Issue 12417: GQAM Observers: inconsistency between domain view and UML representation (marte-ftf) Source: THALES (Mr. Sebastien Demathieu, sebastien.demathieu(at)thalesgroup.com) Nature: Revision Severity: Significant Summary: GQAM Observers: inconsistency between domain view and UML representation Details: In the domain view (Figure 15.4), the Timing/TimedObserver has two properties startObs and endObs, typed by MARTE::Time::TimedInstantObservation. In the UML profile diagram (Figure 15.8), the related GaTiming/TimedObserver has two properties startObs and endObs, typed by UML::TimeObservation. In the UML profile description (p.281), the same GaTiming/TimedObserver has two properties startObs and endObs, typed by MARTE::Time::TimedInstantObservation. We need to choose whether a timed observer relates to a MARTE TimedInstantObservation or UML TimeObservation, and then to align the domain view, profile definition and profile description accordingly. Resolution: In the conceptual domain model, we rely in MARTE "concepts", independently of whatever UML construct. For that reason we use the concept coming from the Time chapter. However, in the profile view, it is permissible to use the UML corresponding metaclass, because there is no reason to constraint such TimedObservers to Observations stereotyped with the MARTE Time concepts. (The time concepts add the notion of clocked Observation (attribute "clock"), which are not essential here). We agreed with Charles Andre that such a stereotype should not be mandatory when defining a TimedObserver. Revised Text: see ptc/2009-05-12 pages 203 - 204 Actions taken: April 24, 2008: received issue October 16, 2009: closed issue Discussion: End of Annotations:===== From: webmaster@omg.org Date: 24 Apr 2008 05:21:56 -0400 To: Subject: Issue/Bug Report -------------------------------------------------------------------------------- Name: Sébastien Demathieu Company: Thales mailFrom: sebastien.demathieu@thalesgroup.com Notification: Yes Specification: UML profile for MARTE Section: 15.3.1 FormalNumber: 07-08-04 Version: Beta 1 RevisionDate: August 2007 Page: 272 Nature: Revision Severity: Significant HTTP User Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; fr; rv:1.8.1.14) Gecko/20080404 Firefox/2.0.0.14 Description GQAM Observers: inconsistency between domain view and UML representation Details: In the domain view (Figure 15.4), the Timing/TimedObserver has two properties startObs and endObs, typed by MARTE::Time::TimedInstantObservation. In the UML profile diagram (Figure 15.8), the related GaTiming/TimedObserver has two properties startObs and endObs, typed by UML::TimeObservation. In the UML profile description (p.281), the same GaTiming/TimedObserver has two properties startObs and endObs, typed by MARTE::Time::TimedInstantObservation. We need to choose whether a timed observer relates to a MARTE TimedInstantObservation or UML TimeObservation, and then to align the domain view, profile definition and profile description accordingly. Subject: RE: RE : [Issue 12417 ] Date: Fri, 30 Jan 2009 10:41:57 +0100 X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: RE : [Issue 12417 ] Thread-Index: AcmAh/x7M4t2vRskTFqmY/VlVHGrLwBbumrAADG/OKA= From: "GERARD Sebastien 166342" To: "marte-ftf" Cc: Sébastien Demathieu , , "ESPINOZA Huascar 218344" , X-OriginalArrivalTime: 30 Jan 2009 09:41:57.0998 (UTC) FILETIME=[FE61A8E0:01C982BE] X-MIME-Autoconverted: from quoted-printable to 8bit by amethyst.omg.org id n0U9gtSZ010558 Hi all, Just to remind you what is the domain model and what is the profile model, and then why it may exist differences between both models: - The domain model: it is a meta-model that denotes the specification of the domain specific language under consideration. It is a refinement of the Request For Proposal (RFP) which denotes the requirements description. It is also called a specification of the domain specific language. - The profile model (in our case the UML profile model): it is a model denoting the design of a specific domain modelling language in terms of UML extensions that is conformed to its domain model. The domain model is then not at the same level of abstraction than the profile dome. It describes the "what", whereas the profile model describes the "how". Both models do not need then to map 1-1. Cheers... Séb. -----Message d'origine----- De : ESPINOZA Huascar 218344 Envoyé : jeudi 29 janvier 2009 10:53 À : 'cmw@sce.carleton.ca' Cc : Sébastien Demathieu; Charles.Andre@sophia.inria.fr; GERARD Sebastien 166342 Objet : RE: RE : [Issue 12417 ] Murray, I would let as is, if there is no a proposal to solve the other issues (not UML in the domain model, and not constrant on the use of Observations together with TimedObserver) You could add a note explaining the mapping domain model vs. profile. In general, the mismatch domain model vs. profile exists through the spec. due to different reasons. A more generic solution (a lot of work) would be to provide a mapping (A table for example) of domain model metaclasses/attributes to Stereotypes. Cheers, Huascar -----Message d'origine----- De : Murray Woodside [mailto:cmw@sce.carleton.ca] Envoyé : mardi 27 janvier 2009 15:03 À : ESPINOZA Huascar 218344 Cc : Sébastien Demathieu; Charles.Andre@sophia.inria.fr Objet : RE: RE : [Issue 12417 ] It seems clear to me that we should always use UML concepts where they apply, and that they apply here, so I would go for UML. Murray Woodside Distinguished Research Professor Dept of Systems and Computer Engineering, Carleton University, 1125 Colonel By Drive, Ottawa K1S 5B6, Canada. (613)-520-5721.....fax (613)-520-5727....cmw@sce.carleton.ca (http://www.sce.carleton.ca/faculty/woodside.html) On Tue, 27 Jan 2009, ESPINOZA Huascar 218344 wrote: > > Séb, please read the reasons I explained for both choices. > > Do you have a concrete proposal to avoid that? According to that, we would be able to decide which one to align. > > Thank you, > Huascar > > -----Message d'origine----- > De : Sébastien Demathieu [mailto:sebastien.demathieu@thalesgroup.com] > Envoyé : mardi 27 janvier 2009 09:07 > À : ESPINOZA Huascar 218344 > Cc : cmw@sce.carleton.ca; Charles.Andre@sophia.inria.fr > Objet : Re: RE : [Issue 12417 ] > > Dear all, > > I strongly support the idea to have the domain model and the UML view > kept consistent. > > The domain model is supposed to provide clear definitions / semantics > for the stereotypes > introduced in the profile. If the gap is too big between the domain > model and the UML > view, it creates more confusion than if we would had no domain model at all. > > Therefore, I would second Murray's statement: We need to choose whether > a timed > observer relates to a MARTE TimedInstantObservation or UML TimeObservation, > and then to align the domain view, profile definition and profile > description accordingly. > > Thanks, > > Sébastien > > > > ESPINOZA Huascar 218344 a écrit : >> Dear Murray, >> >> I know this issue. However, there is an explanation for such a choice. >> >> In the conceptual domain model, we rely in MARTE "concepts", independently of whatever UML construct. For that reason we use the concept coming from the TIme chapter. >> >> In the profile view, we use the UML corresponding metaclass, because there is no reason to constraint such TimedObservers to Observations stereotyped with the MARTE Time concepts. The time concepts add the notion of clocked Observation (attribute "clock"). We agreed with Charles Andre that such a stereotype should not be mandatory when defining a TimedObserver. >> >> So, domain model and profile are different (as many parts in the spec.). I would prefer otherwise, but... >> >> I'd propose to close this issue without change... if you don't have a better solution?? >> >> Cheers, >> Huascar >> >> ________________________________ >> >> De: Murray Woodside [mailto:cmw@sce.carleton.ca] >> Date: lun. 26/01/2009 17:52 >> À: ESPINOZA Huascar 218344 >> Objet : >> >> >> >> Issue 12417 is about the type of timed observations in >> theGaTimingObserver. The issue text is >> >> GQAM Observers: inconsistency between domain view and UML representation >> Details: In the domain view (Figure 15.4), the Timing/TimedObserver has >> two properties startObs and endObs, typed by >> MARTE::Time::TimedInstantObservation. In the UML profile diagram (Figure >> 15.8), the related GaTiming/TimedObserver has two properties startObs and >> endObs, typed by UML::TimeObservation. In the UML profile description >> (p.281), the same GaTiming/TimedObserver has two properties startObs and >> endObs, typed by MARTE::Time::TimedInstantObservation. We need to choose >> whether a timed observer relates to a MARTE TimedInstantObservation or UML >> TimeObservation, and then to align the domain view, profile definition and >> profile description accordingly. >> >> You designed the timing observers, so do you know which is correct? I >> presume the MARTE type... >> >> Murray Woodside >> >> Distinguished Research Professor >> Dept of Systems and Computer Engineering, >> Carleton University, 1125 Colonel By Drive, Ottawa K1S 5B6, Canada. >> (613)-520-5721.....fax (613)-520-5727....cmw@sce.carleton.ca >> (http://www.sce.carleton.ca/faculty/woodside.html) >> >> >> >> > Date: Fri, 30 Jan 2009 11:17:44 +0100 From: Sébastien Demathieu Organization: Thales Research & Technology User-Agent: Thunderbird 1.5.0.9 (Windows/20061207) To: GERARD Sebastien 166342 CC: marte-ftf , Charles.Andre@sophia.inria.fr, ESPINOZA Huascar 218344 , cmw@sce.carleton.ca Subject: Re: RE : [Issue 12417 ] Hi Sébastien, I agree with you: the domain model and the profile definition are not at the same level of abstraction. However, even if the mapping between both is not 1-1, the profile model (the "how") shall not be contradictory with the domain model (the "what"). That seems to be unfortunately the case in the context of issue 12417, where TimedObserver references a MARTE::TimeInstantObservation while its mapping in UML: GaTimedObs references a UML::TimeObservation. Sébastien GERARD Sebastien 166342 a écrit : Hi all, Just to remind you what is the domain model and what is the profile model, and then why it may exist differences between both models: - The domain model: it is a meta-model that denotes the specification of the domain specific language under consideration. It is a refinement of the Request For Proposal (RFP) which denotes the requirements description. It is also called a specification of the domain specific language. - The profile model (in our case the UML profile model): it is a model denoting the design of a specific domain modelling language in terms of UML extensions that is conformed to its domain model. The domain model is then not at the same level of abstraction than the profile dome. It describes the "what", whereas the profile model describes the "how". Both models do not need then to map 1-1. Cheers... Séb. -----Message d'origine----- De : ESPINOZA Huascar 218344 Envoyé : jeudi 29 janvier 2009 10:53 À : 'cmw@sce.carleton.ca' Cc : Sébastien Demathieu; Charles.Andre@sophia.inria.fr; GERARD Sebastien 166342 Objet : RE: RE : [Issue 12417 ] Murray, I would let as is, if there is no a proposal to solve the other issues (not UML in the domain model, and not constrant on the use of Observations together with TimedObserver) You could add a note explaining the mapping domain model vs. profile. In general, the mismatch domain model vs. profile exists through the spec. due to different reasons. A more generic solution (a lot of work) would be to provide a mapping (A table for example) of domain model metaclasses/attributes to Stereotypes. Cheers, Huascar -----Message d'origine----- De : Murray Woodside [mailto:cmw@sce.carleton.ca] Envoyé : mardi 27 janvier 2009 15:03 À : ESPINOZA Huascar 218344 Cc : Sébastien Demathieu; Charles.Andre@sophia.inria.fr Objet : RE: RE : [Issue 12417 ] It seems clear to me that we should always use UML concepts where they apply, and that they apply here, so I would go for UML. Murray Woodside Distinguished Research Professor Dept of Systems and Computer Engineering, Carleton University, 1125 Colonel By Drive, Ottawa K1S 5B6, Canada. (613)-520-5721.....fax (613)-520-5727....cmw@sce.carleton.ca (http://www.sce.carleton.ca/faculty/woodside.html) On Tue, 27 Jan 2009, ESPINOZA Huascar 218344 wrote: Séb, please read the reasons I explained for both choices. Do you have a concrete proposal to avoid that? According to that, we would be able to decide which one to align. Thank you, Huascar -----Message d'origine----- De : Sébastien Demathieu [mailto:sebastien.demathieu@thalesgroup.com] Envoyé : mardi 27 janvier 2009 09:07 À : ESPINOZA Huascar 218344 Cc : cmw@sce.carleton.ca; Charles.Andre@sophia.inria.fr Objet : Re: RE : [Issue 12417 ] Dear all, I strongly support the idea to have the domain model and the UML view kept consistent. The domain model is supposed to provide clear definitions / semantics for the stereotypes introduced in the profile. If the gap is too big between the domain model and the UML view, it creates more confusion than if we would had no domain model at all. Therefore, I would second Murray's statement: We need to choose whether a timed observer relates to a MARTE TimedInstantObservation or UML TimeObservation, and then to align the domain view, profile definition and profile description accordingly. Thanks, Sébastien ESPINOZA Huascar 218344 a écrit : Dear Murray, I know this issue. However, there is an explanation for such a choice. In the conceptual domain model, we rely in MARTE "concepts", independently of whatever UML construct. For that reason we use the concept coming from the TIme chapter. In the profile view, we use the UML corresponding metaclass, because there is no reason to constraint such TimedObservers to Observations stereotyped with the MARTE Time concepts. The time concepts add the notion of clocked Observation (attribute "clock"). We agreed with Charles Andre that such a stereotype should not be mandatory when defining a TimedObserver. So, domain model and profile are different (as many parts in the spec.). I would prefer otherwise, but... I'd propose to close this issue without change... if you don't have a better solution?? Cheers, Huascar ________________________________ De: Murray Woodside [mailto:cmw@sce.carleton.ca] Date: lun. 26/01/2009 17:52 À: ESPINOZA Huascar 218344 Objet : Issue 12417 is about the type of timed observations in theGaTimingObserver. The issue text is GQAM Observers: inconsistency between domain view and UML representation Details: In the domain view (Figure 15.4), the Timing/TimedObserver has two properties startObs and endObs, typed by MARTE::Time::TimedInstantObservation. In the UML profile diagram (Figure 15.8), the related GaTiming/TimedObserver has two properties startObs and endObs, typed by UML::TimeObservation. In the UML profile description (p.281), the same GaTiming/TimedObserver has two properties startObs and endObs, typed by MARTE::Time::TimedInstantObservation. We need to choose whether a timed observer relates to a MARTE TimedInstantObservation or UML TimeObservation, and then to align the domain view, profile definition and profile description accordingly. You designed the timing observers, so do you know which is correct? I presume the MARTE type... Murray Woodside Distinguished Research Professor Dept of Systems and Computer Engineering, Carleton University, 1125 Colonel By Drive, Ottawa K1S 5B6, Canada. (613)-520-5721.....fax (613)-520-5727....cmw@sce.carleton.ca (http://www.sce.carleton.ca/faculty/woodside.html) sebastien.demathieu1.vcf Subject: RE: RE : [Issue 12417 ] Date: Fri, 30 Jan 2009 11:29:18 +0100 X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: RE : [Issue 12417 ] Thread-Index: AcmCxAKdjYgIYlBjRSCRmEcI09ytKAAAYlTg From: "GERARD Sebastien 166342" To: Sébastien Demathieu Cc: "marte-ftf" , , "ESPINOZA Huascar 218344" , X-OriginalArrivalTime: 30 Jan 2009 10:29:18.0321 (UTC) FILETIME=[9B588E10:01C982C5] X-MIME-Autoconverted: from quoted-printable to 8bit by amethyst.omg.org id n0UATMWd022466 Ok, it sounds fair in this case. -----Message d'origine----- De : Sébastien Demathieu [mailto:sebastien.demathieu@thalesgroup.com] Envoyé : vendredi 30 janvier 2009 11:18 À : GERARD Sebastien 166342 Cc : marte-ftf; Charles.Andre@sophia.inria.fr; ESPINOZA Huascar 218344; cmw@sce.carleton.ca Objet : Re: RE : [Issue 12417 ] Hi Sébastien, I agree with you: the domain model and the profile definition are not at the same level of abstraction. However, even if the mapping between both is not 1-1, the profile model (the "how") shall not be contradictory with the domain model (the "what"). That seems to be unfortunately the case in the context of issue 12417, where TimedObserver references a MARTE::TimeInstantObservation while its mapping in UML: GaTimedObs references a UML::TimeObservation. Sébastien GERARD Sebastien 166342 a écrit : > Hi all, > > Just to remind you what is the domain model and what is the profile model, and then why it may exist differences between both models: > - The domain model: it is a meta-model that denotes the specification of the domain specific language under consideration. It is a refinement of the Request For Proposal (RFP) which denotes the requirements description. It is also called a specification of the domain specific language. > > - The profile model (in our case the UML profile model): it is a model denoting the design of a specific domain modelling language in terms of UML extensions that is conformed to its domain model. > > The domain model is then not at the same level of abstraction than the profile dome. It describes the "what", whereas the profile model describes the "how". Both models do not need then to map 1-1. > > Cheers... Séb. > > -----Message d'origine----- > De : ESPINOZA Huascar 218344 > Envoyé : jeudi 29 janvier 2009 10:53 > À : 'cmw@sce.carleton.ca' > Cc : Sébastien Demathieu; Charles.Andre@sophia.inria.fr; GERARD Sebastien 166342 > Objet : RE: RE : [Issue 12417 ] > > Murray, > > I would let as is, if there is no a proposal to solve the other issues (not UML in the domain model, and not constrant on the use of Observations together with TimedObserver) > > You could add a note explaining the mapping domain model vs. profile. > > In general, the mismatch domain model vs. profile exists through the spec. due to different reasons. A more generic solution (a lot of work) would be to provide a mapping (A table for example) of domain model metaclasses/attributes to Stereotypes. > > Cheers, > Huascar > > -----Message d'origine----- > De : Murray Woodside [mailto:cmw@sce.carleton.ca] > Envoyé : mardi 27 janvier 2009 15:03 > À : ESPINOZA Huascar 218344 > Cc : Sébastien Demathieu; Charles.Andre@sophia.inria.fr > Objet : RE: RE : [Issue 12417 ] > > It seems clear to me that we should always use UML concepts where they > apply, and that they apply here, so I would go for UML. > > Murray Woodside > > Distinguished Research Professor > Dept of Systems and Computer Engineering, > Carleton University, 1125 Colonel By Drive, Ottawa K1S 5B6, Canada. > (613)-520-5721.....fax (613)-520-5727....cmw@sce.carleton.ca > (http://www.sce.carleton.ca/faculty/woodside.html) > > > On Tue, 27 Jan 2009, ESPINOZA Huascar 218344 wrote: > > >> Séb, please read the reasons I explained for both choices. >> >> Do you have a concrete proposal to avoid that? According to that, we would be able to decide which one to align. >> >> Thank you, >> Huascar >> >> -----Message d'origine----- >> De : Sébastien Demathieu [mailto:sebastien.demathieu@thalesgroup.com] >> Envoyé : mardi 27 janvier 2009 09:07 >> À : ESPINOZA Huascar 218344 >> Cc : cmw@sce.carleton.ca; Charles.Andre@sophia.inria.fr >> Objet : Re: RE : [Issue 12417 ] >> >> Dear all, >> >> I strongly support the idea to have the domain model and the UML view >> kept consistent. >> >> The domain model is supposed to provide clear definitions / semantics >> for the stereotypes >> introduced in the profile. If the gap is too big between the domain >> model and the UML >> view, it creates more confusion than if we would had no domain model at all. >> >> Therefore, I would second Murray's statement: We need to choose whether >> a timed >> observer relates to a MARTE TimedInstantObservation or UML TimeObservation, >> and then to align the domain view, profile definition and profile >> description accordingly. >> >> Thanks, >> >> Sébastien >> >> >> >> ESPINOZA Huascar 218344 a écrit : >> >>> Dear Murray, >>> >>> I know this issue. However, there is an explanation for such a choice. >>> >>> In the conceptual domain model, we rely in MARTE "concepts", independently of whatever UML construct. For that reason we use the concept coming from the TIme chapter. >>> >>> In the profile view, we use the UML corresponding metaclass, because there is no reason to constraint such TimedObservers to Observations stereotyped with the MARTE Time concepts. The time concepts add the notion of clocked Observation (attribute "clock"). We agreed with Charles Andre that such a stereotype should not be mandatory when defining a TimedObserver. >>> >>> So, domain model and profile are different (as many parts in the spec.). I would prefer otherwise, but... >>> >>> I'd propose to close this issue without change... if you don't have a better solution?? >>> >>> Cheers, >>> Huascar >>> >>> ________________________________ >>> >>> De: Murray Woodside [mailto:cmw@sce.carleton.ca] >>> Date: lun. 26/01/2009 17:52 >>> À: ESPINOZA Huascar 218344 >>> Objet : >>> >>> >>> >>> Issue 12417 is about the type of timed observations in >>> theGaTimingObserver. The issue text is >>> >>> GQAM Observers: inconsistency between domain view and UML representation >>> Details: In the domain view (Figure 15.4), the Timing/TimedObserver has >>> two properties startObs and endObs, typed by >>> MARTE::Time::TimedInstantObservation. In the UML profile diagram (Figure >>> 15.8), the related GaTiming/TimedObserver has two properties startObs and >>> endObs, typed by UML::TimeObservation. In the UML profile description >>> (p.281), the same GaTiming/TimedObserver has two properties startObs and >>> endObs, typed by MARTE::Time::TimedInstantObservation. We need to choose >>> whether a timed observer relates to a MARTE TimedInstantObservation or UML >>> TimeObservation, and then to align the domain view, profile definition and >>> profile description accordingly. >>> >>> You designed the timing observers, so do you know which is correct? I >>> presume the MARTE type... >>> >>> Murray Woodside >>> >>> Distinguished Research Professor >>> Dept of Systems and Computer Engineering, >>> Carleton University, 1125 Colonel By Drive, Ottawa K1S 5B6, Canada. >>> (613)-520-5721.....fax (613)-520-5727....cmw@sce.carleton.ca >>> (http://www.sce.carleton.ca/faculty/woodside.html) >>> >>> >>> >>>