I'm just doing a brain dump of how I think I'd like to see
the 5.0 api
move, where we move from a rule oriented framework to a knowledge
oriented framework for rules and processes.
sounds interesting. how exactly do you see processes, integrated in the
knowledgebase, other than just calling ordinary java classes in the then
part of a rule? Will there be support for other AI-type algorithms?
A rough namespace and interface mapping could be as follows:
KnowledgeBase <- RuleBase
StatefulKnowledgeSession <- StatefulSession
-extends both RuleSession and ProcessSession
-in theory someone could have a rule or process specific impl
although we will continue with the combined approach for now
StatelessKnowledgeSession <- StatelessSession
-contains the rule specific api from WorkingMemory
-contains the process specific api from WorkingMemory
KnowledgeDefinitionPackage <- Package
-Possible just KnowledgePackage
but I like reinforcing that these are definitions
org.drools.knowledge.defintions.rules <- org.drools.rule
-will eventually contain our ontology stuff
-our current pojo model stuff should probably go here too
org.drools.knowledge.engines.rules <- org.drools.reteoo (and common,
etc which will also be refactored)
We will create a drools-api module. This will contain all public user
apis for all
modules in the project, which will allow us to support
service frameworks better, such as OSGi. This will also allow us to more
easily differentiate from public stable apis and internal apis.
To get compability with 4.0.x we'll have a drools4-adapter jar, which
all the 4.0 names and just wrap and delegate to the new impl.
We won't be able to achieve 100% api comptability, but we should be able
to for all the most common public apis.
rules-dev mailing list