On May 13, 2011, at 2:02 PM, Adrian Crum wrote: > On 5/13/2011 1:53 PM, David E Jones wrote: >> On May 13, 2011, at 1:47 PM, Adrian Crum wrote: >> >>> On 5/13/2011 1:36 PM, David E Jones wrote: >>>> On May 13, 2011, at 1:28 PM, Adrian Crum wrote: >>>> >>>>> The CustRequestParty entity seems to be an implementation of the Request >>>>> Role Type entity in The Data Model Resource Book. Besides the name >>>>> difference, the only other difference is using Role Type instead of >>>>> Request Role Type. Reusing Role Type in this way is okay from my >>>>> perspective. The problem is, the CustRequestParty entity isn't related to >>>>> Role Type, instead it is related to PartyRole - which requires a >>>>> PartyRole entry. >>>>> >>>>> That is an extremely limiting relationship - a party can't be related to >>>>> a Request in a particular role unless they are already a member of that >>>>> role. >>>> Pretty much all *Role and *Party entities are setup this way, and in fact >>>> nearly all entities that have pairs of partyId and roleTypeId have a type >>>> one relationship to PartyRole. This is a pattern that goes back to the >>>> beginning of OFBiz and is used throughout the project. >>>> >>>> I agree with making the change so that all of these have fks to Party and >>>> RoleType separately, so not requiring an entry in PartyRole, but keep in >>>> mind that's a big change. I've actually done this in the Mantle UDM, but >>>> that was easy because there aren't any dependencies on that data model >>>> yet... for OFBiz it's a bit more work. >>>> >>>> BTW, this goes back to the original pattern for party roles where the >>>> concept was that a party being in a role (ie with a PartyRole record) >>>> means nothing, and roles should just be used to define how parties are >>>> related to other records in the system. However, no one seems to want to >>>> follow that pattern so by de facto practice it's a moot point, and IMO >>>> ideally we would get rid of PartyRole altogether, or use it for specific >>>> and limited circumstances. The reason is that 99% of the time someone >>>> comes up with a constraint like "Party X is in Role Y" they are forgetting >>>> other important details, like in Role Y for Record Z. >>>> >>> Thanks for the reply! >>> >>> I don't think this particular change is a big one. I suspect the >>> relationship can be changed without much fuss, but I will check into it >>> further. >>> >>> There shouldn't be any confusion about Party Role if we follow the author's >>> reasoning for it: It is intended to describe the party's role in the >>> enterprise or organization. In Request Role, the relationship being >>> described is a party's role in the request, not the party's role in the >>> enterprise. That's why Request Role is related directly to a Role Type and >>> not a Party Role. >> Yeah, this is exactly the sort of misunderstanding I'm referring to. You >> wrote "It is intended to describe the party's role in the enterprise or >> organization" and "not the party's role in the enterprise. That's why >> Request Role is related directly to a Role Type and not a Party Role." >> >> However, a record in PartyRole is NOT meant to represent a Party's role >> within the enterprise or organization, if you want to model that you should >> have a PartyRelationship record going between the Party record for the >> enterprise or organization... not a PartyRole that just ties a partyId to a >> roleTypeId without any consideration of the enterprise or organization. It's >> inflexible and generally bad modeling, and if something in The Data Model >> Resource Book seems to describe it this way I'd be surprised, chances are >> whatever you think means this really means something else. >> >> > > Right. I over-simplified the meaning of Party Role to demonstrate the > differences in this case. Thank you for the clarification. > > Semantics aside, we agree that the existing entity relationship I described > is incorrect, right?
Good question, I don't know that it's "incorrect"... but it is certainly cumbersome, and the initial point of that constraint (to help avoid the use of PartyRole without any context) is totally lost on probably every single person who uses OFBiz, with maybe just a couple of exceptions, so the pattern has failed in its intent. The main point of my first response is that the pattern is used in dozens of places, to see a list look at the WebTools Entity Reference for the PartyRole entity and look at all of the type "many" relationships. Every one of those follows the pattern you described for CustRequestParty. It's a good point that it is cumbersome and is of little use, and so we might as well get rid of those everywhere... -David
