Version 3

    EJB 3 Q3 2009

    This document outlines the strategy and direction of EJB 3 for 2009 quarter 3 (and beyond).

     

    All of the previously outlined strategy still aplies (EJB3Q12009), so I won't repeat those parts here.

     

    Components Life

    To have components which have a minimal impact during the 'Full Support' phase (3 years) the components should be stable and backwards compatible for the duration of at least 3 years. Effectively that would mean we won't have to branch the component.

     

    This does mean that any dependency used by a component must adhere to the same strategy and thus provide a stable API accross multiple JBoss Application Server releases. Currently this is deemed unreleastic as we can expect API changes in VDF, AOP (/ interceptors) and/or MC.

     

    Given that ejb3-core is too large and has (weird) dependencies that can never adhere to the above requirements there is no other alternative than to branch off.

    Performance

    As opposed to previous directives, performance should now be taken into account on all activities. Currently booting up EJB 3 Embeddable takes too much time (~6 secs) while no clear hotspots stand out.

     

    With each code refactoring or code design at least an understanding of performance impact must be known beforehand. If the performance impact is negative an evaluation should be made whether to proceed or not.

    Q3 Goals

    Top priority remains any EAP work.

     

    We should have a full EJB 3.1 compliant implementation before October 31st. Thus for that milestone the following requirements are defined:

    Must have:
    • EJB 3.1 Embeddable (& the new JBoss Embedded) (Andrew)
    • JNDI Binding (Carlo)
    • TimerService (Carlo)
    • Metadata (Jaikiran)
    Should have:
    • Deployment performance (Jaikiran)
    • Graceful Shutdown (Andrew)
    Could have:
    • JMX View (replacing the @Service beans)

    • Managed Object View (the real replacement )
    • Remove AS dependencies
    • Remove AOP dependency

     

    Possibly we will have a high involvement in the JSR-250 compliant injection framework based on MC.