Bogdan asked: > Why the MDR implementation uses synchronized java.util.Map > and synchronized methods in ModelEventPump? Are there any > threading issues?
There are definitely potential threading issues concerning the Model subsystem implementation, but I'm not sure that we have a comprehensive design that addresses them. They've mostly been dealt with on an ad hoc basis. For the MDR implementation, model events are delivered on a separate MDR-specific event delivery thread. Most model modifications are initiated from the AWT event delivery thread where all GUI actions are handled. Some things like model loading happen on a separate worker thread. There is a critics thread which using the Model subsystem, but it should be readonly accesses. If EMF follows similar model to MDR of event delivery, it shouldn't introduce any additional problems, but it might be worth writing up a quick summary (or providing a pointer to one) of how the EMF event delivery works (what thread, when, etc) as well as any synchronization requirements for the EMF/UML2 methods that query/modify the model so that we can review it. An example of something that would cause us problems is if you are not call EMF/UML2 methods from a EMF event listener. If this were the case, we might need to re-dispatch events through a separate thread for serialization and/or decoupling. Not sure that last part is very understandable. If I had to guess though, I'd guess that we won't have any significant problems here. Tom --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
