[indiana-discuss] A different take

Eric J. Ray Eric.Ray at sun.com
Mon Jul 2 14:34:23 PDT 2007


So, all of the current discussions notwithstanding, let's think
about a different approach to Indiana (or call it Ohio, or
Illinois--whatever). Consider that we're really trying to build
a base for others to take and leverage, and we've clearly demonstrated
that everything is subject to a religious war, let's regard
it differently. Any distribution could take Indiana, in some form,
and arbitrarily build on it to have an OpenSolaris distribution
that meets specific needs.

Indiana Core could be the absolute minimal functional system.
It would be a subset of ON plus a subset of Install--the bits
required to successfully boot to a console/text interface and to
upgrade the system from there. In Solaris terms, this would likely
define the list of packages that constitutes a minimal supported
system, plus a core set of drivers.

Indiana Basic could be Indiana Core, plus the rest of ON,
plus more of Install, plus X, plus some reasonably friendly WM/Desktop
setup. Basically, a minimal usable system, to let someone
bootstrap themselves into their happy place. (E.g.,
# indiana-get -repository=gnu-bleeding-edge update-all or the
GUI equivalent)

With these pieces, a distro builder could start with Core and add
pieces, or, if appropriate, start with Basic and add to it. Someone
looking to embed OpenSolaris into an appliance would likely just
use with Core. Someone interested in a single-function, easily
controllable device (like a VMware-type appliance) might use
Basic.

I've been talking with a couple of engineers about what Dave Miner
termed a "distribution constructor", which could be a toolset to help
anyone take the Core/Basic pieces and add components as needed,
then easily end up with a functional set of bits that actually could
be installed and used. It's something that we're really serious
about at least investigating, and current plans include staffing/
funding for it. (Yes, it'd be open.)

An approach like this is starting to make sense to me because it's
clear that we cannot meet everyone's needs, and the next best
thing is to provide the tools to EASILY allow OpenSolaris distros
to meet those various needs. A side effect, of course, is that
various distros based off the same Core are very likely to be
compatible with each other, modulo various package dependencies.

Thoughts?


-- 

Eric J. Ray
Software Engineering Manager
Solaris Install
Sun Microsystems
303-223-7843 (direct)/x81067
eric.ray at sun.com



More information about the indiana-discuss mailing list