Week10

Objectives
none

Progress
Network version based on CNS is working.

However, the CNS server crash when an other process that use the same node address want to use it to create the output end of a channel. I investigated and located a bug at org.jcsp.net2.One2NetChannel.destroy line 621. The One2NetChannel don't hold the eventuality of a local connection here, despite the fact that the class is designed to(probably a small mistakes). I tried to get the last version of the JCSP's jar but only fount earlier version on their official website. I run the server on an other JVM to solve the problem but this solution may not be applicable to our real case.

I will write a version that work without CNS. I initially did that to learn how it work but CNS hold for us the channel instantiation problem at the expense of the efficiency. Since we don't create channel dynamically, the channel location problem should not be solved during run time.

It should not be very complex and will probably do it in parrallel with my repport. It would have been done if I wasn't sick this week end.

Supervisor's Comments
Well done AND after the discussion we both knew that you had done it corectly and in a very sensible manner.

With respect to the net2 problem, the best thing is to talk to Kevin Chalmers - he wrote it!

BEFORE you can have any robots to go on to the next stage you have to WRITE this up

AND the benefit of doing this will be that it will make the explanation of the BT implementation much easier to describe because you will be ablt to state that you are just replacing the IP:port number as a BT address and that all you requuire is the code in each BT node that sends the data to the correct channel number.

I got carried away with the next stage of implementation BUT Nicolas reminded me that he should do some writing - well remembered and even thought you would prefer to code you realise the importance of writing this up.