On Tue, Jun 09, 2009 at 03:38:46PM -0700, Bart Smaalders wrote:
> Nicolas Williams wrote:
> >On Tue, Jun 09, 2009 at 11:52:14PM +0200, Darren Reed wrote:
> >>One might ask the very same question of using SMF for that task.
> >>How can SMF assemble a file it has never seen before?
> >
> >The pkg author provides the start method that knows what to do.
> >
> >>Why can we teach SMF to do this but not pkg?
> >
> >IPS could have it's own SMF-like actuator concept.  It'd be replicating
> >much of the same functionality that's already in SMF.  I think that'd be
> >cleaner, actually, but it'd also require lots more work.
> 
> This would imply that anyone adding new stuff to be merged
> would have to add code to the packaging system, and the packaging system
> would have be upgraded often.

Well, no.  What I had in mind was a generalization of that SMF service I
prototyped for SVR4 pkg scripting -> IPS migration.  Basically: a single
SMF service to run all boot time self-assembly.  Each pkg would deliver
its self-assembly "jobs" (instead of self-assembly SMF services) via an
IPS action that does the Right Thing (meaning: queue up the jobs,
possibly by dropping them into a directory where they will be found by
the IPS service that runs them).

The nice thing about such an approach is this: the self-assembly
actuator is not OS-specific, so IPS packaging can be less OS-specific.

The problem with that approach is that you then need to implement an
interface by which to see the status of each such job, and which can
handle dependencies -- something SMF already has, so this would be code
duplication, to a point.

Nico
-- 
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to