[arc-discuss] Defining OpenSolaris Consolidations

[James Carlson]

Stephen Hahn writes:
> * James Carlson <james.d.carlson at sun.com> [2007-08-07 10:09]:
> > Not only can you not deliver to the same location as some other
> > consolidation, but (unless you get an architectural exemption) you
> > can't redeliver the same thing or a modified version of the same thing
> > to some distinct location.
> 
>   True, but that's still relatively late enforcement.

"ARC early, ARC often."  Perhaps.

>  I'm not sure what
>   the right answer might be:  is it okay to have multiple similar
>   Consolidations?  How many?  Two?  Five? 

I don't think there's a good independent answer to that question.

The smallest unit here would be the "project" -- the old "home
directory consolidation" approach, where every little thing delivers
directly to the WOS.  That's doable and perhaps even reasonable in
some contexts.  In others, it can be a real mess.

And, arguably, a few of our consolidations are too big for their own
good.

>   A not-immediately-emotional example is hard to find.  How about a
>   full-blown device driver Consolidation, overlapping in charter with ON
>   and X11 (if we stretch)?  Does it matter that each might be soliciting
>   Projects for integrations?  (Are they competing?)  Or do we need to be
>   more specific about what causes the need for a new Consolidation?
>   Historical events show a lot of organizational influences, rather than
>   architectural or even low level build-related reasons.

If the device driver consolidation aims at purely DDI-compliant
things, and it can do so in a way that delivers acceptable quality,
then I think that'd be a fine approach.

The architectural issue here is that you don't want to have zillions
of tiny private interfaces dragged across consolidation boundaries, so
it makes sense to draw those boundaries where you have clear, robust,
and stable interfaces, and try to avoid them where you're cutting
across private bits.

Yes, you can use contracts, but that's just a hack.  It doesn't scale
well and the failure modes are pretty well-known.  And, in fact,
contracts are just a way to _force_ stability that doesn't exist.

To move towards more consolidations, we need more stable interfaces,
and I think that's ultimately a good thing for [Open]Solaris in
general, because it's exactly those boundaries and interfaces that
most affect third-party developers.  For instance, a separate "file
system" consolidation would likely be great news.

-- 
James Carlson, Solaris Networking              <james.d.carlson at sun.com>
Sun Microsystems / 1 Network Drive         71.232W   Vox +1 781 442 2084
MS UBUR02-212 / Burlington MA 01803-2757   42.496N   Fax +1 781 442 1677