Re: Implementation questions

[dsiinc![e-mail]]

>I am starting to implement the new fddi draft MIB.  I have a few questions
>and would appreciate any assistance.
>
>Is there any requirement/advantage to keep the MAC, Path and Port
>indices unique across all SMTs or is it sufficient to keep them unique
>with a given SMT?

According to our definition, SNMP indices must match the SMT indices.  
For instance, the definition for fddimibMACIndex states:

	  "Index variable for uniquely identifying the MAC
	  object instances, which is the same as the
	  corresponding resource index in SMT."

In other words, SNMP should use the same indices as SMT uses.
Now, SMT makes no restrictions on numbering MACs and PORTs, but 
PATH indices are restricted as follows:

    Primary PATH must be 1,
    Secondary PATH must be 2,
    Local PATHs must be 3-255.

Thus, you're limited by the indexing scheme used by your SMT.  Our SMT
makes indices unique only within a SMT instance, so multiple SMT 
instances reuse the same MAC, PORT and PATH numbers.

>The fddimibPATHIndex is defined as having a value from 3 to 255.  Do 
>these values have any special meaning or do they just start at 3 instead
>of 1 and get assigned as appropriate to the implementation?

See discussion above.  PRIMARY == 1, SECONDARY == 2, LOCAL == 3-255

>The fddimibSMTNumber, fddimibMACNumber... all indicate that they are static
>for a given initialization of the management system.  Is this being changed
>along the lines of the new ifNumber to allow them to change when a new
>card is swapped in/out?

It doesn't really matter what we put here.  We concluded that an SNMP
manager or agent can't ever really be sure of detecting equipment that 
is hot-swapped anyway.  The SMT MIB only guarantees integrity across 
what it considers "manageability of the object."

In other words, SMT initialization, power cycle, self test, or
transitions of the "hardware present" variables (for MACs and PORTs) all 
lead to discontinuities of the SMT MIB.  Since SNMP relies on SMT 
to get information, SNMP can't promise any better integrity than this.
By SMT's definition, I interpret insertion/deletion of a hot-swapped 
board to be an "initialization" of the management system, so I'm free
to renumber everything I want during a hot-swap.

Since there is no way for the SNMP agent to detect a hot-swap, what 
difference does it make?  Any time you read any two variables, they may
be inconsistent with each other due to hardware transitions.  Why is
renumbering any different?

Ron Mackey			Distributed Systems International, Inc.
[e-mail]			531 W. Roosevelt Road, Suite 2
708-665-4639			Wheaton, IL 60187-5057