Nicolas Williams writes:
> On Mon, Sep 14, 2009 at 04:19:17PM -0600, Tom Whitten wrote:
> > Nicolas Williams writes:
> > > On Wed, Sep 09, 2009 at 01:59:48PM -0600, Tom Whitten wrote:
> > > > The proposed SMF Early Manifest Import (EMI) project will solve these
> > > > problems. startd will run EMI before any services are started, and EMI
> > > > will import manifests and cleanup services that have been removed. EMI
> > > > will allow a new location for manifest and profile files in /etc/svc.
> > > > Since this is part of the root filesystem, it will be available in the
> > > > earliest stages of the boot process.
> > >
> > > So all manifest delivery will move from /var/svc to /etc/svc? Or just
> > > some (e.g., those for services that don't depend on filesystem-minimal
> > > and manifest-import)?
> >
> > Services that require the advantages of early manifest import will be
> > required to move their manifests to /etc/svc. /var/svc will still be
> > supported, but service developers will be encouraged to move their
> > manifests over time. We plan to move the ON manifests, but 3rd party
> > services and non-ON cosolidations will probably continue to deliver
> > manifest into /var/svc for some time to come.
>
> Could SMF assume that any service whose manifest was imported from
> /var/svc necessarily has a dependency on filesystem-minimal? Or would
> that be too risky/disruptive?
I'm not comfortable with the idea of implicit dependencies. I think that
they should be specifically stated in the manifest. A manifest could be in
/var/svc for the simple reason that it hasn't been moved yet.
> Will all Solaris services be fixed by
> either moving their manifests to /etc/svc or by having their
> dependencies updated to record one on filesystem-minimal?
I'm not sure that I understand your question.
[SNIP]
>
> Yay!
>
> Nico
> --
