[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: Figure 7.1 modified/no inheritance (was RE: [regrep-cc-review] BCC as RIM Association)
Please see comments below marked with <JMC>. Diego Ballvé wrote: > > > Please see comments below marked with <JMC>. > > > > Diego Ballvé wrote: > > > > > > > <JMC> > > > > Yes - I am now more convinced (based on the excellent > > information you > > > > gave recently) that we really may need to represent ASCC as an > > > > ExtrinsicObject rather than simply an Association with > > Slots. And the > > > > ASCC would itself be involved in 2 associations (to "join" > > > > the ACCs that > > > > it associates, as per p.12 example). > > > > </JMC> > > > > > > Joe, I'm supporting ASCC as RIM Association! Read the mails I wrote > > > this week. > > > > <JMC> > > I'm not so sure that "ASCC as a RIM Association" is the best approach. > > I'll expand on an response I gave to the e-mails you wrote this week: > > > > CCTS states that Stored Core Components have the following attributes: > > - Unique Identifier > > - Version > > - Dictionary Entry Name > > - Definition > > - Usage Rule > > > > My fundamental question is: Would an ASCC need to have > > different values > > for *at least one of the above attributes* than the BCC from > > which it is > > "derived"? In other words: > > > > - Is it possible that an ASCC may have a different Unique Identifier > > than the BCC from which it is derived? I would say it must. > > - Is it possible that an ASCC may have a different Version > > than the BCC > > from which it is derived? I would say it definitely may (i.e. we may > > want to version ASCCs separately from A. > > > > ...and so on for Dictionary Entry Name, etc. > > > > Having said that: > > > > If we represent ASCCs with Associations, how would we handle the > > scenarios I list above? > > </JMC> > > Exactly the same way as for ACC/BCC - and this is why we're insisting > on it. This is what I meant when pointing out that Associations are > RegistryObjects. It means that Associations have name, desccription > and slots, just like ExtrinsicObject (although they do not have > version - it would have to be a slot). <JMC> Please note that when I use the term "RegistryObject" I most often mean "ExtrinsicObject" rather than "Association" - I always use the term "Association" when I mean association. I truly believe that we should represent the following entities are RegistryObjects - meaning ExtrinsicObjects, not Associations: - BCCs - ACCs - BBIEs - ABIEs - Data Types - CCTs - Supplementary Components - Supplementary Component Restrictions - Content Component Restrictions ...and ASCCs as well. Any proposal to represent these as anything other than ExtrinsicObjects will be considered, but it will have to be (a) highly convincing (which I have not yet seen, to be quite honest), and (b) done very soon - by the end of this week (Friday). Sorry to be so rigid about this, but it's for the best interest of the effort so that we can keep moving forward. </JMC> > > > > I was discussing with Farrukh about it and he supports that > > > too. He asked for an UML diagram without inheritances cause it might > > > be easier to see things, so I'm attaching it (with red > > comments on how > > > we map each "class"). Note that fields are hidden. > > > > > > Also, Farrukh has questioned the need for Associations > > between objects > > > when a Slot containing the UUID can be enough... see the attachment. > > > > <JMC> > > I would welcome more details on this from Farrukh and/or yourself - my > > thought is: Couldn't one *implement* an Association as a Slot > > containing > > the UUID of the Target Object? I would say yes. In other > > words, without > > additional background on this it sounds like we're getting into > > implementation details here, beyond specification. > > </JMC> > > Not really. Slots, Associations and ExtrinsicObjecs are all part > of Registry Specs. If we specify that X is binded to Association, > somebody can *implement* it in any way he/she wants, although > when you get that object from the registry (out of the > implementation) using for instance SOAP interface and RIM/RS > Schemas, you must get an Association. This allows that 2 > implementations can understand eachothers (be compatible). <JMC> Thanks for your thoughts - I completely understand what you are proposing. However, I truly believe that the approach proposed here is not only way too complex for the task at hand (i.e. why "bind X to an Association" when we can simply prescribe use of an Association), it (in my opinion) ignores the Association feature of the registry for (in my opinion) no good reason. Please provide a stronger argument by the end of Friday for this, so that we may consider this approach vs. Associations. </JMC> > > Diego
begin:vcard n:Chiusano;Joseph tel;work:(703) 902-6923 x-mozilla-html:FALSE url:www.bah.com org:Booz | Allen | Hamilton;IT Digital Strategies Team adr:;;8283 Greensboro Drive;McLean;VA;22012; version:2.1 email;internet:chiusano_joseph@bah.com title:Senior Consultant fn:Joseph M. Chiusano end:vcard
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]