Em Qui, 2005-08-18 às 19:19 -0300, Daniel Ruoso escreveu: > It seems to be dealing with the exact same problem I'd like to deal > with Oak2
Well, this is a data dump of my brain at this moment... :) * The application is distributed among nodes. * Each node can see any object (remote or local, in a transparent way) * Use of UDDI servers to see where is what * A GUI client *is* an application node, and can provide services itself (in a store, the manager's node can ask the operator's node for an information) * So is a Web client. * So is a server that is just providing services for the gui and web clients (or for other server nodes). * The information can be spreaded (this is more usefull for e-gov environments, as we don't want to control the citizens, but the gov actions...) * Distributed transactions over different nodes. * SOAP over SSL, using the SSL keys to certify nodes. (I mean, having a SSL private key for each node, so you can know if that message was sent by that node. Also allowing setting trust levels for the keys, so it's possible to treat differently a server in a datacenter and a desktop machine.) * Definition of a business flow using events generated by objects. The listeners for that events can be dinamically attached. (For instance, if I have a purchase order confirmed, the finances dept should be notified about it's value, but I could just disable it, if the finances dept is not using the system yet (during the transition from an old system), or change to the code that integrates with other legacy systems, or even having a different way of dealing with this information) * Meta-data for entities relationships, so new relationships can be defined dinamically. This meta-data defines which type of object would handle it's data when other node is looking at it (a GUI node, for instance). (A account payable can refer to a Supplier, to an Employee or to the Investors, but its logic is the same in all cases). What do you think? daniel