Re: Interface configuration

[Bob Stewart <[e-mail]>]

>I orginally started this with a proposal for an additional table.  This table
>would be indexed by segment and interface.  If what I'm hearing is that
>lack knowledge of what interfaces are connected where then doesn't this table
>answer that problem?


I agree that a network manager must know what interfaces are connected to what
segments.  Keith and I further agreed that such knowledge needs to extend to
bridge and repeater ports, which may not be represented as (MIB-II) interfaces
at all, thus Keith's suggestion for a type.  Without this information, the
proposed table doesn't address enough of the problem.  Further, I believe the
proposed table doesn't work because segment and interface index are not
sufficient table indexes.  Entity index must be included, as interface index
numbers are not unique across entities.

Now, although I agreed that the information is necessary, I'm contending that
the Chassis MIB may not be the right place for it, on the basis that the
contents of interface tables and mapping of ports is not only somewhat
implementation-specific, the information may not be available to the Chassis
MIB agent.  At best, such an agent should obtain the information dynamically,
probably through some out-of-band mechanism.  It would not be good for that
out-of-band mechanism to be a compiler, forcing the chassis agent to have
built-in intimate knowledge of software internals of other entities.

I may have overstated the case a bit to make the concern clear, but I believe
this is a legitimate concern.  If we are to put details of segment mapping in
the Chassis MIB, we must do that in a way that is independent of
implementation details for all types of entities, including those that do not
map segments to MIB-II interfaces.  And we must believe that such detail is
reasonably available to the chassis agent for the hardware implementations we
expect to support the Chassis MIB.

	Bob