Example 3 in A.4 purports to show instances of the Department model defined
in the previous example (it says that "Department is considered the
metamodel for Chemistry"). However Department is not a MOF-compliant
metamodel (it is a fairly basic UML model). So XMI just does not apply!

Similarly the XMI Document Instance in Example 4 in A.5 shows instance data
for the Mail model which again is not a MOF-compliant metamodel.

Question if both MOF and UML were defining the same tag values for initialValue. No change. This is a DTD initialValue which differs from UML's initial value.

I propose that the XMI RTF not regenerate UML and MOF DTDs, for the
following reasons:

1) UML 1.1 and MOF 1.1 are now out of date.
2) Both the UML RTF and the MOF RTF are now creating their own DTDs, so it
is not appropriate for the XMI RTF to continue to create them.

see above

The second and third paragraphs in 6.6.1 define exactly how to use Namespaces in XMI DTDs and serve as both the restriction and the solution. Accept. Clarify to indicate that the following text is the definition of the restriction.

Accept. Clarify to indicate that the following text is the definition of the restriction.

Summary: THe background it is the use of XMI for interchange of warehouse
information. The person prototyping
the exchange is Brain Volpi so if you have follow on questions, please
copy him since I am on vacation for the next two weeks.

1. We are using UML to define the meta models and the created the DTD"s
and document streams.

• our model contains multiple inheritance which is hard to handle
  in XMI. Does anyone have any
  comments or suggestion on the mapping to XMI we should consider.

• our model contains UML interface definitions which we have mapped
  into abstract classes
  any comments on a better way to do this?
• UML has ordered associations and we are currently assuming the
  XMI stream will always be
  in the correct order. Is this valid?

resolved in XMI 1.1 RTF

Summary: 4. XMI has specific rules for attribute ordering. Can this be relaxed?

resolved in XMI 1.1 RTF

Summary: 3. XMI uses containment rules to drive the document generation process.
Is it valid to use alternative
DTD generation rules that allow for reference semantics where we can
use bi-directional references
instead of the single direction references you are limited to with
containment. We need roles at both
ends.

resolved in XMI 1.1 RTF

Summary: 2. How should null values be transmitted. Currently we don"t transmit
them. This is really the issue of
partial models versus null values that Steve raised in the original
submission discussions.

resolved in XMI 1.1 RTF