Hi Marc and mhe,
I'm afraid there are no design documents available, but here's a short description of how it works. A Ge graph is stored in a pwg-file that contains all the graphical objects and objects for dynamics and actions. This file that is read by the java script, so there is no java applet involved, all you have to do is to copy the pwg-file to the web directory. A graph is opened in a web browser by ge.html that takes the name of the pwg-file as an argument, eg 'ge.html?graph=pwrdemo', and if it's an object graph, also the instance object can be stated 'ge.html?graph=pwr_c_av&instance=H1-Av1'. ge.html creates a canvas and starts the script ge.js. ge.js reads the pwg-file and opens a web socket to the host node to get login information, information about attributes and objects, setup subscriptions and receive subscripton data etc. The web socket connection is made to the WebSocketServer process in the host node. This is still written in java but hopefully will be rewritten in c++ some day to get rid of all java depencencies. The demo on proview.se actually doesn't open a web socket, it uses a local database in the java script. This is why some demos that requires a real database and a plc program doesn't work.
/Claes