- From: Sebastian Feuerstack <Sebastian@Feuerstack.org>
- Date: Mon, 01 Oct 2012 11:37:02 -0300
- To: public-mbui@w3.org
- Message-ID: <5069AA8E.8090101@Feuerstack.org>
Dear colleagues, here are my notes regarding the latest AUI class diagram. I have attached a class diagram that just illustrates my change requests and does not cover the entire AUI. I would like to propose: - to consider requirement GR5 “Mappability” by two classes “Entity” and “Mapping”. A Mapping defines a relation between source and target entities. An Entity contains an unique identifier that can be referred to. In the attached diagram i have aggregated Entities into UserInterfaceModel but i would prefer to see a Task to be derived from an Entity as well, so that a task can be mapped to e.g. an AbstractInteractionUnit. - that Documentation can be linked to a Model and all classes and attributes that we consider as documentation are derived from this class to better separate machine-processable data from human-processable data. The latter might not be relevant at system run-time. - an AbstractInteractionUnit can aggregate Behavior conforming to specific requirement 3 - that Behavior is a special kind of a mapping and optional - consider the composite pattern http://en.wikipedia.org/wiki/Composite_pattern and therefore explicitly consider the “leaf” as an abstract class (e.g. ElementaryIU) I am continuously trying to relate each proposal of the AUI class diagram to our AUI implementation and realized that - in general - we would require less effort in adapting our model to the W3C one if new concepts are introduced by sub-classes instead of attributes. From the implementation point of view it is much easier to introduce a new class into an existing hierarchy than new attributes that would require code changes in the existing classes. Further on, abstract classes can be used as elegant “extension points” for all concepts that have not been considered in the W3C AUI. Therefore i would vote against the “minimize abstract classes requirement”. Regards, Sebastian -- Sebastian Feuerstack Department of Computer Science Federal University of Sao Carlos - Brazil http://www.feuerstack.org Check out MINT 2010 - the Multimodal INTeraction Framework http://www.multi-access.de
Attachments
- image/jpeg attachment: AUI-011012-sfeu.jpg
Received on Monday, 1 October 2012 14:37:37 UTC