Re: ISSUE-66 (Semantics of actions): Operational semantics of actions as covered by PRD? [PRD ]

I replied to a similar issue that we need Modify.  Actually, I'm not so 
sure.  Maybe Retract + Assert are fine.  I think the semantics of 
actions should be that the sequence of actions in the Do(ACTION*) part 
of the rule is executed "atomically" in the sense that rule conditions 
are only evaluated after all the ACTIONs execute.  I think all PR 
engines work this way.

I don't know how to handle Execute except to say that it MUST NOT change 
the KB.  It effectively becomes a way to compute some new variable 
values that subsequent actions will use to change the KB. But then, it 
may as well be a "user defined builtin" in the condition part.  And this 
would be more compatible with BLD. 

Rule Interchange Format Working Group Issue Tracker wrote:
> ISSUE-66 (Semantics of actions): Operational semantics of actions as covered by PRD? [PRD ]
>
> http://www.w3.org/2005/rules/wg/track/issues/
>
> Raised by: Christian de Sainte Marie
> On product: PRD 
>
> - What should is the intended semantics of the actions covered by PRD?
> - What when the intended action is the creation or the deletion of an individual (object): is that part of the semantics of Assert/Retract? How?
> - If (when) PRD covers some kind of Modify action, how should the intended semantics differ from Retract+Assert?
> - Any kind of Execute action is, by definition, opaque: how should its semantics be specified, esp. wrt possible side effects on the state of facts wrt which the semantics of a PR system is specified in PRD?
> - etc...
>
>
>
>

Received on Tuesday, 24 June 2008 06:05:46 UTC