[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [wsbpel] Dead Path Elimination and Join Conditions
I agree that the syntax is straighforward. It appears that we just
disagree on the semantics. I would prefer that another statement be in
the spec that says that an explicit join condition MUST NOT evaluate to
false in any case where the implicit join condition evaluates to true.
In other words, it can not be true when all incoming links are false. |
I think it is ugly to have a construct such as a transition that is meant to express control dependencies, and then allow its use for expressing "anti-dependencies."
Dieter Koenig1 wrote:
Hi Danny, I meant *any* join condition, either implicit or explicit. This applies to *every* activity, either with one or with more incoming links. This is also how I read the spec: "(...) propagating negative link status transitively along entire paths formed by consecutive links until a join condition is reached that evaluates to true." "Every activity that is the target of a link has an implicit or explicit “join condition” associated with it. This applies even when an activity has exactly one incoming link" "If the explicit join condition is missing, the implicit condition requires the status of at least one incoming link to be positive" Maybe I am still missing the point, but with these three statements in the spec, I see no more room for interpretation. Kind Regards DK Danny van der Rijn <firstname.lastname@example.org To > Dieter Koenig1/Germany/IBM@IBMDE cc 27.01.2005 19:51 email@example.com, firstname.lastname@example.org Subject Re: [wsbpel] Dead Path Elimination and Join Conditions Do you mean that DPE always ends at "an explicit join condition"? Every activity in a flow has at least an implicit join condition. While I agree that an implementation can do what you describe, the especially pathological case that I describe doesn't even meet the (implied) semantics of the word "join," since there is only one control path. I am bothered more by the ability to create extremely confusing semantics than by any non-implementability. Danny Dieter Koenig1 wrote:Formally, I see no problem as DPE always ends at a join condition, so the effect caused by the "SecondToThird" transition is perfectly valid. Maybe this is something that should be made more clear in the spec language. Kind Regards DKDanny van derRijn<email@example.com Tocc26.01.2005 22:23 firstname.lastname@example.orgSubjectRe: [wsbpel] Dead Path Eliminationand Join ConditionsThe contradiction is one of semantics. There can be no "dead paths" in such a case, since a join condition later in the path can "resuscitate"thepath. The sentinel case you describe can easily be coded differently, say as the 2nd activity in a sequence, where the first is a flow. After the flow completes, the sentinel can check conditions. Of course, it can't check link status, but I don't see that as a huge obstacle. Danny email@example.com wrote: Hello Danny: This is in contradiction with my understanding of dead-path elimination. I would prefer to disallow joinConditions whose expression does not require a true input in order that the join condition evaluate to true. Comments? I do not see how your example contradicts sections 12.5.1 (link semantics) or 12.5.2 (dead path elimination)? I think your example is strange but not pathological. Let us pretend the programmer does not like fault handlers and structured the process as a graph with one end activity: Third. In turn, the programmer wants activity "Third" to be a sentinel or assert of sorts, executing only if activity "Second" failed. If my understanding is correct, if "Second" executes and sets "secondToThird"'s transitionCode to true, Third's joinCondition will evaluate to false, not run, and the process is finished: after all nothing bad happened .... and this is what the programmer intended. Cheers, Andrew To unsubscribe from this mailing list (and be removed from the roster of the OASIS TC), go to http://www.oasis-open.org/apps/org/workgroup/wsbpel/members/leave_workgroup.php.To unsubscribe from this mailing list (and be removed from the roster of the OASIS TC), go to http://www.oasis-open.org/apps/org/workgroup/wsbpel/members/leave_workgroup.php .