DSDP-TM Connector Meeting Salzburg 2005x11x14
Revision as of 13:17, 19 December 2005 by Moberhuber
|Meeting Title:||Meeting to continue discussion on Target Connection Adapter proposal|
|Date & Time:||Monday November 14, 2005 at 3pm CET|
|Location:||WR Office, Salzburg, Austria|
- Peter Lachner, Intel
- Martin Oberhuber, Wind River
- Felix Burton, Wind River
- Michael Scharf, Wind River
- Peter shows slide with new scenarios for connectors.
- Felix: The most difficult thing with the TCA proposal is defining the interfaces.
- Hard to think about the generic connector concept
- Start by defining only interfaces that TM needs for registering connectors, and defining dependencies between them.
- Additional interfaces to be defined by connectors themselves
- TM to provide a GUI to show basic services on a target
- Selecting a particular entity on a target might reveal new services (with interfaces defined by the parent)
- We need to separate the protocol from the connection (somebody might want to do a protocol without a connection, just for translating protocols).
- Michael: Connectors are like resolvers - they resolve "I want to... with target" by providing the right connection module and protocols. There might be some preconfigured resolutions, plus some autodetected resolutions.
- With our connector chain represented by some "black boxes", we want as little intelligence as possible to reside in the individual boxes.
- Michael: Interfaces for the various protocols provided by connectors could be specified in an "action-like" manner: each action to be executed is an interface in its own right, which may be available or not
- Martin: TM should only hold the meta-information for connectors (that can also be programs outside Eclipse): provided-protocol, required-protocol. Example: Wind River chain of servers (Eclipse - dfwserver - tgtsvr - target agent) is mostly not Java
- Felix Burton worked on "Receptacles" 1980 -- showcase that composition and configuration of connection chains can actually work