Liane Praza wrote:
> I'm not yet sure how this approach will pan out in terms of both
> administrative interface and implementation, but think the best way to
> figure that out is through prototyping and further discussion. A
> project would be an ideal way to allow collaboration in doing that.
>
I agree, I am interested in where this will lead. I also think SMF
is the right area for the project. But, I think the filesystem mount
dependencies are examples of a more general problem with the large grain
and highly static service dependency relationships which lead to trouble
in other areas.
A recent example in another area is shared instance zones with
associated routes. If we had a more dynamic model, I think each zone
would be a global zone service which in turn satisfy their corresponding
logical interfaces which could in turn satisfy the dependencies of other
"micro" services like a newly appropriate route.
I think if the project demonstrates a means to dynamically create
system configuration specific "micro" services/dependencies from the
systems vfstab (or its equivalence) then it could create the framework
to better model configuration specific dependencies in other areas as
well. It could then also lead to better modeling of cross dependencies
between filesystem mounting and other areas in the future.
-Will
