Outcome of allday meeting at Aberdeen Wed 17th April 2002 ========================================================= Present; P.Gray, A.Preece, K.Hui (Abdn), Jessica Chen-Burger (Edin) ==================================================================== The basic plan for the demo was as given on Kit's web page: http://www.csd.abdn.ac.uk/~khui/akt/ix/ We looked back at the presentations that Jessica and Kit gave at the Edinburgh January AKT meeting (on Intranet). We are in the process of fleshing these out. Moving Constraints from Panel to Solver ======================================= Kit has made considerable progress in automating the transformation of constraints from a Daplex form to an RDF form suitable for transmission by AKTBus or on the semantic Web. This goes via an intermediate ICode form that can be passed easily to the KRAFT constraint fusion and solving software. This fulfils much of the requirements for the Aberdeen side. Representation of Constraints in Process Language on IX panels ============================================================== Jessica's constraints are basically conjunctions of conditions, represented as two-place terms. She showed a classification ontology of terms used in her process description language: http://www.aiai.ed.ac.uk/project/akt/work/jessicac/pc-config/onto-model/top-level.html - click on Contents frame onto2.dia See also example instances in prolog syntax: http://www.aiai.ed.ac.uk/project/akt/work/jessicac/pc-config/jess.pl Kit has developed a simple parser for extracting Daplex declarations and constraints from this form, demonstrated in January. We need to use this for the July demo, so any changes to Jessica's constraint format should be done in a way that allows this parser to evolve gracefully! Jessica did not see the need for big changes. Jessica agreed to include information on Variable Types in an extra argument in her constraint format. This would make it more self-contained and act as a redundant check. Currently it has to be inferred by looking for named variables and matching them up with template declarations. Construction of the "Aberdeen Panel" ==================================== During the demo this should run on a separate workstation/computer from the Edinburgh panel. It will be constructed in Java, with Workflow facilities being defined and provided by Jessica, and with callouts possible to Prolog (probably via sicstus/Jasper). Calls may also be made to AKTBus. The Panel should basically be a "clone" of the Edinburgh panel, with different state, and specialised facilities for dealing with Technical constraints {Edinburgh deals with Cost constraints}. An Issue (in INCA model) raised on the panel, which is to be passed to the Aberdeen solver, will be sent over the AKTbus using a simple FIPA-like performative and constraints. Aberdeen are re-designing this FIPA protocol with Soton, and will shortly make a proposal. The Issue and associated constraints will be initially in Jessica's IX form, and Kit's software will be used to transform it to RDF and then send it down AKTBus to the solver. Panel status Visualisation ========================== This is being done by Jeff Dalton at Edin. Jessica demo'ed the latest release and distributed a Quick Start guide to IX Process panels, which was much appreciated. She showed Kit how to download it for experiment. Use of RDF between IX panels ============================ It is not clear whether Edinburgh should use an RDF style message for transmission between IX panels. For July, something simpler may suffice. Jessica will reflect on it. Kit can make his software available for Jessica to download. She could use some of his Prolog as a template to adapt for IX-panel work. Alun pointed out that if Edinburgh used RDF internally then it could use XML stylesheets to display constraints and workflow state on the panels, which is good practice. Integration of Broker software with Panels ========================================== To add credibility, the Edinburgh and Aberdeen panels should contact some kind of Broker to discover the identities of other agents to which Activities can be passed (for display in the Actions box). For the July demo, this Broker may be Steve's fully-fledged Broker, or may be something very simple that we should be able to implement in a very short time period. In either case, someone at Edinburgh who knows the panel code well needs to implement a message call-and-response when the Actions box is generated. This should be decoupled from the Aberdeen IX-TIE work for the moment, in order to get parallel development and avoid delays from non-essential software dependencies. Once the demo is working and reviewed, a tighter coupling may well be designed. See attached Jessica Powerpoint slide + diagram emailed at end of day. meeting-summary.txt ------------------------------------------------------------------------- Prof Peter M.D. Gray email: pgray@csd.abdn.ac.uk Department of Computing Science pgray@computing-science.aberdeen.ac.uk University of Aberdeen tel: (+44) 1224 272292 Kings College, ABERDEEN fax: (+44) 1224 273422 Scotland, UK, AB24 3UE www: http://www.csd.abdn.ac.uk/~pgray/