20050208 Editor's call minutes.

[Gary P Cole <Gary.P.Cole@Sun.COM>]

NOTE: These minutes are from *last* Tuesday. 
I'll publish the minutes from this week's Editor's call soon.  (Short 
call--won't take long.)
Participants in this morning's working group call discussed:
- reference capability questions

Attendees:
- Jeff Bohren (Open Networks)
- Darran Rolls (Sun)
- Gary Cole (Sun)

After reviewing the reference examples from the Houston Face-to-Face,
Gary Cole expressed several concerns about the Reference Capability.

1)  Without discoverable meta-data describing the connection types
    and which schema entities each connection type may connect,
    Reference Capability does not promote interoperability.

    Jeff Bohren answered this concern by saying that the approach
    agreed at the Houston Face-to-Face DID include discoverable meta-data. 
    We simply failed to capture this facet in our notes and examples.

2)  Since references may span targets, it may be problematic
    for a provider to declare as part of a target's support
    for the ReferenceCapability any type of Reference 
    that refers to a schema entity on another target.

    Jeff Bohren addressed this concern by saying that
    it should be possible to declare as part of a target A
    a referenceType that refers to an entity on another target B.
    It may not be as simple as referring to an entity
    that is supported by the target A (since one of 
    target A's SupportedSchemaEntity must already define 
    a convenient "alias" for that entity), but
    targetA could specify the 'toEntity' for that referenceType
    using the more canonical syntax of the SchemaEntityRefType.

3)  Aside from the scalability benefit 
    (which results from the ability to exclude references 
    from the schema-defined XML 'data' of each object),
    the ReferenceCapability seems to offer little benefit 
    over schema-defined attributes that "just happen" 
    to refer to arbitrary objects.  In particular,
    there is little benefit to interoperability.

    Jeff Bohren addressed this concern by saying that
    the Reference Capability would be very valuable
    in conjunction with a "standard" schema (i.e., in
    any profile or binding that assumes a certain schema).
    We must define the Reference Capability now
    in order to exploit it with subsequent profiles.


Action Items
------------

1)  Jeff Bohren to supply examples of defining named reference types
    (i.e., discoverable meta-data for references) and modifying references.

2)  Gary Cole to work offline with Jeff Bohren
    to reconstruct agreement wrt Reference Capability
    and then update draft spec accordingly.