[cosmos-dev] Review of enhancement 200222

["Ali Mehregani" <ali.mehregani@xxxxxxxxx>]

Summary of the design review for enhancement 200222

The following design document was reviewed during the architecture meeting on September 6th, 2007:
http://wiki.eclipse.org/COSMOS_Design_200222.  Here's a summary of the review:

1. Mark: General question about the design document of enhancements:
Should a design document include detail about namespaces/package names/WSDM capabilities/etc… that an enhancement will introduce?

    - That level of detail should only be included if it affects other subproject and/or clients who intend to extend COSMOS.
    - The information should also be captured under a separate wiki page and be accompanied as part of the Eclipse help with the WSDM tooling.

2. Mark/Joel: The last figure under the architecture section is misleading.  The connection and the operation fetch are not network operations.  The method of fetching and using the query service for MDRs should be flushed out under enhancement:  
https://bugs.eclipse.org/bugs/show_bug.cgi?id=200244

    - Renamed "client" to "query service provider"
    - Added a paragraph after the figure to better explain the scope
    - Further generalized enhancement: https://bugs.eclipse.org/bugs/show_bug.cgi?id=200244

3. David: There is no mention of applying the property subset directive in the flow diagram:

    - This step was missed in the diagram.  It is a post fetch operation that will be handled for itemTemplates.

4. Sheldon:  A mirror operation of the output transformer should be provided for the client to be able to convert an XML query response into a POJO graph representation.

    - Agreed: Modified the design document accordingly

5. Hubert:  More detail needs to be provided about the common components that will reside under the data collection subproject.

    - David is working on the input/output transformations and he will work closely with Hubert to flush out the common components.
    - The outcome of their discussions will be posted on the mailing list.

6. Hubert:  How does the client know what to query for without any pre-knowledge of the data stored in a repository?

    - This falls outside the scope of this enhancement but it is an open question that needs to be addressed by the data collection subproject.

7. Hubert:  Why does applying the instance and record type selectors before property and xpath selectors make a difference in terms of performance?
    - This is completely a repository implementation detail.  The SML repository is implemented in a way that will process instance and record type selectors faster than the property and the xpath selectors.  It is not a common sequence of steps that all MDR query services are expected to follow.


Please use the talk page included at the bottom of the design document to add questions/comments/concerns.