On Nov 13, 2007, at 12:28, James Carlson wrote: > Simon Phipps writes: >> There's a third view that looks beyond our needs in this community. >> View #1 is the "toolkit for co-developers" view, view #2 is the "one >> distro is king" view; both focus on us in one way or another. View #3 >> is the "applications work here" view. What ultimately will matter to >> end users, after initial user experience, is having confidence that >> the applications they want to use will work on the distro they are >> selecting. What will matter to ISVs is that there is a stable OS they >> can test against or even certify to. > > I think that when you start adding in the complexities of "middleware" > (libraries and the like), "View #3" actually devolves into "View #2," > or the concept of separate distributions itself dies. > > The problem is that in many cases you have to choose one set of tools > over another as a core -- is it KDE or GNOME? are we using Tomcat or > SunOneWhateverIt'sCalled? IPS or SysV? OpenSSL or PKCS#11? -- and > there are a large number of these choices throughout the system. In > addition, it means that you maximize the amount of content. You have > to carry every library that might be used by some application. > > Each one of those decisions that we make further constrains what other > distributions may do, if we assume that "applications work here" has > any value at all.
While I agree with you to a certain extent, I then look elsewhere - Ubuntu specifically - and see a range of distros[1] where an end-user can be confident that the programs in the repositories linked by default will run. The idea of multiple distributions is not dead there despite use of a "common core" that targets it-just-works simplicity. That sort of it-just-works confidence is what drives adoption in today's market and it's the confidence I believe we should target in creating the concept of a "reference" for OpenSolaris. > > Thus, I don't think that "View #3" works as a separate view for the > long term. Instead, what I'd like to see happen (and what I think > John Plocher was hinting at) would be to break Indiana into two > distributions: one that's the community-recognized "must carry" set of > bits for all other distributions. That's the "View #1" bits. It may > be hard for _some_ application developers to target, but others with > minimal requirements can use it. While it will be useful to designate a set of "must carry componentry" like this, I am not sure it's a construct worth exposing outside the co-developer community. I'd probably avoid calling it a "distribution" as a consequence, since deployer-developers tend to associate that with "ready-to-deploy" code. > The second distribution would be the > "Sun OpenSolaris" one, with a rich set of optional components. While your approach has merit, it takes as read there There Can Only Be One in the "user-ready OS" category. I am not yet convinced that is the case. > If other distributions follow that second target, then perhaps we end > up with that golden "View #3." I suspect, though, that if it were to > happen, then we'd have essentially zero difference between > distributions, and no reason to choose any. The others go away in > that scenario. On the contrary, Ubuntu is seeing derivative distributions flourish, without fragmenting the application space significantly. S. [1] http://www.ubuntu.com/products/whatisubuntu/derivatives
