checks, than Blastwave's current processes can. That means it will be possible to deliver a better product with less work. I know Dennis and the many package maintainers spend tens of thousands of hours every year in support of Blastwave. With better tools - tools which they won't have to build and maintain themselves and which will in time become familiar to and comfortable for all OpenSolaris contributors - much of that time might be saved. The same could be said for Sunfreeware. Certainly Blastwave is way ahead of us in terms of distribution - if you look at the Sun pages for the Companion, they're laughable, especially dependency management (or lack thereof). I'm not convinced that pkg-get is the best solution to this problem, despite its readily apparent efficacy. In many ways it seems that pkg-get exists mainly because the right fix would have required changes to the packaging tools themselves, and the source was simply not available at the time. With that problem addressed, this issue deserves a second look. But we all recognise that the future of third-party software distribution is a network-enabled one, and it's embarrassing that Sun has done so little to move in that direction with the Companion. Even in the worst case, Sun will need to continue to offer something like the Companion. If we can't work together on this project with contributors to other efforts, there's little chance that management would desire to co-bundle the product of one of those other efforts, either. Even if that unfortunate outcome results here, it's still preferable that the Companion be developed openly. Thus, this project is still needed; projects are the basic containers for work. But more hopefully, we invite and encourage the Blastwave team and any other interested parties to participate - this is not a reimplementation or usurpation but rather an exercise in improvement and unification. One thing I think everyone will agree with is that users benefit greatly from having a single place to get the latest software, linked consistently and in a way that avoids both duplication and conflicts. In the best case, this project will deliver on this goal. We're a long way from there if the bitterness in this thread is any indication. So let me make clear that there are no preconceived notions about how the endgame is to be reached. If everyone agrees that Blastwave is ahead of the Companion in every conceivable way and that every solution chosen by Blastwave is clearly correct, perhaps this project becomes a thin wrapper around the existing Blastwave effort, providing the Sun stamp of approval that some Solaris users demand and defraying some of the costs associated with that effort by providing centralised infrastructure as well as top-notch development process and tools to ease maintenance and improve scalability. Maybe that's the way this all works out. And if the reasons are right, I'll be very pleased by that outcome. But it's not yet clear that this is the most desirable result. There are a lot of tough technical challenges between where we are and where we want to be, and I'd like to see a collaborative effort to get past them. That's why this project exists. If you'd like to help, feel free to join the newly-created companion-discuss mailing list. -- Keith M Wesolowski "Sir, we're surrounded!" Solaris Kernel Team "Excellent; we can attack in any direction!" _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
