Hi Wes, On Mon, 2006-08-21 at 13:53 -0500, Wes Felter wrote: > If One TCK is so important, let me propose a semi-heretical idea: don't > open it. Certainly the 400 JVM developers should be able to run the TCK, > but it's not clear that they need to distribute modified versions. (Of > course, they could still submit bug fixes and such to the TCK > maintainers.) I don't think the Linux distros need to distribute the TCK > either, so their "OSI or die" policy doesn't appear to affect the TCK. Of course the Linux distros need it! One of the great strenghts of distributions is their ability to run the testsuite for all packages they build. And to give their users the power to repeat the build process and do their own checking of the results. It is a critical part of quality control. So when you build your deb/rpm/etc (possibly with distro specific patches) you run the testsuite and fail if some tests don't give clean results. That is how GCC for example works. There are even distros that only ship (easily buildable) source code so as to optimize any binary for your personal system. Then it is even more important that the user also gets the testsuites to make sure their special customized build on their own hardware is perfect. And it might very well be that to do that you have to adapt, modify and hack the testsuite to run and build in your particular packaging environment (and run it non-interactively). Don't forget that free software breaks down the artificial barrier between users and developers. Everybody is a core library, runtime, compiler, package, etc hacker now. So everybody should also take the responsibility and run any testsuites. Your users will demand it from you. Cheers, Mark