I think that is an example of a "pragmatic" way to solve something that backwards chaining could possibly achieve, but in a perhaps more intuitive way. I don't think its what is talked about in the literature is quite the same, but still, this is very useful.

Thanks for the kind words !


On 2/27/07, Rich Halsey <rich_halsey@bellsouth.net> wrote:
Mark,
 
In yet another point of interest, I also see the JBoss Rules "rules flow" as a very elegant form of backward chaining implementation/modeling. For example, in the PDF file from Kris Verlaenen, the ultimate goal of  "Process Order" cannot be accomplished without the completion of "Check Order" which may/may not (?) complete without "Update Order". As opposed to the engine hunting through all of the conditions and actions to find an appropriate path for the backward chaining, the model implies how it is to be done and the system will guarantee the flow with its split and join nodes via forward chaining - but, if and only if the backward chaining pre-conditions are satisfied.
 
I'm not sure who will agree with what I have said, but this is the way that it appears to me as a rules engineering person.
 
Thanks,
 
Rich Halsey
 
 
 
 
"GENIUS IS THE ULTIMATE WEAPON"
 
....God grant me...
The senility to forget the people I never liked
The good fortune to run into the ones that I do
And the eyesight to tell the difference."
----- Original Message -----
From: Mark Proctor
Sent: Monday, February 26, 2007 7:58 AM
Subject: Re: [rules-dev] Re: RuleFlow preview

Our implementation is a light layer to provide "wait states" for one or more rules, it uses a similar principle to agenda-groups (Clips modules) to partition the execution. Activated rules are placed in temporary buckets (rule-flow-groups), instead of onto the agenda, when the rule-flow-group is activated the bucket empties onto the Agenda for normal execution, when all the emptied rules are fired the next rule-flow-groups are activated.

The system is still "parallel" in nature, in that the agenda is still responsible for executing rules and the agenda can have more than one rule on it at at time. In our implementation all the rules in the rule-flow-group will be put onto the agenda for execution, at the same time standard rules can also continue to be managed and executed by the agenda, and agenda groups (clips modules) still continue to operate - all in parallel.

A rule that is specified to execute as part of a rule-flow-group can also be part of an agenda-group, but that use case is discouraged because it can get quite hairy unless you really know what you are doing :) As it means a rule-flow-group can be activated, the rules moved onto their respective agenda-groups, where any rules not in agenda-groups that do not have focus will not fire, the next rule-flow-group will not activate untill all rules for the current rule-flow-group have fired, regardless of the agenda-groups they are in.

The limitation at the moment is that the temporary bucket has no ability to handle different start instances and differentiate between the rules in it's bucket of the same rule-flow, but you can have multiple different rule flows executing in parallel. We purposefuly kept it simple for "version 1" to build up the functionality needed for rule flow. The use cases for parallel execution of the same flow are not easy - as one instance can catch up and over take another instance on the same flow. Also if a rule in a rule-flow-group activates which of the two current instances for the same rule flow are responsible for firing it? The same issue arrises for when you have the same rule-flow-group in multiple rule-flows. We are currently not sure how best to handle these types of situations; maybe you could help us on those use cases? Or even provide a patch :)

Mark
Rich Halsey wrote:
 
Hi Mark,
 
The part in the document where it says:

"At this point, ruleflow-groups should not be reused in more than one ruleflow, and you should not

start a new instance of a process before the previous one has ended."

will be the weak link in the chain, i.e. there should not be any reason why rule-flow-groups should not be reused nor having multiple instances since rules are implicitly parallel in operation. This was what I found to be the problem with ILOG's JRules back in the v4.0 edition. It turned JRules into a clunky procedural processing engine (which was not what we needed at that time).

However, I am very proud to see that Jboss Rules (JBRules) has successfully evolved to this point. You (and your team) are to be commended for your efforts.

Tally-ho !!

Rich Halsey

 

 

 

 
 
 
"GENIUS IS THE ULTIMATE WEAPON"
 
....God grant me...
The senility to forget the people I never liked
The good fortune to run into the ones that I do
And the eyesight to tell the difference."
----- Original Message -----
From: Mark Proctor
Sent: Monday, February 26, 2007 5:12 AM
Subject: RuleFlow preview

I thought everyone on the dev list would be interested in reviewing and providing feedback on Kris' excellent work on RuleFlow - includes screenshots :)

Mark
-------- Original Message --------
Date: Mon, 26 Feb 2007 01:51:29 +0100
From: Kris Verlaenen <kris.verlaenen@gmail.com>
Subject: Ruleflow


I've attached a document describing how ruleflow is implemented /
could be used in the future. If anyone has got any suggestions or
improvements (on the API I'm proposing, or things you would like to
see differently), just let me know asap.

I think I'll be able to commit a first working version on svn soon.
Still have to include conditional connections (where a connection is
only selected if its condition evaluates to true), and some smaller
stuff.

Kris


_______________________________________________
rules-dev mailing list
rules-dev@lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-dev


_______________________________________________
rules-dev mailing list
rules-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-dev