Issue 5918: CCM Spec: attributes are listed in the ports section? (components-ftf) Source: Raytheon (Mr. Craig Rodrigues, nobody) Nature: Uncategorized Issue Severity: Summary: In section 1.1.2 of the CCM specification: 1.1.2 Ports =========== ..... The component model supports four basic kinds of ports: - Facets - Receptacles - Event sources - Event sinks - Attributes Well, that list includes five things, not four. So, is an attribute considered a port or not? The wording in this section needs to be clarified in the CCM specification, because it is not clear if an attribute is a port or not. Resolution: Revised Text: Actions taken: April 28, 2003: received issue Discussion: End of Annotations:===== Date: Mon, 28 Apr 2003 13:00:21 -0400 From: Craig Rodrigues To: issues@omg.org Subject: CCM Spec: attributes are listed in the ports section? User-Agent: Mutt/1.4i Hi, In section 1.1.2 of the CCM specification: 1.1.2 Ports =========== ..... The component model supports four basic kinds of ports: - Facets - Receptacles - Event sources - Event sinks - Attributes Well, that list includes five things, not four. So, is an attribute considered a port or not? The wording in this section needs to be clarified in the CCM specification, because it is not clear if an attribute is a port or not. -- Craig Rodrigues Distributed Systems and Logistics, Office 6/325A crodrigu@bbn.com BBN Technologies, a Verizon company (617) 873-4725 Cambridge, MA From: ritter@fokus.fraunhofer.de X-Authentication-Warning: saturn.fokus.fraunhofer.de: nobody set sender to tri@fokus.fhg.de using -f To: ccm@moriarty.dif.um.es Subject: Re: [CCM] Are attributes ports? Date: Thu, 01 May 2003 21:13:27 +0200 (MET DST) Cc: components-rtf@omg.org User-Agent: IMP/PHP IMAP webmail program 2.2.7 Hi Craig, Thanks for the issue. But I do not fully agree with Philippe and Scott. I believe attributes are quite different from facet, receptacles, event sinks, and event sources. Attributes are used for configuration only. I think we shouldn't call them ports. I can't connect an attribute with another 'attribute'. But maybe we should add the component equivalent interface (component reference) as port. Component attributes are included in the component reference. Cheers, Tom Zitiere Craig Rodrigues : > On Wed, Apr 30, 2003 at 07:05:42PM +0200, Philippe Merle wrote: > > > If you consider a "port" to represent access to component behavior > through > > > a well-defined mechanism then attributes would be ports, IMHO. I'm > not > > > sure that connecting with components should be considered the main > purpose > > > for a port - I personally don't consider providing a facet to be a > > > connection in the same sense as receptacles and events. > > > > I agree with you. > > > > > > > > >Perhaps this is another issue to be clarified in the spec to avoid > as much > > > >confusion as possible. > > > > > > Yes, the spec should be explicit about attributes being ports if > there is a > > > consensus among the RTF members that this is the case. > > > > Here I think we have just a typo in the CCM specification > > 'four' must be read 'five'. > > > Hi, > > I submitted this to OMG as issue 5918: > http://www.omg.org/issues/components-ftf.html#Issue5918 > > Can someone follow up on this? > > Thanks. > -- > Craig Rodrigues Distributed Systems and Logistics, Office 6/325A > crodrigu@bbn.com BBN Technologies, a Verizon company > (617) 873-4725 Cambridge, MA > > _______________________________________________ > CCM mailing list > CCM@moriarty.dif.um.es > http://moriarty.dif.um.es/mailman/listinfo/ccm > Supported by The CCM Page: http://www.ditec.um.es/~dsevilla/ccm > X-Dreamscape-Track-Mars-A: gymer.mc.com [192.148.197.11] X-Dreamscape-Track-Mars-B: Thu, 1 May 2003 15:29:57 -0400 (EDT) Date: Thu, 01 May 2003 15:30:03 -0400 From: James Kulp Organization: Mercury Computer Systems, Inc. User-Agent: Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US; rv:1.3) Gecko/20030312 X-Accept-Language: en,pdf To: ritter@fokus.fraunhofer.de CC: ccm@moriarty.dif.um.es, components-rtf@omg.org Subject: Re: [CCM] Are attributes ports? I completely agree. In the deployment & configuration submission (PSM CCM) we found it natural and simpler to consider the specific interface of a CCM component as a slightly special port. Of course since configuration properties are in general (in the PIM sense) quite different from ports, we chose to consider them as different from ports since they are not necessarily valid at runtime. Thus we take the CCM attributes as both configuration properties as well as part of the (runtime-usable) specific interface. Jim ritter@fokus.fraunhofer.de wrote: Hi Craig, Thanks for the issue. But I do not fully agree with Philippe and Scott. I believe attributes are quite different from facet, receptacles, event sinks, and event sources. Attributes are used for configuration only. I think we shouldn't call them ports. I can't connect an attribute with another 'attribute'. But maybe we should add the component equivalent interface (component reference) as port. Component attributes are included in the component reference. Cheers, Tom Zitiere Craig Rodrigues : On Wed, Apr 30, 2003 at 07:05:42PM +0200, Philippe Merle wrote: If you consider a "port" to represent access to component behavior through a well-defined mechanism then attributes would be ports, IMHO. I'm not sure that connecting with components should be considered the main purpose for a port - I personally don't consider providing a facet to be a connection in the same sense as receptacles and events. I agree with you. Perhaps this is another issue to be clarified in the spec to avoid as much confusion as possible. Yes, the spec should be explicit about attributes being ports if there is a consensus among the RTF members that this is the case. Here I think we have just a typo in the CCM specification 'four' must be read 'five'. Hi, I submitted this to OMG as issue 5918: http://www.omg.org/issues/components-ftf.html#Issue5918 Can someone follow up on this? Thanks. -- Craig Rodrigues Distributed Systems and Logistics, Office 6/325A crodrigu@bbn.com BBN Technologies, a Verizon company (617) 873-4725 Cambridge, MA _______________________________________________ CCM mailing list CCM@moriarty.dif.um.es http://moriarty.dif.um.es/mailman/listinfo/ccm Supported by The CCM Page: http://www.ditec.um.es/~dsevilla/ccm X-Sender: evans@mail.cpi.com X-Mailer: QUALCOMM Windows Eudora Version 5.1 Date: Thu, 01 May 2003 16:54:28 -0400 To: ccm@moriarty.dif.um.es From: "J. Scott Evans" Subject: Re: [CCM] Are attributes ports? Cc: components-rtf@omg.org Hi Tom, At 09:13 PM 5/1/03 +0200, ritter@fokus.fraunhofer.de wrote: Hi Craig, Thanks for the issue. But I do not fully agree with Philippe and Scott. I believe attributes are quite different from facet, receptacles, event sinks, and event sources. Attributes are used for configuration only. I think we shouldn't call them ports. I can't connect an attribute with another 'attribute'. But maybe we should add the component equivalent interface (component reference) as port. Component attributes are included in the component reference. From formal/02-06-65 page 1-2: "Components support a variety of surface features through which clients and other elements of an application environment may interact with a component. These surface features are called ports. The component model supports four basic kinds of ports:" I still don't understand the association between "port" and "connection." IMHO, an attribute is a legitimate surface feature of a component and therefore is a port according to the description of a port. That attributes are used for configuration does not necessarily demote them in relation to facets, events, or receptacles. Isn't state an important surface feature of a component and access to it via attributes a valid form of interaction with a component? To define a component itself as a port is not consistent with the definition of a port as given above and poses semantic problems since it implies that a specific port (the component) itself has ports in a composition like relationship. This would not be consistent for all ports and is not implied by the definition of either a port or a component. So, IMHO, we should change the "four basic kinds of ports" to "five basic kinds of ports" or vote to change the definition of a port, a component, or both so that attributes do not satisfy the definition of a port (other changes to the spec would also be required for consistency). Scott Cheers, Tom Zitiere Craig Rodrigues : > On Wed, Apr 30, 2003 at 07:05:42PM +0200, Philippe Merle wrote: > > > If you consider a "port" to represent access to component behavior > through > > > a well-defined mechanism then attributes would be ports, IMHO. I'm > not > > > sure that connecting with components should be considered the main > purpose > > > for a port - I personally don't consider providing a facet to be a > > > connection in the same sense as receptacles and events. > > > > I agree with you. > > > > > > > > >Perhaps this is another issue to be clarified in the spec to avoid > as much > > > >confusion as possible. > > > > > > Yes, the spec should be explicit about attributes being ports if > there is a > > > consensus among the RTF members that this is the case. > > > > Here I think we have just a typo in the CCM specification > > 'four' must be read 'five'. > > > Hi, > > I submitted this to OMG as issue 5918: > http://www.omg.org/issues/components-ftf.html#Issue5918 > > Can someone follow up on this? > > Thanks. > -- > Craig Rodrigues Distributed Systems and Logistics, Office 6/325A > crodrigu@bbn.com BBN Technologies, a Verizon company > (617) 873-4725 Cambridge, MA > > _______________________________________________ > CCM mailing list > CCM@moriarty.dif.um.es > http://moriarty.dif.um.es/mailman/listinfo/ccm > Supported by The CCM Page: http://www.ditec.um.es/~dsevilla/ccm > From: ritter@fokus.fraunhofer.de X-Authentication-Warning: saturn.fokus.fraunhofer.de: nobody set sender to tri@fokus.fhg.de using -f To: "J. Scott Evans" Subject: Re: [CCM] Are attributes ports? Date: Fri, 02 May 2003 17:23:31 +0200 (MET DST) Cc: ccm@moriarty.dif.um.es, components-rtf@omg.org User-Agent: IMP/PHP IMAP webmail program 2.2.7 Hi Scott, I do not stick to the names. Though I think that the other four 'ports' are somehow different from attributes. And I think that the components equivalent interface (component reference) is more of a port than the attributes. The access to state is a surface feature (not the state itself). Since the access goes via the component reference I would consider this as part of this port. I don't see why taking the component reference as port contradicts the quoted definition. I think it perfectly fits. What about a supported interface. Is a supported interface not a port (part of the component reference port)? We need to be aware of the clear distinction between component and the component reference. Therefore a component reference can be a port of the component. E.g. the beginning of 1.1.3 seem s to indicate that a component reference is somehow similar to the other component facets and this implies that the component reference should be considered as a port. Currently I'm not totally aware of all the text of the spec but I suppose adding the component reference as port (in 1.1.2) would imply only minor changes in the spec. This certainly needs serious checking. Doing so would include attributes as part of the component reference port. I think we do not need to make substantial changes to the definition of port except to add the component reference explicitly. I admit the fact that attributes are used for configuration does not really touch our discussion and the quoted definition of port. Cheers, Tom P.S. Hope CC to Diego's CCM list works this time. > From formal/02-06-65 page 1-2: > > "Components support a variety of surface features through which clients > and > other elements of an application environment may interact with a > component. > These surface features are called ports. The component model supports > four > basic kinds of ports:" > > I still don't understand the association between "port" and > "connection." IMHO, an attribute is a legitimate surface feature of a > component and therefore is a port according to the description of a > port. That attributes are used for configuration does not necessarily > demote them in relation to facets, events, or receptacles. Isn't state > an > important surface feature of a component and access to it via attributes > a > valid form of interaction with a component? To define a component > itself > as a port is not consistent with the definition of a port as given above > > and poses semantic problems since it implies that a specific port (the > component) itself has ports in a composition like relationship. This > would > not be consistent for all ports and is not implied by the definition of > > either a port or a component. So, IMHO, we should change the "four > basic > kinds of ports" to "five basic kinds of ports" or vote to change the > definition of a port, a component, or both so that attributes do not > satisfy the definition of a port (other changes to the spec would also > be > required for consistency). > > Scott > > >Cheers, > > Tom > > > > > >Zitiere Craig Rodrigues : > > > > > On Wed, Apr 30, 2003 at 07:05:42PM +0200, Philippe Merle wrote: > > > > > If you consider a "port" to represent access to component > behavior > > > through > > > > > a well-defined mechanism then attributes would be ports, IMHO. > I'm > > > not > > > > > sure that connecting with components should be considered the > main > > > purpose > > > > > for a port - I personally don't consider providing a facet to be > a > > > > > connection in the same sense as receptacles and events. > > > > > > > > I agree with you. > > > > > > > > > > > > > > >Perhaps this is another issue to be clarified in the spec to > avoid > > > as much > > > > > >confusion as possible. > > > > > > > > > > Yes, the spec should be explicit about attributes being ports > if > > > there is a > > > > > consensus among the RTF members that this is the case. > > > > > > > > Here I think we have just a typo in the CCM specification > > > > 'four' must be read 'five'. > > > > > > > > > Hi, > > > > > > I submitted this to OMG as issue 5918: > > > http://www.omg.org/issues/components-ftf.html#Issue5918 > > > > > > Can someone follow up on this? > > > > > > Thanks. > > > -- > > > Craig Rodrigues Distributed Systems and Logistics, Office > 6/325A > > > crodrigu@bbn.com BBN Technologies, a Verizon company > > > (617) 873-4725 Cambridge, MA > > > > > > _______________________________________________ > > > CCM mailing list > > > CCM@moriarty.dif.um.es > > > http://moriarty.dif.um.es/mailman/listinfo/ccm > > > Supported by The CCM Page: http://www.ditec.um.es/~dsevilla/ccm > > > > > X-Sender: evans@mail.cpi.com X-Mailer: QUALCOMM Windows Eudora Version 5.1 Date: Fri, 02 May 2003 15:00:35 -0400 To: ccm@moriarty.dif.um.es From: "J. Scott Evans" Subject: Re: [CCM] Are attributes ports? Cc: components-rtf@omg.org Hi Craig, At 01:48 PM 5/2/03 -0400, Craig Rodrigues wrote: On Fri, May 02, 2003 at 05:23:31PM +0200, ritter@fokus.fraunhofer.de wrote: > Hi Scott, > > I do not stick to the names. Though I think that the other four 'ports' are > somehow different from attributes. And I think that the components equivalent > interface (component reference) is more of a port than the attributes. The fact that there is fundamental disagreement on very basic CCM terminology is very worrisome. Having terminology that is agreed upon is very important when you have to explain this new and emerging technology to development teams and managers. I have to give a CCM presentation at the OMG Real-time Workshop in July, so I would like to have this particular issue clarified before I give my presentation. http://www.omg.org/issues/components-ftf.html#Issue5918 Based on the discussion thread on this mailing list, I am still very confused whether an attribute is a port or not. I wouldn't be too worried by the current discussion because it really amounts to a clarification in semantics with no expected impact on interactions with a component in an application environment. Tom, Jim, and Marco introduce valid points and the concept of the component as a port is not entirely inconsistent with the implementation framework concept, from 3.2.9.1, that a "segmented executor has a distinguished segment associated with the component", which manages the components attributes. I'd propose that the next step would be a draft of the changes suggested by Tom so that the interested RTF members have something concrete to discuss as a possible resolution for the issue. Tom, are you willing to submit a resolution for the issue with specific changes in wording in the spec? When is the next Components revision task force? I suppose the next RTF will be chartered at the next OMG technical meeting (June?). At the present time we can only informally agree on a proposed resolution which may or may not be approved through a vote by the next CCM RTF. Scott -- Craig Rodrigues Distributed Systems and Logistics, Office 6/325A crodrigu@bbn.com BBN Technologies, a Verizon company (617) 873-4725 Cambridge, MA _______________________________________________ CCM mailing list CCM@moriarty.dif.um.es http://moriarty.dif.um.es/mailman/listinfo/ccm Supported by The CCM Page: http://www.ditec.um.es/~dsevilla/ccm From: ritter@fokus.fraunhofer.de X-Authentication-Warning: saturn.fokus.fraunhofer.de: nobody set sender to tri@fokus.fhg.de using -f To: ccm@moriarty.dif.um.es, Craig Rodrigues Subject: Re: [CCM] Are attributes ports? Date: Fri, 02 May 2003 20:59:19 +0200 (MET DST) Cc: components-rtf@omg.org User-Agent: IMP/PHP IMAP webmail program 2.2.7 Hello Craig, first of all, I do believe that the discussion about calling attributes ports or not is not that much important as it seems to be. It is important that components can have attributes and on that all do agree, I hope ;-) . A final answer to that issue 5918 will be given in the next report of the next RTF. I'm afraid this will be after your presentation. Hopefully, we find a consensus within the mailing list soon. It is most likely that this will be the answer to the issues in the RTF report. If you insist on giving a statement about that in your presentation, you need to find your own interpretation of the 'bug' in the spec. A final answer will be in the RTF report. Finding and resolving such bugs in the spec is important to improve the acceptance of CCM. Thanks to all who help in this process. Cheers, Tom Zitiere Craig Rodrigues : > On Fri, May 02, 2003 at 05:23:31PM +0200, ritter@fokus.fraunhofer.de > wrote: > > Hi Scott, > > > > I do not stick to the names. Though I think that the other four > 'ports' are > > somehow different from attributes. And I think that the components > equivalent > > interface (component reference) is more of a port than the attributes. > > > > The fact that there is fundamental disagreement on very basic > CCM terminology is very worrisome. Having terminology > that is agreed upon is very important when you have to explain > this new and emerging technology to development teams and managers. > I have to give a CCM presentation at the OMG Real-time Workshop > in July, so I would like to have this particular issue clarified > before I give my presentation. > > http://www.omg.org/issues/components-ftf.html#Issue5918 > > Based on the discussion thread on this mailing list, > I am still very confused whether an attribute is a port or not. > > When is the next Components revision task force? > > -- > Craig Rodrigues Distributed Systems and Logistics, Office 6/325A > crodrigu@bbn.com BBN Technologies, a Verizon company > (617) 873-4725 Cambridge, MA > > _______________________________________________ > CCM mailing list > CCM@moriarty.dif.um.es > http://moriarty.dif.um.es/mailman/listinfo/ccm > Supported by The CCM Page: http://www.ditec.um.es/~dsevilla/ccm > Date: Fri, 2 May 2003 15:31:52 -0400 From: Craig Rodrigues To: ccm@moriarty.dif.um.es Cc: components-rtf@omg.org Subject: Re: [CCM] Are attributes ports? User-Agent: Mutt/1.4i On Fri, May 02, 2003 at 08:59:19PM +0200, ritter@fokus.fraunhofer.de wrote: > Hello Craig, > > first of all, I do believe that the discussion about calling attributes ports or > not is not that much important as it seems to be. It is important that > components can have attributes and on that all do agree, I hope ;-) . It may not be important if an attribute is a port or not. However, correct and precise terminology is very important to me because I will be responsible for educating a room full of people about CCM. If the terminology is not clear, then it will be difficult to explain the fundamental concepts behind the terms. At first I thought that section 1.1.2 contained a typographical error, but was further confused by the discussion on the list. I don't care whether an attribute is a port or not....I just want attributes to be clearly defined in the CCM spec with wording that is not ambiguous or confusing to newcomers. For now, I will refer to attributes as "surface features" of a component. Thanks. -- Craig Rodrigues Distributed Systems and Logistics, Office 6/325A crodrigu@bbn.com BBN Technologies, a Verizon company (617) 873-4725 Cambridge, MA From: ritter@fokus.fraunhofer.de X-Authentication-Warning: saturn.fokus.fraunhofer.de: nobody set sender to tri@fokus.fhg.de using -f To: ccm@moriarty.dif.um.es, "J. Scott Evans" Subject: Re: [CCM] Are attributes ports? Date: Fri, 02 May 2003 22:00:00 +0200 (MET DST) Cc: components-rtf@omg.org User-Agent: IMP/PHP IMAP webmail program 2.2.7 > resolution for the issue. Tom, are you willing to submit a resolution > for > the issue with specific changes in wording in the spec? Yes I am. I will do this when the next RTF is established. Tom