Stephen,
Thanks for your comments on the ARC materials. You raised a number of
interesting questions that I would like to address. I know that Tony has
talked to you offline, but I wanted to follow up on the alias.
A central issue to many of your questions is Early Manifest Import's status
as a service. In the the design document early-manifest-import was not a
service, because we were trying to solve two problems:
1. If a system is upgraded to the EMI code, there will be no
representation of the EMI service in the repository when the
system boots.
2. We needed to insure that EMI gets run before any other services
are started.
As we were talking about your mail, however, Sean Wilcox and Tony Nguyen
came up with a way to solve these problems and also allow EMI to be a
transient service. We have this coded now, and for the most part it
works. I have a little bit more work to do on svcadm disable. We'll be
removing the term pseudo-service and references to the specialness of the
service from the document.
Your suggestion to use the value of SMF_FMRI rather than -e to determine
whether the script is running as early or late manifest import is a good
suggestion.
It has never been a goal of this project to use service/instance properties
to specify which directories should be scanned for manifests. It is a good
idea, though, and I will file an RFE for the work.
You asked "How do the Xincluded profiles get handled?" Currently, there do
not appear to be any included profiles outside of the ON consolidation. So
it's probably safe to move the location of profiles without the need for
cross-consolidation coordination. Once EMI is integrated, other
consolidations will deliver new included profiles into the new location.
Thus, it should not be necessary to support generic and platform profiles
in both locations.
We are looking at your suggestion to make early manifest import be an
instance of the manifest-import service.
mfstscan still exists. We use it in the manifest-import script to
determine if we need to do cleanup work. We also occasionally use it for
debugging purposes.
As to the other suggestions that you made for clarifying the document, we
will include those clarifications. Thanks for your help.
tom
