Minutes of meeting 18 Jan 2002

Meeting opened 1.05pm

Present: Raphael Goldberg, Peter Howarth, Daniel Justice, Michelle Osmond, Zhu Tan, Aleksandar Trifunovic, Nicholas Dingle

Discussion with Nick Dingle about Medusa

• Room 332 opposite 308 & 311 ("Antonio Garcez" on the door)
• Runtime.exec() to run external programs (like Dnamaca) from the command line: worked without problems.
• Borland JBuilder is good (can get on magazines)
• Simulation module would be useful for comparison with analytical results from Dnamaca
• Graph theory module: was difficult to construct the coverability graph
• PNML: an existing tool that uses it is the PNK (Petri Net Kernel), which can deal with different types of nets, and has modules that communicate back to the main program.
• SAX was easy to use, but other ways may be better (didn't investigate early on). Similarly with the method he used for transferring data to modules.
• Swing is thread-safe - ie don't want multiple threads accessing Swing components. Had problems with it's circles!

Further discussion - what we've done since the last meeting

• Daniel: XML: what it is. DTD's, SAX, DOM, JDOM. (Word document)
• Peter: Reflection, GUI - probably not too hard, but it's worth putting effort into making it intuitive and elegant.
• Alex: Dnamaca, Runtime.exe. Threads - seem to be relevant for animation. Invariant analysis, looked at the matrices needed and algorithms.
• Raphael: read the two MSc reports from last year, looked at invariance analysis, Markov and graph theory.
• Tan: looked at subnets in Renew (graphical handling), and Java.
• Michelle: the new module (subnet) implementation in PNML and method used in last year's project. Summary of XML and PNML.

Discussion with William Knottenbelt

Key functionality and possible extensions for the projects are described below.

• Display and graphical editor for petri nets. Could write a class for one 'pane', then could open multiple instances (eg edit multiple nets, or subnets).
• Animator: click on a transition to see it fire. Automatic animated simulator
• Subnets
• Modules:
  • Invariant Analysis
  • Graph Theory. Also for nets with immediate transitions
  • Simulator. Also with inhibitor arcs, which enable a transition when the input place is empty (not expect to support them for the other modules, it's hard!)
  • Interface to Markov chain analyser (Dnamaca) - takes text files as inputs and outputs
  • General interface - communication between modules and main program, eg for animation. Maybe return special object, or call functions?
• Coloured petri nets
• Threads - for editing different nets?

To decide in the next week or 2:

• write requirements
• high level OMT diagram
• GUI: what should it look like
• module interface: parameters and return types

Weekly meetings with wjk: 12:30 on Fridays

Further discussion

Very likely to use JDOM for main internal data representation. Could be wrapped in a custom class with useful access functions that we determine could be useful to modules (eg a function to return a particular matrix of the net). If we pass this class to modules, they will have access to both the 'convenience' functions and the entire JDOM, allowing them to extract anything they like, including future elements that may be defined in PNML later. They can also modify the net if necessary (eg putting it into a particular marking) - may or may not be a good thing.

Need to determine interface for modules. People will look at modules and see what they require, how to do them (broadly) and what they'll return (what sort of results, and how they could be presented to the user).

We'd like a clearer view of how JDOM actually works. Also if it will work on the college machines ok.

We haven't investigated the simulation aspect properly yet.

Need to think about possible features/layouts of the GUI. Inspiration from the concurrency program? (tabs, one pane has the code and another has the graph) Would be nice to be able to view the PNML from within the program - most users will not want to deal with it (the editor is meant to make it unnecessary), but it would be nice to provide the option. Subnets: opening a new window to edit a subnet seems necessary. The subnet may be within the same pnml file as the main net, or it may be in an external file or even on the internet (the pnml allows for this: theoretically, this means there could be a library of publically available subnets on websites which could be accessed and used).

To investigate:

• Aleksandar: Invariance analysis module
• Raphael: Graph Theory module
• Peter: Simulation module
• Daniel: Details of the JDOM
• Michelle: Swing, JDOM in practice
• Tan: GUI (also other tools' GUIs) and overall design

Next meeting:Monday (21st), 1pm in the common room.

Meeting closed 3.30pm