My apologies for the late delivery of this contribution. For the moment I'm locked out of the PictureTel server, so here's the text:
TITLE: Implications Of Context Attribute Descriptors
___________________
Summary
This contribution is written to express concern over the proposed inclusion of context attribute descriptors in H.248v2. This innovation is seen upon reflection as a major change in the H.248 protocol architecture, and threatens the coherence of the protocol. It is proposed as an alternative that we gain further experience by seeing if it is possible to avoid the specification of new context properties while solving the problems of MCU decomposition. When we are satisfied that we can enunciate clear criteria for adding properties to contexts as opposed to terminations and further, that we are aware of all of the implications of context properties for the protocol, we can proceed to update the protocol accordingly.
1. Introduction
H.248v1 defines three context properties: the Topology Descriptor, the Emergency flag, and the Priority flag. These additions were desirable, but were made at a fairly late stage in the development of the protocol. As a result, the work was done in an ad hoc manner and introduced irregularities into the operation of the protocol.
Delayed Contribution 129 (Porto Seguro) proposed the introduction of a context attributes descriptor to contain new context properties and to allow addition of such properties in packages. On the one hand, the present contribution argues that this step, while desirable, does not go far enough to assimilate context properties into the overall protocol structure. On the other hand, introduction of the ability to define context properties in packages at this stage in the development of the protocol may cause confusion and threaten protocol interoperability at a practical level.
2. A Complete Implementation Of Context Properties
A complete implementation of context properties requires that the protocol satisfy the following requirements: · there is a well-defined method for setting property values; · there is a well-defined method for modifying or clearing property values; · there is a well-defined method for auditing values that have been set on a context; · it would be highly desirable to have a method for determining what context properties an MG supports.
Beyond these protocol matters come questions of semantics. We must decide whether the permissible properties of a context can depend on what media are flowing in it, or on the properties of the terminations currently in the context. Since the active streams are a property of a context, we must decide whether context properties can be associated with specific streams, and if so, what syntax is used for the purpose. There are undoubtedly other issues which may come to light.
As a possibly controversial final step, it would seem desirable to assimilate the existing context properties into the new syntactic structures which would evolve. With appropriate requirements for transitional interworking, the existing syntax for topology and the emergency and priority flags should be deprecated and counterparts should be defined within the context attribute descriptor.
3. Potential Confusion
The final point this contribution makes is that if we do not establish a clear definition within the protocol specification of which properties should be context attributes and which should be termination properties, packages will inevitably be defined which solve the same problem with differing mechanisms. This will be unhelpful to achieving the interoperability of different implementations.
4. Conclusion and Proposal
Context attributes should be formalized only after careful study of all of the issues. In concrete terms, this contribution proposes either that the generalization of context attributes should be left to H.248v3 or else that H.248v2 be delayed beyond the proposed February 2002 target.
Tom Taylor taylor@nortelnetworks.com Ph. +1 613 736 0961 (ESN 396 1490)
_______________________________________________ Megaco mailing list Megaco@ietf.org http://www1.ietf.org/mailman/listinfo/megaco