Present: Francois, Kris (minutes), Nikolai, Steen, Vito
Excused: Alessandro, Jens, Marc, Vito
1.2. Server Framework
The C++ has been finished according to the design changes made 2 weeks ago. Steen is working now on the C API and on the default poller for PS (PLS poller).
The technical student which should work on the Java implementation will arrive next week. (to be confirmed by Alessandro).
1.3. RDA
Two additional days were spent on Java implementation as the Data/DataEntry objects did not serialize properly. This was because the CorbaAny serialization is not implemented in ORBacus. This is a flaw in the Java binding specification which did not require it explicitly. CORBA 2.4 spec corrected it.
Nikolai has made an implementation of Data/DataEntry which is CORBA independent and can be serialized (not only for JMS but also to save it on file, etc). He is then copying the data to use CorbaAny for transport as before. Nikolai measured the overhead of copying which is negligible.
Nikolai reckons that he can maintain the delivery date of the C++ implementation for the end of October. He will make the first implementation on Linux and then port it to LynxOS.
1.4. JMS client and MoM Agent
The JMS client is using Data/DataEntry now. The MoM agent is ready and waiting for the CORBA server on LynxOS.
1.5. Public API
It is almost finished. All calls are wrapped in independent classes and the packages to be used can be selected at runtime to choose the subscribe between Corba and JMS implementations. Documentation is completed 80% and will be finished next week.
1.6. Naming service
Implementation has started. Kris will implement the access to the Equip DB as the first approach.
Kris reported from the 5 meetings in which he, Vito, Marc and Francesco from the MW team participated. For Kris the conclusions were not satisfactory as no edge-cutting decision was taken. Pierre Charrue has requested the following:
Kris reckons that there is no technical problem to fulfill the requirements, except that the "dynamic devices" would have to be better defined.
Everybody has agreed that the first priority is to deliver the first version for the end of the year as it was planned.
It has been requested that a better common repository and file space is made available. AFS is not convenient to work from LynxOS and NT. Kris will propose to the Linux manager to make an NFS directory on pcslux8 which is backed-up and mountable on PS machines.
There is need for a common approach to diagnostics and tracing. The servers which are under development or to be developed soon (PS, OPC, SL Equip) would profit from a common approach. To start with we can extend the diagnostics interface already defined in IDL by Nikolai.
Next meeting Wednesday 18 Oct 9h