>I feel maybe we're getting closer to the issue. I think a lot of >people are put off by the perceived complexity of submitting code to >OpenSolaris. I also think that not compromising on quality is a good >goal, but that you have to understand that the community only >produces Very Good Code only through an iterative process. I think it >has to start with "barely works", proceed to "works in most cases", >on to "works all the time on all platforms", followed by "we made it >relatively fast" to finally "it's screaming fast". I think that >expecting one person to follow through from start to finish by >himself is too harsh, and only a few very motivated people will do >it. I think you need to look at something like Debian, maybe having a >unstable release, where everybody and their brother can go check in >code that may break stuff, and then the community members can come in >and find bugs, talk about them, fix them, and encourage/support each >other. That would, I think, really chafe the Sun Engineering Way, but >I am fairly confident it would get more people involved, regardless >of license.
At Sun we have no tolerance for "works some of the time" or even "works in most cases" code. At least not in the main branch. This is why we have "projects" which are by and large experimental branches. In the main branch, I can't see any reason for code which does not work. When we created the development process, we very much went for a "cut to size" approach; the bigger the project the bigger the overhead; consider it a flat rate tax. ksh93 is huge as a project and so it pays a large tax; changes limited to a few lines pay a much smaller tax (though there's a minimum fee, I would gather) >As far as the patent/NDA encumbered libs, can could OpenSolaris >Unstable even run without them, or with empty/minimal functionality >wrappers? I'm asking because someone will want to compile code >locally and see if it even runs before sending in diffs. You will need and can have the binary. >Also, I would look at allowing people to check in either x86 or sparc >code, letting the community work on the other. As someone else >mentioned, I imagine very few people will have both types of >machines. I think there's supposed to be a test pool in future. Casper _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
