Kacheong Poon writes:
> Now thinking of it, I probably asked the wrong question
> about connectivity and milestone as I misunderstood the idea
> of a milestone. Does reaching a milestone mean that it
> is now guaranteed that certain operation can work? Or does
> it simply mean that some services have started, but it does
> not guarantee that those services work correctly? For
> example, a name service, say NIS, has been started. But it
> does not mean that it can return results as the NIS server
> is not responding.
I think that's an excellent question, as it points out the difference
between networking and (say) launching a door-using server, but I
suspect that the real answer must also depend on what the those using
this milestone require or assume.
There's not much point in defining a milestone that nobody cares
about.
That is what I find to be a difficulty here. By saying that we need
to have some event (leave SMF, milestones, and the like out for the
moment) signify when an external interface is made available, we're
essentially saying that there's some application that:
- cares about this event, because it will not run properly until
that event occurs, and
- can't find out about the event through any conventional means, and
must instead be started and stopped so that it can never "see" the
system when there are no external interfaces.
I doubt that this is true for any application or service. If it is,
then I'd suspect that it's a bug rather than a feature, and I'd very
much like to know about it. Nobody has mentioned such a service yet.
It seems quite likely to me that some people will want to have
"networking applications" (suitably defined) disabled when there's no
external network interface available. That strikes me as just a
degenerate case of the more general issue that people will want to
have a profile specifying what applications should run in any given
situation: when I'm on network A, I use this set; on network B, some
other set; when on no network at all, this third set.
Given that most (perhaps all) applications and services, even sshd and
xntpd, are usable when all you have is loopback, it doesn't seem
reasonable to say that those things have a dependency on external
interfaces and cannot run when there aren't any. That sounds like an
effort to codify some elements of the "widely-understood default
policy" into SMF's dependency mechanism.
I do not believe that these things in fact represent dependencies as
SMF defines and uses them.
--
James Carlson, KISS Network <james.d.carlson at sun.com>
Sun Microsystems / 1 Network Drive 71.232W Vox +1 781 442 2084
MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677
