>Obviously, guaranteeing a compatibility baseline for the whole system >isn't practically possible.
Quite; having a "Reference Distro" to develop on would help, though, as it allows customers to determine whether an ISV supplied package works on the reference distro as promised and whether the issue is unique to their distro or not. While "standards conformance tests" are nice, the experience with the Linux Standards Base referenced by Ian in his OGB concall was "millions spend and not much to show for it". You can write standards tests until you are blue in the face, but it does not allow you to give any form of guarantee that applications which pass the tests will actually work. That would also require you to verify that the application only uses bits covered by the standards tests. I think that we'll sooner solve the Halting Problem than that. >What should be made sure is that there's a conformance test where you >can hand out sort of a e-badge, that tells an user that the kernel >hasn't been screwed with custom patches (unlike what every major Linux >distro does). This would be of interest for driver developers. That's still a hard problem to solve and I think that that is not where I would expect divergence to occur first and foremost. I would expect libraries to be missing, be in different locations, have different versions, different "SONAME"s and that sort of thing; different version of GNOME etc. >While I don't expect distro makers doing their own kernel tweaking on >their distros yet, you have to plan ahead so that this conformance thing >is in place for the case OpenSolaris actually takes off like the Sun >management (i.e. JSchwartz, Murdock and cohorts) hopes. I'm not sure Indiana is the proper place for conformance tests or a reference distribution; quite the contrary in fact. Casper _______________________________________________ opensolaris-discuss mailing list [email protected]
