AD Evaluation requested changes

"David B Harrington" <[email protected]> Thu, 21 Jul 2005 14:27:55 -0400
Newsgroups gmane.ietf.bridge
Message-ID <[email protected]>
Hi,

During AD Evaluation, Bert requested that all read-write objects
include persistence expectations in the descriptions.

During the IEEE 802.1 WG meeting this week, they discussed the objects
and decided that all the read-write objects except
dot1dStpPortProtocolMigration should be persistent, and will provide
clarifications to the relevant IEEE standards.

I suggest the following changes to the documents:

1) The RSTP-MIB read-write objects dot1dStpVersion,
dot1dStpTxHoldCount,  
dot1dStpPortAdminEdgePort, dot1dStpPortAdminPointToPoint, and 
dot1dStpPortAdminPathCost should include the statement "The value of
this object MUST be retained across reinitializations of the
management system."

2) The new P-BRIDGE-MIB and Q-BRIDGE-MIB objects
dot1dPortRestrictedGroupRegistration and
dot1qPortRestrictedVlanRegistration should include the statement "The
value of
this object MUST be retained across reinitializations of the
management system."

3) The P-BRIDGE-MIB and Q-BRIDGE-MIB read-write objects from RFC2674 -

dot1dTrafficClassesEnabled, dot1dGmrpStatus,
dot1dPortDefaultUserPriority, dot1dPortNumTrafficClasses,
dot1dRegenUserPriority, dot1dTrafficClass, dot1dPortGarpJoinTime,
dot1dPortGarpLeaveTime, dot1dPortGarpLeaveAllTime,
dot1dPortGmrpStatus, dot1qGvrpStatus, dot1qStaticUnicastAllowedToGoTo,

dot1qStaticUnicastStatus, dot1qStaticMulticastStaticEgressPorts,
dot1qStaticMulticastForbiddenEgressPorts, dot1qStaticMulticastStatus,
dot1qPvid, and 
dot1qPortAcceptableFrameTypes should include the statement "The value
of
this object MUST be retained across reinitializations of the
management system."

4) I originally suggested that for all objects that were already
defined in RFC2674, it would probably be wrong to make conformant
implementations non-conformant by
adding a new requirement, so I suggested the text: "The value of this
object SHOULD be retained across reinitializations of the management
system."

Bert responded, "I'd add some text then as to why we use SHOULD
instead of MUST.
In other words, make it clear that the SHOULD is there to accomodate
old (compliant implementations) but the new implementations MUST 
use persistent storage and that old implementations are
encouraged to also make it persistent."

Since this is a question of compliance, I now recommend using MUST in
the description, and adding text to the compliance description
clauses:

qBridgeCompliance: "RFC2674 was silent about the expected persistence
of the read-write objects in this MIB module. Applications MUST NOT
assume that the values of the read-write objects are persistent across
reinitializations of the management system, and MUST NOT assume that
the values are not persistent across reinitializations of the
management system."

qBridgeCompliance2: "This document clarifies the persistence
requirements for the read-write objects in this MIB module. All
implementations claiming compliance to qBridgeCompliance2 MUST retain
the values of those read-write objects that specify this requirement."
 
Is this acceptable to the WG?

David Harrington
[email protected]