In search of a good term/name for a metaclass [message #477304] |
Wed, 30 April 2008 18:21 |
Andreas Maier Messages: 32 Registered: July 2009 |
Member |
|
|
Hi,
in a standards organization (DMTF), we have a metamodel called CIM that
is very close to UML. In fact, there is a UML profile for CIM published
by DMTF.
We do have a concept of "management profiles" in the DMTF (not to be
confused with "UML profiles"), which define how exactly classes out of a
large pool of classes called the "CIM schema" are used in a particular
context. A particular class from the CIM schema may be used in multiple
ways by multiple such mgmt profiles. For example, a CIM_Sensor class may
be a speed sensor in one mgmt profile and a temperature sensor in
another. Both of these usages could even be defined in the same mgmt
profile.
The reason classes are not implemented directly as defined in the CIM
schema is that mgmt profiles usually need to specify additional
constraints on them (plus mgmt profiles define the set of classes to be
implemented and how they map to the real world).
In a particular implementation, the classes implemented in multiple mgmt
profiles come together in a single run-time infrastructure that provides
a mgmt service, so regardless of whether these multiple different usages
of CIM_Sensor are defined in the same or different mgmt profiles, they
will both be present in the same implementation.
Basically, each usage of that class CIM_Sensor represents a set of
instances that somehow belong together. So it is not really the class
CIM_Sensor that represents these sets of instances, it is each usage of
CIM_Sensor in a mgmt profile that does.
The reason I am discussing this here, is that we are in search of a good
term for a metaclass that represents the concept of a usage of a class
in the context of a mgmt profile.
One term that was proposed in our DMTF working group was "class
adaptation", based on the idea that a CIM schema class is adapted by a
mgmt profile, but not everybody in our group was happy with that term
(reason: we really do not want to suggest that the mgmt profile modifies
the class defined in the CIM schema, however "adaptation" somehow
suggests that).
Another proposal for that term was "instance specification", based on
the UML InstanceSpecification metaclass, but I personally think that the
UML concept of an instance specification is not very close to our
concept here.
-> If anyone out there has a good idea on such a term, post it here.
-> Also, please comment on the two terms discussed above.
Andy
|
|
|
Powered by
FUDForum. Page generated in 0.03439 seconds