On Fri, 2007-11-02 at 12:40 -0700, John Plocher wrote:
> One of my comments on the wiki definition was along the lines of:
>
> We could, as a starting place for defining compatibility,
> simply assert that there is a baseline (installer and
> a set of versioned packages; a "recipe", if you will)
> that must exist in any distro if it wants to claim
> compatibility.
>
> Of course, this type of definition is poor, from many perspectives.
> It is, however, easy to implement :-)
This may in fact be close to the right starting place.
At the higher levels I'd want to see a goal of:
"every piece of your distribution built from opensolaris sources should
be independently reproduceable from source, and changes between the base
opensolaris sources and the sources used in the build must be clearly
identified".
If the source changes can be identified, then an expert user can make an
independant judgement about whether the changes are sufficiently
compatible for their purposes.
we need a tight definition of "reproduceable" that excludes things that
are expected to change from build to build (such as elf timestamps and
elfsign signatures).
At the lower levels ("built with OpenSolaris") it should suffice to
identify the specific versions of opensolaris sources which were used to
build the product.
> It presumes significant sameness between distros - not a bad thing
> from a compatibility perspective. Same installer, same kernel, same
> packaging system, same repositories...
Given the current state of pkg, I think it's premature to require
everyone use the same packaging system. What really matters for binary
compatibility is what the packaging system delivers onto the running
system.
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org
