COSMOSDatatCollectionMeeting20070326-TPTPInterlock

TPTP & COSMOS Interlock on Data Collection

March 26, 2007

Attendees

Don Ebright

Joel Hawkins

Valentina Popescu

Harm Sluiman

Mark Weitzel

Ali Mehregani

Minutes

The intent of the teleconference was to further understand and document the relationship b/t TPTP and COSMOS. This call focused on the data collection interfaces. Harm has been working on documenting the Data Management Services that will be implemented in TPTP. That document is located here: http://wiki.eclipse.org/index.php/TPTP_DMS

Joel began with an initial discussion of what exists today, and what he is currently working on, and general thoughts for how the remaining work will be implemented. We talked through the initial wrapping of GLA as a tactical step in being able to demonstrate some immediate value. We would like to be able to integrate the TPTP RAC as soon as possible.

We discussed two different types of APIs. One that is "strongly typed", e.g. getTopology, and one that is more "loosely typed", e.g. "executeQuery(someXPathString)". There is a recognition that both are needed, and that, in general, the "strongly typed" APIs could exercise the "loosely typed" queries as their implementation.

We briefly reviewed the dependencies of the current work and agreed a more formal document capturing these is necessary as part of the project. Currently, the dependencies for data collection are:

• Java 1.5 (Sun)
• Apache Muse 2.2

Note:

• There was a concern that many commercial infrastructures are still on Java 1.4, that this would be a barrier of acceptance for the data collection layer.
  • What is the plan for commercial OMPs to be on a Java 1.5 base?

A brief discussion was held on the Data Collection F2F. It would be beneficial to host this in Toronto at the IBM site so that if we needed to interact with the TPTP team, we could do so.

Future Direction Joel is initially thinking of using the Java Persistence API (JPA) as a means to talk to relational back ends. It may be possible to use code that will be contributed to Eclipse through Oracle. This is a key decision and one that has not been worked through by the entire team at this point.

• Note: There is no implementation of JPA at Eclipse.

We discussed the desire to be able to use the same models between TPTP and COSMOS. At a minimum, these will be documented in WSDL and XSDs so that they can be bound to remotely via a Web service. In optimized deployment scenarios, it would be beneficial to share the same POJO model.

• These models are in the process of being defined by the TPTP team and we will need to schedule another call to review them in detail

Action Items for the COSMOS Data Collection Team

• It was agreed that we needed to document the "out of the box" experience for the data collection layer. This would include the deployment scenarios.
  • Include the set of "profiles" which will dictate how COSMOS is packaged & installed
• Produce a set of dependency graphs that clearly articulate the pre-requisites of the data collection layer
• Describe using WSDL & XSD the agreed upon set of models that interface b/t TPTP and COSMOS.
  • A stretch goal is to produce a Java POJO interface as well