Issue 15167: composite tags (uml2-rtf) Source: No Magic, Inc. (Mr. Nerijus Jankevicius, nerijus(at)nomagic.com) Nature: Uncategorized Issue Severity: Summary: Are composite tag definitions allowed? Resolution: Revised Text: Actions taken: April 6, 2010: received issue Discussion: End of Annotations:===== ted-NM: yes From: "Nerijus Jankevicius" To: Subject: composite tags Date: Tue, 6 Apr 2010 20:30:04 +0300 X-Mailer: Microsoft Outlook Express 6.00.2900.5843 X-EsetScannerBuild: 6883 Anyone can help on subject? Are composite tag definitions allowed? See attached document with problem statement. Thanks in advance, -- Nerijus Jankevicius SysML Product Manager OMG-Certified UML Professional No Magic Europe Savanoriu pr. 363, LT 49425 Kaunas, Lithuania P.O. box 2166, LT- 3000, Kaunas Phone: +370-37-324032 Fax: +370-37-320670 e-mail: nerijus@magicdraw.com WWW: http://www.magicdraw.com -- MagicDraw - Architecture made simple! Tagged_values_with_aggregation_composite.pdf DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:sender:received:in-reply-to :references:from:date:x-google-sender-auth:received:message-id :subject:to:cc:content-type; bh=UUIQj3Pn5KYKK+1lslt8W1Ry+/ZA7daIOaBntLWJb5o=; b=KtYdR+FnTb3HJgQkcnYnRar59lPGYlk1HwXUrsg3Mj3t4NwdZz3ledvwLbtR7GdjP6 WYMC10afS7vlH2n6yxsUEWGtBofFemj8Iga/ImAbv8N8ZJnG5GLt1KchcI07KByGp4Fv iT9dOT+x9kRdq9CLepX4VcyqjHPwf9YDq+LJs= DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:sender:in-reply-to:references:from:date :x-google-sender-auth:message-id:subject:to:cc:content-type; b=gQkmzb+E63Tg4Vg9VyGKQOky+kDqTmCVMFr3pKMyeoXlWTx8UEV2kQG+ESo3etRrpz Kc9IpmkxZ88a0ZIvbOK41NbT/z/sR6yedSAZ1eDQ+B7aRM69d5YoeStbl0KW9U6T7GI8 QZQM02Qg88513sZeW43FelZe7Rn7/zEFA55s4= Sender: bran.selic@gmail.com From: Bran Selic Date: Tue, 6 Apr 2010 19:50:51 +0200 X-Google-Sender-Auth: cabf4362fee4d28c Subject: Re: composite tags To: Nerijus Jankevicius Cc: uml2-rtf@omg.org Good question, Narijus. I hadn't thought of it before. However, once we opened the door for adding meta-associations in a profile, I see no reason to exclude this possibility (at least not on first analysis). Can you explain the reason why you folks thought it should not be possible? Cheers...Bran 2010/4/6 Nerijus Jankevicius Anyone can help on subject? Are composite tag definitions allowed? See attached document with problem statement. Thanks in advance, -- Nerijus Jankevicius SysML Product Manager OMG-Certified UML Professional No Magic Europe Savanoriu pr. 363, LT 49425 Kaunas, Lithuania P.O. box 2166, LT- 3000, Kaunas Phone: +370-37-324032 Fax: +370-37-320670 e-mail: nerijus@magicdraw.com WWW: http://www.magicdraw.com -- MagicDraw - Architecture made simple! X-Trusted-NM: yes From: "Nerijus Jankevicius" To: "Bran Selic" Cc: Subject: Re: composite tags Date: Wed, 7 Apr 2010 17:43:42 +0300 X-Mailer: Microsoft Outlook Express 6.00.2900.5843 X-EsetScannerBuild: 6889 Bran, Imagine I can define new composition between stereotyped Class and Package, so Class is able to own a Package. It is not possible to own such element in UML metamodel, I have no ideas how it should be saved into XMI also. Is it standard way to store such elements in "profile application" part of XMI as Eclipse do? How do you suppose it should work? More primitive case - when composition is defined between stereotypes which metaclasses can own each other in UML. In this case composite property should subset some standard property, like "ownedAttributes" or similar. But, the problem is that Extension is not Generalization, so there is no inheritance (between metaclass and stereotype) and subsets and redefinitions can't be used. Regards, -- Nerijus Jankevicius SysML Product Manager OMG-Certified UML Professional No Magic Europe Savanoriu pr. 363, LT 49425 Kaunas, Lithuania P.O. box 2166, LT- 3000, Kaunas Phone: +370-37-324032 Fax: +370-37-320670 e-mail: nerijus@magicdraw.com WWW: http://www.magicdraw.com -- MagicDraw - Architecture made simple! ----- Original Message ----- From: Bran Selic To: Nerijus Jankevicius Cc: uml2-rtf@omg.org Sent: Tuesday, April 06, 2010 8:50 PM Subject: Re: composite tags Good question, Narijus. I hadn't thought of it before. However, once we opened the door for adding meta-associations in a profile, I see no reason to exclude this possibility (at least not on first analysis). Can you explain the reason why you folks thought it should not be possible? Cheers...Bran 2010/4/6 Nerijus Jankevicius Anyone can help on subject? Are composite tag definitions allowed? See attached document with problem statement. Thanks in advance, -- Nerijus Jankevicius SysML Product Manager OMG-Certified UML Professional No Magic Europe Savanoriu pr. 363, LT 49425 Kaunas, Lithuania P.O. box 2166, LT- 3000, Kaunas Phone: +370-37-324032 Fax: +370-37-320670 e-mail: nerijus@magicdraw.com WWW: http://www.magicdraw.com -- MagicDraw - Architecture made simple! Date: Wed, 07 Apr 2010 17:56:26 +0100 From: Dave Hawkins User-Agent: Thunderbird 2.0.0.6 (X11/20070728) To: Nerijus Jankevicius CC: Bran Selic , uml2-rtf@omg.org Subject: Re: composite tags X-Source-IP: acsmt354.oracle.com [141.146.40.154] X-Auth-Type: Internal IP X-CT-RefId: str=0001.0A090207.4BBCB94F.01E1:SCFMA4539814,ss=1,fgs=0 If the composition is from a stereotype of Class to a Package I don't see any issue with doing that. The XMI would be of a similar form to the example you give in your PDF, eg: Eclipse isn't really storing this in the "profile application" section of the XMI. The stereotype instances could appear anywhere at the root level of the XMI. I'd quite like to see stereotype instances owned by the relevant profile application, ie: Many of the profiles I've seen defined seem to be making ExtensionEnds composite, which essentially means that the stereotype instance is owned by the object it is extending. This is much more like the issue you are describing in that objects unknowingly have ownership of other objects, which raises the issue of how an implementation should cope with that situation. In general I don't think it's a good idea for owned ends of meta-associations to be composite. In fact, I think this should be invalid in MOF. Cheers, Dave Nerijus Jankevicius wrote: Bran, Imagine I can define new composition between stereotyped Class and Package, so Class is able to own a Package. It is not possible to own such element in UML metamodel, I have no ideas how it should be saved into XMI also. Is it standard way to store such elements in "profile application" part of XMI as Eclipse do? How do you suppose it should work? More primitive case - when composition is defined between stereotypes which metaclasses can own each other in UML. In this case composite property should subset some standard property, like "ownedAttributes" or similar. But, the problem is that Extension is not Generalization, so there is no inheritance (between metaclass and stereotype) and subsets and redefinitions can't be used. Regards, -- Nerijus Jankevicius SysML Product Manager OMG-Certified UML Professional No Magic Europe Savanoriu pr. 363, LT 49425 Kaunas, Lithuania P.O. box 2166, LT- 3000, Kaunas Phone: +370-37-324032 Fax: +370-37-320670 e-mail: nerijus@magicdraw.com WWW: http://www.magicdraw.com -- MagicDraw - Architecture made simple! ----- Original Message ----- *From:* Bran Selic *To:* Nerijus Jankevicius *Cc:* uml2-rtf@omg.org *Sent:* Tuesday, April 06, 2010 8:50 PM *Subject:* Re: composite tags Good question, Narijus. I hadn't thought of it before. However, once we opened the door for adding meta-associations in a profile, I see no reason to exclude this possibility (at least not on first analysis). Can you explain the reason why you folks thought it should not be possible? Cheers...Bran 2010/4/6 Nerijus Jankevicius > Anyone can help on subject? Are composite tag definitions allowed? See attached document with problem statement. Thanks in advance, -- Nerijus Jankevicius SysML Product Manager OMG-Certified UML Professional No Magic Europe Savanoriu pr. 363, LT 49425 Kaunas, Lithuania P.O. box 2166, LT- 3000, Kaunas Phone: +370-37-324032 Fax: +370-37-320670 e-mail: nerijus@magicdraw.com WWW: http://www.magicdraw.com -- MagicDraw - Architecture made simple! -- Dave Hawkins | Principal Software Engineer | +44 118 924 0022 Oracle JDeveloper Development Oracle Corporation UK Ltd is a company incorporated in England & Wales. Company Reg. No. 1782505. Reg. office: Oracle Parkway, Thames Valley Park, Reading RG6 1RA. DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:sender:received:in-reply-to :references:from:date:x-google-sender-auth:received:message-id :subject:to:cc:content-type; bh=bcq04lrhORaq2t9slArmR2qSVapG7s36rk8v2r5Auec=; b=mfunwj4cFjQVhwBi5x3iiEgZRPCzzXKOAM0raY4JLbpUfvy1+DShAkendNcINcvhdL zowKzo2dxhHcA18ehr8mea+vOPrfW1qFPuuXCbum0+s1+PAi9mer85ecUQHHV7d6abs2 /TVale30Jd9OWzSedpe9mzt+39sBIVmbU8SPs= DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:sender:in-reply-to:references:from:date :x-google-sender-auth:message-id:subject:to:cc:content-type; b=d+RUHtUyo3UHlPkoJpp7Qn/Z9wXPLuH/OnBC4jJiGkYkdHqx+cKiDz2P5V7a2PU27Z iDf7iaBJTCklldl+ISi3rM0D06PaFk8CPHIMb1VYHpI+4jt3//3FJ0KHS3VsrpNBIl3q wdVJang4z2AGvdtzwgFEDKiD/DCJUhvphMObo= Sender: bran.selic@gmail.com From: Bran Selic Date: Thu, 8 Apr 2010 09:57:25 +0200 X-Google-Sender-Auth: 744edcc509368a3e Subject: Re: composite tags To: Nerijus Jankevicius Cc: uml2-rtf@omg.org On Wed, Apr 7, 2010 at 4:43 PM, Nerijus Jankevicius wrote: Bran, Imagine I can define new composition between stereotyped Class and Package, so Class is able to own a Package. [bvs] Well, we did open up a Pandora's box when we allowed the addition of new meta-associations in profile tags. The problem of the possibility of non-sensical meta-relationships exists whether or not they involve composition. It is not possible to own such element in UML metamodel, [bvs] But, this is precisely the issue that was created once we allowed meta-associations in profiles. So, your question really has to do with the broader issue of whether we should allow typing of tags by meta-classes. (Too late to change this, I suspect.) I have no ideas how it should be saved into XMI also. Is it standard way to store such elements in "profile application" part of XMI as Eclipse do? How do you suppose it should work? [bvs] Ask the IBM folks -- apparently they have a way, since it seems that they deal with it as most people would expect. [bvs] Don't get me wrong, I think there is an issue here. However, I suspect that we cannot do much about it without redesigning the profile mechanism. Unfortunately, when we defined profiles, we really had no deep understanding of the mechanism or any theoretical foundations behind them. One of the principles was that a profile should be compatible with UML, but we did not state clearly what is meant by "compatible". In UML 1, we were a bit more conservative and did not allow tags to be typed by meta-classes, but we changed our minds in UML 2, without really reflecting too much on how this impacted compatibility. [bvs] I realize this is not helping your problem, but, perhaps we can get someone to do proper research into the profile concept that would allow us to do it right in the future. Cheers...Bran Subject: RE: composite tags Date: Thu, 8 Apr 2010 07:50:06 -0700 X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: composite tags Thread-Index: AcrXEuy44YKKFeL1TYiEYAFtUOvGiwAEGrtQ From: "Pete Rivett" To: "Nerijus Jankevicius" , "Ed Seidewitz" Cc: , Ø Imagine I can define new composition between stereotyped Class and Package, so Class is able to own a Package. Not sure what you mean here . are you defining a composite association between the stereotypes? In which case that.s invalid since a stereotype instance can only be owned by the element to which it.s applied. Or are you talking about defining a new association, owned by the Profile, between the UML metaclasses? With both ends owned by the Association? Ø I have no ideas how it should be saved into XMI also. Ø Is it standard way to store such elements in "profile application" part of XMI as Eclipse do? Ø How do you suppose it should work? Assuming nothing else owned a given package that would in theory be possible and the XMI serialization rules do cover this. If the profile is named Pro with namespace prefix the same and the association is named A_C_P with ends c:UML::Class and p:UML::Package then you.d have the following XMI element for having class with id c1 owning package with id p1 (assuming they are in the same XMI file)

Note that this form of serialization is the same regardless of whether the association is composite. Though a tool would have to respect the composite semantics and delete p1 if c1 is deleted. And prevent p1 being owned by something else at the same time. Note that there would be no notation defined for this meta-association so it would not be possible to draw it. unless a given UML tool has a DSL-like capability . you.d have to use property sheets. I see this as pushing the boundaries so do not see it as important for model interchange testing . do you have a scenario where you see this being useful? Have you come across a Profile doing this? Pete From: Nerijus Jankevicius [mailto:nerijus@nomagic.com] Sent: Thursday, April 08, 2010 4:59 AM To: Ed Seidewitz; Pete Rivett Cc: model-interchange@omg.org Subject: Fw: composite tags Ed, Pete, any feedback on the issue below? It is highly related to Model Interchange and any Profile develpment. Thanks. Regards, -- Nerijus Jankevicius SysML Product Manager OMG-Certified UML Professional No Magic Europe Savanoriu pr. 363, LT 49425 Kaunas, Lithuania P.O. box 2166, LT- 3000, Kaunas Phone: +370-37-324032 Fax: +370-37-320670 e-mail: nerijus@magicdraw.com WWW: http://www.magicdraw.com -- MagicDraw - Architecture made simple! ----- Original Message ----- From: Bran Selic To: Nerijus Jankevicius Cc: uml2-rtf@omg.org Sent: Thursday, April 08, 2010 10:57 AM Subject: Re: composite tags On Wed, Apr 7, 2010 at 4:43 PM, Nerijus Jankevicius wrote: Bran, Imagine I can define new composition between stereotyped Class and Package, so Class is able to own a Package. [bvs] Well, we did open up a Pandora's box when we allowed the addition of new meta-associations in profile tags. The problem of the possibility of non-sensical meta-relationships exists whether or not they involve composition. It is not possible to own such element in UML metamodel, [bvs] But, this is precisely the issue that was created once we allowed meta-associations in profiles. So, your question really has to do with the broader issue of whether we should allow typing of tags by meta-classes. (Too late to change this, I suspect.) I have no ideas how it should be saved into XMI also. Is it standard way to store such elements in "profile application" part of XMI as Eclipse do? How do you suppose it should work? [bvs] Ask the IBM folks -- apparently they have a way, since it seems that they deal with it as most people would expect. [bvs] Don't get me wrong, I think there is an issue here. However, I suspect that we cannot do much about it without redesigning the profile mechanism. Unfortunately, when we defined profiles, we really had no deep understanding of the mechanism or any theoretical foundations behind them. One of the principles was that a profile should be compatible with UML, but we did not state clearly what is meant by "compatible". In UML 1, we were a bit more conservative and did not allow tags to be typed by meta-classes, but we changed our minds in UML 2, without really reflecting too much on how this impacted compatibility. [bvs] I realize this is not helping your problem, but, perhaps we can get someone to do proper research into the profile concept that would allow us to do it right in the future. Cheers...Bran X-Trusted-NM: yes From: "Nerijus Jankevicius" To: "Pete Rivett" , "Ed Seidewitz" Cc: , Subject: Re: composite tags Date: Thu, 8 Apr 2010 18:11:22 +0300 X-Mailer: Microsoft Outlook Express 6.00.2900.5843 X-EsetScannerBuild: 6895 Pete, please see document attached in my first email. Compositions are used in profiles. Still used in MARTE, was used in SysML also (e.g. Part owned by Block). Resending document again. Thanks, Nerijus ----- Original Message ----- From: Pete Rivett To: Nerijus Jankevicius ; Ed Seidewitz Cc: model-interchange@omg.org ; uml2-rtf@omg.org Sent: Thursday, April 08, 2010 5:50 PM Subject: RE: composite tags Ø Imagine I can define new composition between stereotyped Class and Package, so Class is able to own a Package. Not sure what you mean here . are you defining a composite association between the stereotypes? In which case that.s invalid since a stereotype instance can only be owned by the element to which it.s applied. Or are you talking about defining a new association, owned by the Profile, between the UML metaclasses? With both ends owned by the Association? Ø I have no ideas how it should be saved into XMI also. Ø Is it standard way to store such elements in "profile application" part of XMI as Eclipse do? Ø How do you suppose it should work? Assuming nothing else owned a given package that would in theory be possible and the XMI serialization rules do cover this. If the profile is named Pro with namespace prefix the same and the association is named A_C_P with ends c:UML::Class and p:UML::Package then you.d have the following XMI element for having class with id c1 owning package with id p1 (assuming they are in the same XMI file)

Note that this form of serialization is the same regardless of whether the association is composite. Though a tool would have to respect the composite semantics and delete p1 if c1 is deleted. And prevent p1 being owned by something else at the same time. Note that there would be no notation defined for this meta-association so it would not be possible to draw it. unless a given UML tool has a DSL-like capability . you.d have to use property sheets. I see this as pushing the boundaries so do not see it as important for model interchange testing . do you have a scenario where you see this being useful? Have you come across a Profile doing this? Pete From: Nerijus Jankevicius [mailto:nerijus@nomagic.com] Sent: Thursday, April 08, 2010 4:59 AM To: Ed Seidewitz; Pete Rivett Cc: model-interchange@omg.org Subject: Fw: composite tags Ed, Pete, any feedback on the issue below? It is highly related to Model Interchange and any Profile develpment. Thanks. Regards, -- Nerijus Jankevicius SysML Product Manager OMG-Certified UML Professional No Magic Europe Savanoriu pr. 363, LT 49425 Kaunas, Lithuania P.O. box 2166, LT- 3000, Kaunas Phone: +370-37-324032 Fax: +370-37-320670 e-mail: nerijus@magicdraw.com WWW: http://www.magicdraw.com -- MagicDraw - Architecture made simple! ----- Original Message ----- From: Bran Selic To: Nerijus Jankevicius Cc: uml2-rtf@omg.org Sent: Thursday, April 08, 2010 10:57 AM Subject: Re: composite tags On Wed, Apr 7, 2010 at 4:43 PM, Nerijus Jankevicius wrote: Bran, Imagine I can define new composition between stereotyped Class and Package, so Class is able to own a Package. [bvs] Well, we did open up a Pandora's box when we allowed the addition of new meta-associations in profile tags. The problem of the possibility of non-sensical meta-relationships exists whether or not they involve composition. It is not possible to own such element in UML metamodel, [bvs] But, this is precisely the issue that was created once we allowed meta-associations in profiles. So, your question really has to do with the broader issue of whether we should allow typing of tags by meta-classes. (Too late to change this, I suspect.) I have no ideas how it should be saved into XMI also. Is it standard way to store such elements in "profile application" part of XMI as Eclipse do? How do you suppose it should work? [bvs] Ask the IBM folks -- apparently they have a way, since it seems that they deal with it as most people would expect. [bvs] Don't get me wrong, I think there is an issue here. However, I suspect that we cannot do much about it without redesigning the profile mechanism. Unfortunately, when we defined profiles, we really had no deep understanding of the mechanism or any theoretical foundations behind them. One of the principles was that a profile should be compatible with UML, but we did not state clearly what is meant by "compatible". In UML 1, we were a bit more conservative and did not allow tags to be typed by meta-classes, but we changed our minds in UML 2, without really reflecting too much on how this impacted compatibility. [bvs] I realize this is not helping your problem, but, perhaps we can get someone to do proper research into the profile concept that would allow us to do it right in the future. Cheers...Bran Tagged_values_with_aggregation_composite1.pdf Subject: RE: composite tags Date: Thu, 8 Apr 2010 09:20:17 -0700 X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: composite tags Thread-Index: AcrXLdN+19ER3y3fRya4qLPaFvCO4QACGI3Q From: "Pete Rivett" To: "Nerijus Jankevicius" , "Ed Seidewitz" Cc: , OK so this is an additional third alternative to the ones I mentioned: a UML metaclass owned by a Stereotype. The Eclipse approach quoted in the document you attached seems correct to me: is there anything specific you object to in that? Pete From: Nerijus Jankevicius [mailto:nerijus@nomagic.com] Sent: Thursday, April 08, 2010 8:11 AM To: Pete Rivett; Ed Seidewitz Cc: model-interchange@omg.org; uml2-rtf@omg.org Subject: Re: composite tags Pete, please see document attached in my first email. Compositions are used in profiles. Still used in MARTE, was used in SysML also (e.g. Part owned by Block). Resending document again. Thanks, Nerijus ----- Original Message ----- From: Pete Rivett To: Nerijus Jankevicius ; Ed Seidewitz Cc: model-interchange@omg.org ; uml2-rtf@omg.org Sent: Thursday, April 08, 2010 5:50 PM Subject: RE: composite tags Ø Imagine I can define new composition between stereotyped Class and Package, so Class is able to own a Package. Not sure what you mean here . are you defining a composite association between the stereotypes? In which case that.s invalid since a stereotype instance can only be owned by the element to which it.s applied. Or are you talking about defining a new association, owned by the Profile, between the UML metaclasses? With both ends owned by the Association? Ø I have no ideas how it should be saved into XMI also. Ø Is it standard way to store such elements in "profile application" part of XMI as Eclipse do? Ø How do you suppose it should work? Assuming nothing else owned a given package that would in theory be possible and the XMI serialization rules do cover this. If the profile is named Pro with namespace prefix the same and the association is named A_C_P with ends c:UML::Class and p:UML::Package then you.d have the following XMI element for having class with id c1 owning package with id p1 (assuming they are in the same XMI file)

Note that this form of serialization is the same regardless of whether the association is composite. Though a tool would have to respect the composite semantics and delete p1 if c1 is deleted. And prevent p1 being owned by something else at the same time. Note that there would be no notation defined for this meta-association so it would not be possible to draw it. unless a given UML tool has a DSL-like capability . you.d have to use property sheets. I see this as pushing the boundaries so do not see it as important for model interchange testing . do you have a scenario where you see this being useful? Have you come across a Profile doing this? Pete From: Nerijus Jankevicius [mailto:nerijus@nomagic.com] Sent: Thursday, April 08, 2010 4:59 AM To: Ed Seidewitz; Pete Rivett Cc: model-interchange@omg.org Subject: Fw: composite tags Ed, Pete, any feedback on the issue below? It is highly related to Model Interchange and any Profile develpment. Thanks. Regards, -- Nerijus Jankevicius SysML Product Manager OMG-Certified UML Professional No Magic Europe Savanoriu pr. 363, LT 49425 Kaunas, Lithuania P.O. box 2166, LT- 3000, Kaunas Phone: +370-37-324032 Fax: +370-37-320670 e-mail: nerijus@magicdraw.com WWW: http://www.magicdraw.com -- MagicDraw - Architecture made simple! ----- Original Message ----- From: Bran Selic To: Nerijus Jankevicius Cc: uml2-rtf@omg.org Sent: Thursday, April 08, 2010 10:57 AM Subject: Re: composite tags On Wed, Apr 7, 2010 at 4:43 PM, Nerijus Jankevicius wrote: Bran, Imagine I can define new composition between stereotyped Class and Package, so Class is able to own a Package. [bvs] Well, we did open up a Pandora's box when we allowed the addition of new meta-associations in profile tags. The problem of the possibility of non-sensical meta-relationships exists whether or not they involve composition. It is not possible to own such element in UML metamodel, [bvs] But, this is precisely the issue that was created once we allowed meta-associations in profiles. So, your question really has to do with the broader issue of whether we should allow typing of tags by meta-classes. (Too late to change this, I suspect.) I have no ideas how it should be saved into XMI also. Is it standard way to store such elements in "profile application" part of XMI as Eclipse do? How do you suppose it should work? [bvs] Ask the IBM folks -- apparently they have a way, since it seems that they deal with it as most people would expect. [bvs] Don't get me wrong, I think there is an issue here. However, I suspect that we cannot do much about it without redesigning the profile mechanism. Unfortunately, when we defined profiles, we really had no deep understanding of the mechanism or any theoretical foundations behind them. One of the principles was that a profile should be compatible with UML, but we did not state clearly what is meant by "compatible". In UML 1, we were a bit more conservative and did not allow tags to be typed by meta-classes, but we changed our minds in UML 2, without really reflecting too much on how this impacted compatibility. [bvs] I realize this is not helping your problem, but, perhaps we can get someone to do proper research into the profile concept that would allow us to do it right in the future. Cheers...Bran From: "Rouquette, Nicolas F (316A)" To: Pete Rivett CC: Nerijus Jankevicius , Ed Seidewitz , "model-interchange@omg.org" , "uml2-rtf@omg.org" Date: Thu, 8 Apr 2010 11:38:55 -0700 Subject: Re: composite tags Thread-Topic: composite tags Thread-Index: AcrXSr70Iqh0t1bmRJKLS0/LBs9iEA== Accept-Language: en-US X-MS-Has-Attach: yes X-MS-TNEF-Correlator: acceptlanguage: en-US X-Source-IP: altvirehtstap01.jpl.nasa.gov [128.149.137.72] X-Source-Sender: nicolas.f.rouquette@jpl.nasa.gov X-AUTH: Authorized First, Nerijus' example of a stereotype extending uml::Class with a composite association to a uml::Package is not a well-formed stereotype; I'll explain this below. Second, this isssue is related to 14961 which we resolved in ballot 4 for UML 2.4 with the following replacement for a paragraph in 18.3.6: .Stereotypes can participate in binary associations. The opposite class can be another stereotype, a non-stereotype class that is owned by a profile, or a metaclass of the reference metamodel. For these associations there must be a property owned by the Stereotype to navigate to the opposite class. Where the opposite class is not a stereotype, the opposite property must be owned by the Association itself rather than the other class/metaclass.. Nerijus' example is invalid and MagicDraw's behavior is correct and Eclipse' behavior is incorrect. Attached is an example in Eclipse Helios 3.6M6 and MagicDraw 16.8 beta of Nerijus' case: a stereotype extending class with a composite association to a package. Eclipse lets you construct an ill-formed profile as Nerijus described; MagicDraw doesn't. CompositeTagProfile.uml MDCompositeTagTest.mdzip My-after.uml My-before.uml ATT000011.txt PastedGraphic-5.tiff ATT00002.txt From: "Tran, Henry" To: "'Rouquette, Nicolas F (316A)'" , Pete Rivett CC: Nerijus Jankevicius , Ed Seidewitz , "model-interchange@omg.org" , "uml2-rtf@omg.org" Date: Thu, 8 Apr 2010 13:05:02 -0700 Subject: RE: composite tags Thread-Topic: composite tags Thread-Index: AcrXSr70Iqh0t1bmRJKLS0/LBs9iEAAC3P+Q Accept-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: acceptlanguage: en-US x-tm-as-product-ver: SMEX-10.0.0.1412-6.000.1038-17308.001 x-tm-as-result: No--55.074100-0.000000-31 x-tm-as-user-approved-sender: Yes x-tm-as-user-blocked-sender: No X-MIME-Autoconverted: from quoted-printable to 8bit by amethyst.omg.org id o38Jtr4v029257 Alls, May I ask that we always add the "MIWG" in the email subject so that I can set rule to forward the model based interchange activity to the appropriate folder? Best regards, -----Original Message----- From: Rouquette, Nicolas F (316A) [mailto:nicolas.f.rouquette@jpl.nasa.gov] Sent: Thursday, April 08, 2010 11:39 AM To: Pete Rivett Cc: Nerijus Jankevicius; Ed Seidewitz; model-interchange@omg.org; uml2-rtf@omg.org Subject: Re: composite tags First, Nerijus' example of a stereotype extending uml::Class with a composite association to a uml::Package is not a well-formed stereotype; I'll explain this below. Second, this isssue is related to 14961 which we resolved in ballot 4 for UML 2.4 with the following replacement for a paragraph in 18.3.6: "Stereotypes can participate in binary associations. The opposite class can be another stereotype, a non-stereotype class that is owned by a profile, or a metaclass of the reference metamodel. For these associations there must be a property owned by the Stereotype to navigate to the opposite class. Where the opposite class is not a stereotype, the opposite property must be owned by the Association itself rather than the other class/metaclass." Nerijus' example is invalid and MagicDraw's behavior is correct and Eclipse' behavior is incorrect. Attached is an example in Eclipse Helios 3.6M6 and MagicDraw 16.8 beta of Nerijus' case: a stereotype extending class with a composite association to a package. Eclipse lets you construct an ill-formed profile as Nerijus described; MagicDraw doesn't. DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:sender:received:in-reply-to :references:from:date:x-google-sender-auth:received:message-id :subject:to:cc:content-type; bh=gsNRn0i5P8koxZsB1ombBVDyuA7TgH75OInhmPz1Bt4=; b=tMFuCcBvicqJgytMLKsUH437+XZJfHlPTmw7DcBQ1rIQ0xbCxgERoWXsTc7o2HL4mV aF+64PkZm2c2xe36i5guHtH+09zQQVXsrzD2nPqDjggCyrm9IHDPN09Y0U5Uv1W09iDp PPqulbqUyi45ON3HVBRKQeovOjb0tYndMmJP0= DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:sender:in-reply-to:references:from:date :x-google-sender-auth:message-id:subject:to:cc:content-type; b=dcvoMzJhhGL99+gPHeO0VwJKUbXnuCxjnO//qSNNJ5dsb95G//hcqP496eY1lYr095 inD08e62oxdo3WEThO9UzBhboWQA3gNmHaMJmxj6JxRxkVDsVkHyUxM75TQTBTDO86Sw h2r0T73MlivQBENouMIR0W5wt6xz1iQVwVUr4= Sender: bran.selic@gmail.com From: Bran Selic Date: Fri, 9 Apr 2010 09:09:08 +0200 X-Google-Sender-Auth: 2f69b6f0695ca359 Subject: Re: composite tags To: "Rouquette, Nicolas F (316A)" Cc: Pete Rivett , Nerijus Jankevicius , Ed Seidewitz , "model-interchange@omg.org" , "uml2-rtf@omg.org" Nicolas, I think we got too focused on an incidental example brought up by Nerijus. The original issue is whether or not it is possible to have composite tag definitions. It seems to me that nothing you've cited so far prevents that. Cheers...Bran On Thu, Apr 8, 2010 at 8:38 PM, Rouquette, Nicolas F (316A) wrote: First, Nerijus' example of a stereotype extending uml::Class with a composite association to a uml::Package is not a well-formed stereotype; I'll explain this below. Second, this isssue is related to 14961 which we resolved in ballot 4 for UML 2.4 with the following replacement for a paragraph in 18.3.6: .Stereotypes can participate in binary associations. The opposite class can be another stereotype, a non-stereotype class that is owned by a profile, or a metaclass of the reference metamodel. For these associations there must be a property owned by the Stereotype to navigate to the opposite class. Where the opposite class is not a stereotype, the opposite property must be owned by the Association itself rather than the other class/metaclass.. Nerijus' example is invalid and MagicDraw's behavior is correct and Eclipse' behavior is incorrect. Attached is an example in Eclipse Helios 3.6M6 and MagicDraw 16.8 beta of Nerijus' case: a stereotype extending class with a composite association to a package. Eclipse lets you construct an ill-formed profile as Nerijus described; MagicDraw doesn't. X-Trusted-NM: yes Subject: Re: composite tags From: Nerijus Jankevicius Date: Fri, 9 Apr 2010 11:15:02 +0300 Cc: "Rouquette, Nicolas F (316A)" , Pete Rivett , Ed Seidewitz , "model-interchange@omg.org" , "uml2-rtf@omg.org" To: Bran Selic X-Mailer: Apple Mail (2.1078) X-PMX-Version: 5.5.7.378829, Antispam-Engine: 2.7.2.376379, Antispam-Data: 2010.4.9.81542 X-PerlMx-Spam: Gauge=IIIIIIIII, Probability=9%, Report=' HTML_NO_HTTP 0.1, LEO_OBFU_SUBJ_RE 0.1, SUPERLONG_LINE 0.05, BODY_SIZE_5000_5999 0, BODY_SIZE_7000_LESS 0, CTYPE_MULTIPART_NO_QUOTE 0, __BOUNCE_CHALLENGE_SUBJ 0, __BOUNCE_NDR_SUBJ_EXEMPT 0, __C230066_P3_2 0, __CT 0, __CTYPE_HAS_BOUNDARY 0, __CTYPE_MULTIPART 0, __CTYPE_MULTIPART_ALT 0, __DATE_TZ_RU 0, __HAS_HTML 0, __HAS_MSGID 0, __HAS_X_MAILER 0, __MIME_HTML 0, __MIME_VERSION 0, __MIME_VERSION_APPLEMAIL 0, __MSGID_APPLEMAIL 0, __SANE_MSGID 0, __TAG_EXISTS_HTML 0, __TO_MALFORMED_2 0, __URI_NO_PATH 0, __URI_NO_WWW 0, __URI_NS , __USER_AGENT_APPLEMAIL 0, __X_MAILER_APPLEMAIL 0' X-PerlMx-SIL: Medium X-OriginalArrivalTime: 09 Apr 2010 08:21:17.0115 (UTC) FILETIME=[A04508B0:01CAD7BD] As always, simple question becomes so complex that nobody even understands what we are trying to solve. Let's rephrase the issue: If we have composition between two stereotypes or one stereotype and metaclass (I see no difference), who is the owner of the "part" (the owned METACLASS instance) in the M2 (user level) model? 1. if null, it is invalid 2. if not null, it implies the other end of owner-owned element association is set and element must be added into real subset of owned elements which is invalid also as this composition is not specialized association and role can't subset UML property. Where such UML element (M2) is stored in XMI ? In the "profile application" part? Nerijus On Apr 9, 2010, at 10:09 AM, Bran Selic wrote: Nicolas, I think we got too focused on an incidental example brought up by Nerijus. The original issue is whether or not it is possible to have composite tag definitions. It seems to me that nothing you've cited so far prevents that. Cheers...Bran On Thu, Apr 8, 2010 at 8:38 PM, Rouquette, Nicolas F (316A) wrote: First, Nerijus' example of a stereotype extending uml::Class with a composite association to a uml::Package is not a well-formed stereotype; I'll explain this below. Second, this isssue is related to 14961 which we resolved in ballot 4 for UML 2.4 with the following replacement for a paragraph in 18.3.6: .Stereotypes can participate in binary associations. The opposite class can be another stereotype, a non-stereotype class that is owned by a profile, or a metaclass of the reference metamodel. For these associations there must be a property owned by the Stereotype to navigate to the opposite class Where the opposite class is not a stereotype, the opposite property must be owned by the Association itself rather than the other class/metaclass.. Nerijus' example is invalid and MagicDraw's behavior is correct and Eclipse' behavior is incorrect. Attached is an example in Eclipse Helios 3.6M6 and MagicDraw 16.8 beta of Nerijus' case: a stereotype extending class with a composite association to a package. Eclipse lets you construct an ill-formed profile as Nerijus described; MagicDraw doesn't. This mail has originated outside your organization, either from an external partner or the Global Internet. Keep this in mind if you answer this message. The information in this e-mail is confidential. The contents may not be disclosed or used by anyone other than the addressee. Access to this e-mail by anyone else is unauthorised. If you are not the intended recipient, please notify Airbus immediately and delete this e-mail. Airbus cannot accept any responsibility for the accuracy or completeness of this e-mail as it has been sent over public networks. If you have any concerns over the content of this message or its Accuracy or Integrity, please contact Airbus immediately. All outgoing e-mails from Airbus are checked using regularly updated virus scanning software but you should take whatever measures you deem to be appropriate to ensure that this message and any attachments are virus free. X-Trusted-NM: yes Subject: Re: composite tags From: Nerijus Jankevicius Date: Fri, 9 Apr 2010 11:15:02 +0300 Cc: "Rouquette, Nicolas F (316A)" , Pete Rivett , Ed Seidewitz , "model-interchange@omg.org" , "uml2-rtf@omg.org" To: Bran Selic X-Mailer: Apple Mail (2.1078) As always, simple question becomes so complex that nobody even understands what we are trying to solve. Let's rephrase the issue: If we have composition between two stereotypes or one stereotype and metaclass (I see no difference), who is the owner of the "part" (the owned METACLASS instance) in the M2 (user level) model? 1. if null, it is invalid 2. if not null, it implies the other end of owner-owned element association is set and element must be added into real subset of owned elements which is invalid also as this composition is not specialized association and role can't subset UML property. Where such UML element (M2) is stored in XMI ? In the "profile application" part? Nerijus On Apr 9, 2010, at 10:09 AM, Bran Selic wrote: Nicolas, I think we got too focused on an incidental example brought up by Nerijus. The original issue is whether or not it is possible to have composite tag definitions. It seems to me that nothing you've cited so far prevents that. Cheers...Bran On Thu, Apr 8, 2010 at 8:38 PM, Rouquette, Nicolas F (316A) wrote: First, Nerijus' example of a stereotype extending uml::Class with a composite association to a uml::Package is not a well-formed stereotype; I'll explain this below. Second, this isssue is related to 14961 which we resolved in ballot 4 for UML 2.4 with the following replacement for a paragraph in 18.3.6: .Stereotypes can participate in binary associations. The opposite class can be another stereotype, a non-stereotype class that is owned by a profile, or a metaclass of the reference metamodel. For these associations there must be a property owned by the Stereotype to navigate to the opposite class. Where the opposite class is not a stereotype, the opposite property must be owned by the Association itself rather than the other class/metaclass.. Nerijus' example is invalid and MagicDraw's behavior is correct and Eclipse' behavior is incorrect. Attached is an example in Eclipse Helios 3.6M6 and MagicDraw 16.8 beta of Nerijus' case: a stereotype extending class with a composite association to a package. Eclipse lets you construct an ill-formed profile as Nerijus described; MagicDraw doesn't. From: "Rouquette, Nicolas F (316A)" To: Pete Rivett , Nerijus Jankevicius CC: Ed Seidewitz , "model-interchange@omg.org" , "uml2-rtf@omg.org" Date: Thu, 8 Apr 2010 22:35:34 -0700 Subject: MIWG - Re: composite tags Thread-Topic: MIWG - Re: composite tags Thread-Index: AcrXpnvSKIp87zVcRCeGr6Hl5lVqIQ== Accept-Language: en-US X-MS-Has-Attach: yes X-MS-TNEF-Correlator: acceptlanguage: en-US X-Source-IP: altvirehtstap01.jpl.nasa.gov [128.149.137.72] X-Source-Sender: nicolas.f.rouquette@jpl.nasa.gov X-AUTH: Authorized Below is a diagram of the MOF-equivalent construction of Nerijus' example involving a stereotype, ClassPackage, that extends uml::Class and has a composite association to uml::Package, A_nestedPackage_classPackage. The MOF-equivalent construction makes ClassPackage a metaclass specializing uml::Class instead of a stereotype extending uml::Class. The association A_nestedPackage_classPackage becomes a meta association. Both ClassPackage and A_nestedPackage_classPackage could be defined in a package that, when merged with the UML metamodel, would extend it with the capability to nest a uml::Package in a ClassPackage. Is this MOF-equivalent construction legal as a profile? If ClassPackage + A_nestedPackage_classPackage is more than a specialization of the UML metamodel, it extends the UML with a non-trivial capability that isn't available in the UML metamodel and this would be a clear violation of the intent of profile in UML as described in clause 18.1.2 as I indicated below. To understand why ClassPackage + A_nestedPackage_classPackage is more than a specialization of the UML metamodel, it is useful to look at the taxonomy of meta-association specializations, particularly A_ownedElement_owner, which is the workhorse meta-association for defining ownership relations, but not the only one as Ed observed for the activities metamodel. The diagram below shows the relevant taxonomy of specializations of A_ownedElement_owner in the UML metamodel up to uml::Class and uml::Package. I hope this diagram suffices to show that one cannot specialize both association ends of as redefinitions/subsets of association ends that ultimately specialize those of A_ownedElement_owner. If you defined the profile where <> extends uml::Class and had the association A_nestedPackage_classPackage, then you get the strange behavior that I showed with Eclipse Helios 3.6M6 where, a nestedPackage becomes a uml::Element that has completely lost its context with the rest of the model because the profile induces an incompatible extension of the uml metamodel. PastedGraphic-7.tiff Subject: RE: composite tags Date: Fri, 9 Apr 2010 07:49:46 -0700 X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: composite tags Thread-Index: AcrXzokDvqJEzzlEQg+c7xjE3HHrRgAIyM0A From: "Pete Rivett" To: "Nerijus Jankevicius" , "Bran Selic" Cc: "Rouquette, Nicolas F (316A)" , "Ed Seidewitz" , , My previous response applies: the Eclipse approach seems correct. Nicolas. claim that this is not well-formed is only if you try to specialize the existing UML Element ownership association. That may be the source of confusion in your email too: just because you define a new association as a composition does not mean it specializes the UML Element ownership association: in the UML metamodel all such association properties do that explicitly.. I don.t understand your latest email Nerijus : what are you referring to when you say .if null.? And BTW M2 is the MetaModel (2 letter Ms) level, M1 is the user level model. The basic principle is to treat the stereotype definition at the same level as the UML metamodel. So if there.s a composition from a stereotype then the instance of the stereotype will own the linked element. That is ownership in the sense of deletion semantics, restriction to one owner not inclusion in the UML .ownedElements. derived union. With respect to XMI there is no such thing as the .profile application part. . there are only elements belonging in different XML namespaces. XMI does not mandate that such elements are grouped together in a physical section of the file. The elements .owned. by a stereotype instance can either be nested in that instance (as per the Eclipse example you sent . though I.d prefer to see an explicit xmi:type on that element) or can be at the root level linked to their owner using the Association XMI syntax I showed in my original reply. In either case a UML element will remain in the UML namespace. Pete From: Nerijus Jankevicius [mailto:nerijus@nomagic.com] Sent: Friday, April 09, 2010 1:15 AM To: Bran Selic Cc: Rouquette, Nicolas F (316A); Pete Rivett; Ed Seidewitz; model-interchange@omg.org; uml2-rtf@omg.org Subject: Re: composite tags As always, simple question becomes so complex that nobody even understands what we are trying to solve. Let's rephrase the issue: If we have composition between two stereotypes or one stereotype and metaclass (I see no difference), who is the owner of the "part" (the owned METACLASS instance) in the M2 (user level) model? 1. if null, it is invalid 2. if not null, it implies the other end of owner-owned element association is set and element must be added into real subset of owned elements which is invalid also as this composition is not specialized association and role can't subset UML property. Where such UML element (M2) is stored in XMI ? In the "profile application" part? Nerijus On Apr 9, 2010, at 10:09 AM, Bran Selic wrote: Nicolas, I think we got too focused on an incidental example brought up by Nerijus. The original issue is whether or not it is possible to have composite tag definitions. It seems to me that nothing you've cited so far prevents that. Cheers...Bran On Thu, Apr 8, 2010 at 8:38 PM, Rouquette, Nicolas F (316A) wrote: First, Nerijus' example of a stereotype extending uml::Class with a composite association to a uml::Package is not a well-formed stereotype; I'll explain this below. Second, this isssue is related to 14961 which we resolved in ballot 4 for UML 2.4 with the following replacement for a paragraph in 18.3.6: .Stereotypes can participate in binary associations. The opposite class can be another stereotype, a non-stereotype class that is owned by a profile, or a metaclass of the reference metamodel. For these associations there must be a property owned by the Stereotype to navigate to the opposite class Where the opposite class is not a stereotype, the opposite property must be owned by the Association itself rather than the other class/metaclass.. Nerijus' example is invalid and MagicDraw's behavior is correct and Eclipse' behavior is incorrect. Attached is an example in Eclipse Helios 3.6M6 and MagicDraw 16.8 beta of Nerijus' case: a stereotype extending class with a composite association to a package. Eclipse lets you construct an ill-formed profile as Nerijus described; MagicDraw doesn't. This mail has originated outside your organization, either from an external partner or the Global Internet. Keep this in mind if you answer this message. The information in this e-mail is confidential. The contents may not be disclosed or used by anyone other than the addressee. Access to this e-mail by anyone else is unauthorised. If you are not the intended recipient, please notify Airbus immediately and delete this e-mail. Airbus cannot accept any responsibility for the accuracy or completeness of this e-mail as it has been sent over public networks. If you have any concerns over the content of this message or its Accuracy or Integrity, please contact Airbus immediately. All outgoing e-mails from Airbus are checked using regularly updated virus scanning software but you should take whatever measures you deem to be appropriate to ensure that this message and any attachments are virus free. From: "Rouquette, Nicolas F (316A)" To: Pete Rivett CC: Nerijus Jankevicius , Bran Selic , Ed Seidewitz , "uml2-rtf@omg.org" , "model-interchange@omg.org" Date: Fri, 9 Apr 2010 17:23:19 -0700 Subject: MIWG -- Re: composite tags Thread-Topic: MIWG -- Re: composite tags Thread-Index: AcrYRAUOYsqkuLjQTQOPrvt8KlcEIw== Accept-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: acceptlanguage: en-US X-Source-IP: altvirehtstap01.jpl.nasa.gov [128.149.137.72] X-Source-Sender: nicolas.f.rouquette@jpl.nasa.gov X-AUTH: Authorized On Apr 9, 2010, at 7:49 AM, Pete Rivett wrote: My previous response applies: the Eclipse approach seems correct. Nicolas. claim that this is not well-formed is only if you try to specialize the existing UML Element ownership association. The issue has nothing to do with trying to specialize a meta-asosciation. The issue is whether it is possible to do this specialization at all for an association defined in a profile, regardless of the aggregation characteristics of its association ends. That may be the source of confusion in your email too: just because you define a new association as a composition does not mean it specializes the UML Element ownership association: in the UML metamodel all such association properties do that explicitly.. This is incorrect. Association end composition and element ownership in the sense of A_ownedElement_owner are completely different concepts. It so happens that most composite association ends are parts of associations that do/should specialize A_ownedElement_owner but there are some exceptions. Ed was very clear about this last year when we discussed a related issue to this for clause 12. I don.t understand your latest email Nerijus : what are you referring to when you say .if null.? And BTW M2 is the MetaModel (2 letter Ms) level, M1 is the user level model. The basic principle is to treat the stereotype definition at the same level as the UML metamodel. So if there.s a composition from a stereotype then the instance of the stereotype will own the linked element. That is ownership in the sense of deletion semantics, restriction to one owner not inclusion in the UML .ownedElements. derived union. Pete, since you like the MOF-equivalent perspective, I encourage you to look at Nerijus' example from that point of view. The stereotype ClassPackage becomes a metaclass: ClassPackage. The profile association becomes a meta-association. At the MOF level, there is no problem but this allows you to ask whether the MOF extension is limited or not. Unfortunately, clause 18 hasn't clearly defined what a limited extension of the UML metamodel is explicitly; however, there is a clear intent in 18.1.2 about the fact that a profile is allowed to specialize the UML metamodel only. Since the UML metamodel is a collection of classifiers -- metaclasses, meta-associations and datatype -- these classifiers can be specialized. My point then is that the composite association between the ClassPackage metaclass and UML::Package *cannot* be defined properly according to the metamodeling rules we've used for constructing the UML itself as a specialization of *any* existing UML meta-association. From this point of view, I say that Nerijus example cannot be called a limited extension of the UML; this kind of extension should be done using a MOF tool, not a UML profile. The fact that this example involves a composite association is only incidental. The same analysis applies to *any* association defined in a profile. For the record, I should also clarify that the folks who defined the SysML profile were amazingly *wise* in avoiding like the plague using any kind of association between SysML stereotypes and metaclasses or other stereotypes. Check carefully -- SysML stereotypes *only* define properties, not associations. This saved SysML a lot of of grief and this makes SysML a reasonable profile to use for model interchange. In MIWG, we will have a lot more issues when we attempt to interchange profiles that do define associations. There, the kind of problems Nerijus asked about with composite tags will surface but as far as the interchange of SysML models for MIWG, the question of composite tags is irrelevant because SysML didn't make this mistake. - Nicolas. With respect to XMI there is no such thing as the .profile application part. . there are only elements belonging in different XML namespaces. XMI does not mandate that such elements are grouped together in a physical section of the file. The elements .owned. by a stereotype instance can either be nested in that instance (as per the Eclipse example you sent . though I.d prefer to see an explicit xmi:type on that element) or can be at the root level linked to their owner using the Association XMI syntax I showed in my original reply. In either case a UML element will remain in the UML namespace. Pete - Nicolas. From: Nerijus Jankevicius [mailto:nerijus@nomagic.com] Sent: Friday, April 09, 2010 1:15 AM To: Bran Selic Cc: Rouquette, Nicolas F (316A); Pete Rivett; Ed Seidewitz; model-interchange@omg.org; uml2-rtf@omg.org Subject: Re: composite tags As always, simple question becomes so complex that nobody even understands what we are trying to solve. Let's rephrase the issue: If we have composition between two stereotypes or one stereotype and metaclass (I see no difference), who is the owner of the "part" (the owned METACLASS instance) in the M2 (user level) model? 1. if null, it is invalid 2. if not null, it implies the other end of owner-owned element association is set and element must be added into real subset of owned elements which is invalid also as this composition is not specialized association and role can't subset UML property. Where such UML element (M2) is stored in XMI ? In the "profile application" part? Nerijus On Apr 9, 2010, at 10:09 AM, Bran Selic wrote: Nicolas, I think we got too focused on an incidental example brought up by Nerijus. The original issue is whether or not it is possible to have composite tag definitions. It seems to me that nothing you've cited so far prevents that. Cheers...Bran On Thu, Apr 8, 2010 at 8:38 PM, Rouquette, Nicolas F (316A) wrote: First, Nerijus' example of a stereotype extending uml::Class with a composite association to a uml::Package is not a well-formed stereotype; I'll explain this below. Second, this isssue is related to 14961 which we resolved in ballot 4 for UML 2.4 with the following replacement for a paragraph in 18.3.6: .Stereotypes can participate in binary associations. The opposite class can be another stereotype, a non-stereotype class that is owned by a profile, or a metaclass of the reference metamodel. For these associations there must be a property owned by the Stereotype to navigate to the opposite class Where the opposite class is not a stereotype, the opposite property must be owned by the Association itself rather than the other class/metaclass.. Nerijus' example is invalid and MagicDraw's behavior is correct and Eclipse' behavior is incorrect. Attached is an example in Eclipse Helios 3.6M6 and MagicDraw 16.8 beta of Nerijus' case: a stereotype extending class with a composite association to a package. Eclipse lets you construct an ill-formed profile as Nerijus described; MagicDraw doesn't. This mail has originated outside your organization, either from an external partner or the Global Internet. Keep this in mind if you answer this message. The information in this e-mail is confidential. The contents may not be disclosed or used by anyone other than the addressee. Access to this e-mail by anyone else is unauthorised. If you are not the intended recipient, please notify Airbus immediately and delete this e-mail. Airbus cannot accept any responsibility for the accuracy or completeness of this e-mail as it has been sent over public networks. If you have any concerns over the content of this message or its Accuracy or Integrity, please contact Airbus immediately. All outgoing e-mails from Airbus are checked using regularly updated virus scanning software but you should take whatever measures you deem to be appropriate to ensure that this message and any attachments are virus free. ATT000011.txt