Marilyn Shoemaker wrote:
We plan to push development for the Solaris Companion out into the community,
Why the Companion?
... It's supplemental to Solaris and consequently not subject to reviews by
internal architecture committees,
Ummm,
While the end result may be the same (no explicit ARC reviews needed), this
statement
does not reflect the correct reason why it is so.
The "Companion CD" is a consolidation in the same way that OpenSolaris/ON is
one.
As distinct consolidations, each can (and does) have its own rules and
behaviors.
ON has chosen stability as their prime constraint, leading to the discussions we
had earlier about "no major releases". The implication is that incompatible
change
is strongly discouraged, if not forbidden.
The CCD has chosen to "always do major releases", but historically has only done
them in conjunction with a Solaris release so that consumers can plan for "flag
days"
rather than being surprised by them.
In addition, all of the component "interfaces" found in the CCD consolidation
are
classified as "changeable in incompatible ways in a micro release of the
consolidation".
(The label in use at the time was "External", the interface taxonomy is in the
process
of being updated to change it to "Volatile").
The combination of a release vehicle that does major releases and a component
that
allows change at any time results in an environment that
1) allows all types of change to the contents of the CCD consolidation
to be "self reviewed" because they - by definition - have no
architectural
impact, and
2) effectively prohibits any other consolidation that has stricter
stability requirements from importing or using the CCD's interfaces.
Thus, the reason that the reviews are all Architectural SelfReview is because
the architecture of the component allows it, not because it is somehow
"supplemental to Solaris".
-John
_______________________________________________
opensolaris-discuss mailing list
[email protected]
