RE: Lack of complete Instance semantics in SysML

[edi...@uml-forum.com]

> I find this particularly odd given the fact that the SysML Partners submission

> of the SysML spec (SysML Spec v1.0a) DID NOT contain semantics for

> instance diagrams while the original SysML Submission Team ...

Actually, it's more ironic than odd. During the time when the two SysML submission teams were actively competing Instance syntax/semantics was a key design issue on which they seemed to agree at a high level. Indeed, the SysML Partners planned to incorporate Instance syntax/semantics in the next (beta) version of their specification.

Unfortunately, politics appears to have prevailed over technology here as well as elsewhere in the OMG SysML final spec. Of the three visual modeling standardization efforts I chaired between 1997-2006 (UML1, UML2, and SysML1) SysML was by far the most political and least technical effort. Many of the persons contributing to SysML did not seem to understand basic language design principles, nor did they appear to comprehend the syntax and semantics of SysML's parent language (UML2). Consequently, it shouldn't be surprising that there is a lot of Visio <<keyword>> imagineering (a.k.a. voodoo semantics) in the final spec.

Overall, I'd rate OMG SysML1 as a 3.5/5 overall in terms of language design and practicality.

So what does this mean in practical terms? Does it mean that I am reluctant to recommend OMG SysML for large scale systems engineering projects? No, but I certainly need to caveat its limitations and provide workarounds far more than I would like!

What can we do to improve OMG SysML? Please send constructive criticism to the OMG (iss...@omg.org) and your preferred tool vendor.

/Cris

__________________________________________________
Cris Kobryn
Editor, UML Forum
edi...@UML-Forum.com

-------- Original Message --------
Subject: [UML Forum] Re: [SysML Forum] Lack of complete Instance
semantics in SysML
From: "Estefan, Jeff A (3100)" <jeffrey....@jpl.nasa.gov>
Date: Tue, October 20, 2009 10:06 am
To: "sysml...@googlegroups.com" <sysml...@googlegroups.com>,
"umlf...@googlegroups.com" <umlf...@googlegroups.com>


Cris et al., I find this particularly odd given the fact that the SysML Partners submission of the SysML spec (SysML Spec v1.0a) DID NOT contain semantics for instance diagrams while the original SysML Submission Team (SST) spec submissions (SysML Spec v0.98) did yet was later dropped when the official OMG SysML spec v1.0 was ratified. Please enlighten us. - JAE

[edi...@sysmlforum.com]

Laurent,

> I also would be interested to know more about why this is a huge "design
> flaw". Can you please elaborate Cris?

If you didn't understand my previous comments or the SysML FAQ regarding
this topic, I recommend that you carefully read Russell Peak's 22
October response to this thread, in which he recognizes "the need to
support instances explicitly" and notes that "the OMG SysML Revision
Task Force (RTF) has proposed the addition of instance notation to SysML
1.2."

Re your claims about "Context-Specific Properties": Where are the syntax
and semantics for these precisely defined in the OMG SysML
specification? The technical content re this topic in OMG SysML 1.1
tends to be trivial, and appears to be yet another case of voodoo
semantics that can be interpreted by N tool vendors in N different ways.

BTW, I noticed that your signature indicates you are "IBM's SysML
Representative" and Russell's signature indicates that he is the "OMG
SysML RTF representative for Georgia Tech". Your separate replies to
this topic appear to be inconsistent at best, and contradictory at
worst. (Don't RTF members discuss basic design issues such as this with
each other before proposing minor revisions?) What should we expect to
see in OMG SysML 1.2 to support Instance syntax/semantics? More
"Context-Specific Properties" imagineering, the reuse of UML2 Instance
syntax/semantics, or something else?

Best,

Cris

__________________________________________________
Cris Kobryn
Editor, SysML Forum
mailto:edi...@SysMLforum.com
www.SysMLforum.com

> -------- Original Message --------
> Subject: [SysML Forum] Re: Lack of complete Instance semantics in SysML
> From: Laurent L Balmelli <balm...@us.ibm.com>
> Date: Tue, October 20, 2009 8:49 pm
> To: sysml...@googlegroups.com
> Cc: edi...@SysMLforum.com, sysml...@googlegroups.com
>
>
> I also would be interested to know more about why this is a huge "design
>
> flaw". Can you please elaborate Cris?
>
>
>
> The assertion in this thread that "you cannot assign value to constraint
>
> blocks because there is not instance semantics" is actually not correct.
>
>
>
> In this case, SysML replaces the need for instances by the concept of
>
> Context-Specific Properties. When this is supported by the tool
>
> environment, you can assign values to the parameters in constraint
>
> properties created from the constraint block that are specific (i..e
>
> local) to its usage. This is much simpler that the concept of creating
>
> instances and provides the same benefits. There always has been an
>
> ambiguity in UML between the concept of a role of a classifier and an
>
> instance. SysML provides a nice middle ground to it. IBM has an
>
> implementation that will be part of a future tool release, but that we
>
> already provide it as a service to our customers, in particular in the
>
> scope of parametric models simulation.
>
>
>
> There is certainly applications that make good usage of instances,
>
> especially in embedded systems design, but I have not seen so far in this
>
> thread an example that was not covered by the actual SysML specification.
>
>
>
> Please share your examples, that would be very interesting!
>
>
>
> ----
>
> Laurent Balmelli, Ph.D (春芽利　楼蘭）
>
> Manager, IBM's SysML Representative
>
> Systems and Product Lifecycle Research and Development
>
> Rational Development & Services, Yamato Software Lab
>
> 1623-14, Shimotsuruma, Yamato-shi, Kanagawa-ken
>
> 242-8502, Japan
>
> Phone: +81 46 215 4868, Cell +81 80 6597 0578
>
>
>
>
>
>
>
>
>
> nsowatsk <nsow...@cisco.com>
>
> Sent by: sysml...@googlegroups.com
>
> 2009/10/19 22:10
>
> Please respond to
>
> sysml...@googlegroups.com
>
>
>
>
>
> To
>
> <sysml...@googlegroups.com>, <edi...@SysMLforum.com>
>
> cc
>
>
>
> Subject
>
> [SysML Forum] Re: Lack of complete Instance semantics in SysML
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> Hi Cris
>
>
>
> Did you send a follow up on the details of the instance semantics in
>
> SysML?
>
>
>
> Many thanks
>
>
>
> Nathan
>
>
>
>
>
> On 15/10/2009 10:51, "Dragos DOBRE" <dobre....@gmail.com> wrote:
>
>
>
> >
>
> > Hello Cris,
>
> >
>
> > I would like to have your opinion about the lack of complete Instance
>
> > semantics in SysML.
>
> >
>
> > Regards,
>
> >
>
> > Dragos DOBRE
>
> > Doctorant - Nancy Université - Université Henri Poincaré
>
> > Centre de Recherche en Automatique de Nancy - CRAN
>
> > Faculté des Sciences, BP 70239
>
> > 54506 Vandoeuvre-lès-Nancy
>
> >
>
> >
>
> >
>
> > -----Original Message-----
>
> > From: sysml...@googlegroups.com [mailto:sysml...@googlegroups.com]
>
> On
>
> > Behalf Of edi...@SysMLforum.com
>
> > Sent: jeudi 15 octobre 2009 02:36
>
> > To: sysml...@googlegroups.com
>
> > Cc: UMLf...@googlegroups.com
>
> > Subject: [SysML Forum] Re: Multiple element instances in a deployment
>
> diagram
>
> >
>
> >
>
> > Hello Paul,
>
> >
>
> > The questions you pose conflate modeling language and tool issues, so in
>
> > my reply I will endeavor to separate the two.
>
> >
>
> >> I would like to be able to have multiple instances of an element
>
> >> (node, device, etc.) in a deployment diagram (for example, multiple
>
> >> client stations connected to a server).
>
> >
>
> > First, we need to clarify that Deployment diagrams are part of the UML
>
> > language rather its SysML dialect, where SysML is defined as a UML
>
> > Profile. (For this reason I am cc-ing my response to the UML Forum, so
>
> > that other UML experts can also contribute to this thread.) In addition,
>
> > the lack of complete Instance semantics in SysML is a critical language
>
> > design flaw, which was noted before the language was formally adopted by
>
> > the OMG, but which has not yet been corrected. (If you want to
>
> > understand this critical design flaw better, please let me know and I
>
> > will be glad to elaborate.) On the other hand, the lack of Deployment
>
> > diagrams in SysML can be viewed as a design virtue rather than a flaw,
>
> > since deployment relationships (cf. structure-to-structure allocations)
>
> > can be easily defined in SysML by defining appropriate stereotypes for
>
> > selected SysML structural elements and allocation dependencies.
>
> >
>
> > You are correct that, according to the UML specification, you should be
>
> > able to define multiple instances of Nodes, Components, or other
>
> > allowable Classifiers on a UML Deployment diagram.
>
> >
>
> >> I am using EA and it, at least, will not let me do so. Is there a way
>
> to get
>
> >> around this or do
>
> >> what I want to do?
>
> >
>
> > There is no need for a workaround in Enterprise Architect (EA) since for
>
> > the most part EA's Instance syntax and semantics comply with the UML
>
> > specification. I use most of the mainstream UML/SysML modeling tools
>
> > frequently, including EA, and I just double-checked this capability
>
> > using EA v. 7.5 Deployment diagrams. It works fine! I suspect you may be
>
> > confused by, or misusing, EA's "Paste Element" dialog. When
>
> > copy-and-pasting a Node or other allowable Classifier (non-instance)
>
> > Element from the Project Browser to the target Deployment diagram, you
>
> > need to click on the second radio button labeled "an Instance of Element
>
> > (Object)" rather than the first radio button labeled "as Simple Link".
>
> > The former will allow you to define multiple instances of a Node or
>
> > other Classifier, the other will not since it is a reference link, not a
>
> > bona fide Instance.
>
> >
>
> >> I'd also like to represent two (or more) environments separated by
>
> >> time. These would represent sequential aspects of a workflow (from a
>
> >> deployment perspective). ...
>
> >
>
> > You are correct that chronologically ordering Instance "snapshots" can
>
> > be a useful way to compare and contrast environments as they evolve over
>
> > time. Here, however, I suspect you may encounter problems with how EA
>
> > handles Instance semantics over time, since this is a relatively
>
> > sophisticated technique.
>
> >
>
> > /Cris
>
> >
>
> > __________________________________________________
>
> > Cris Kobryn
>
> > Editor, SysML Forum
>
> > mailto:edi...@SysMLforum.com
>
> > www.SysMLforum.com
>
> >
>
> >> -------- Original Message --------
>
> >> Subject: [SysML Forum] Multiple element instances in a deployment
>
> >> diagram
>
> >> From: Paul Ebert <a2e...@gmail.com>
>
> >> Date: Wed, October 14, 2009 6:36 am
>
> >> To: SysML Forum <sysml...@googlegroups.com>
>
> >>
>
> >>
>
> >> I would like to be able to have multiple instances of an element
>
> >> (node, device, etc.) in a deployment diagram (for example, multiple
>
> >> client stations connected to a server). I am using EA and it, at
>
> >> least, will not let me do so. Is there a way to get around this or do
>
> >> what I want to do?
>
> >>
>
> >> I'd also like to represent two (or more) environments separated by
>
> >> time. These would represent sequential aspects of a workflow (from a
>
> >> deployment perspective). An example might be a sales person on the
>
> >> road, giving presentations and gathering customer data at customer
>
> >> sites and then connecting to the home office network at the end of the
>
> >> day to load the data into an application or to access email. Again,
>
> >> in this situation, I'll need multiple instances of an element (well,
>
> >> they are actually the same instance at the abstract level, but are
>
> >> multiple instances from the tool's perspective).
>
> >>
>
> >> Any suggestions? Thanks.
>
> >>
>
> >>
>
> >
>
> >
>
> >
>
> > >
>
>
>
>
>
>
>
>
>
>
>
>
>
>

[Russell Peak - GIT]

For those on this list who may not be familiar with the OMG revision task
force (RTF) process, you can see something about it here and on other
similar OMG pages:
- http://www.omg.org/technology/documents/spec_catalog.htm

It should be clear that any individual OMG RTF member speaks only for
himself/herself, and that it is RTF reports and other documents such as
ptc/09-08-13 that capture the official collective positions of the RTF
(which in turn are not official OMG positions until approved by the OMG
board).

Those familiar with standards/spec processes in general will know that there
are often competing and contradictory opinions on the same committee, as by
their very nature such committees are composed of diverse organizations and
individuals.

Another key point is that a standard/spec intended for broad usage usually
attempts to support a variety of usage styles. For example, regarding this
current topic, one organization may prefer to use context-specific
properties because of their particular tools and methods, while another may
prefer instances. Thus a given organization may not need some capabilities
of the spec that another organization heavily relies on.

Therefore, what would be surprising is if all RTF members had the same
opinions and viewed the spec the same way. And if there were no issues.

As others have said "What can we do to improve OMG SysML? Please send

constructive criticism to the OMG (iss...@omg.org) and your preferred tool
vendor."

Here are some useful links regarding issues:
- http://www.omg.org/spec/SysML/
- http://www.omg.org/issues/sysml-rtf.open.html
- http://www.omg.org/technology/issuesform.htm

Russell

OMG SysML RTF representative for Georgia Tech

---------------------------------------------------
Russell S. Peak, PhD
Senior Researcher

Director
Modeling & Simulation Lab
http://www.msl.gatech.edu/

Associate Director
Product & Systems Lifecycle Management Center
http://www.pslm.gatech.edu/

Georgia Institute of Technology
813 Ferst Drive, MARC 373
Atlanta, Georgia 30332-0560 USA
voice +1-404-894-7572
Russel...@gatech.edu
---------------------------------------------------

-----Original Message-----
From: sysml...@googlegroups.com [mailto:sysml...@googlegroups.com] On
Behalf Of edi...@SysMLforum.com
Sent: Sunday, October 25, 2009 2:29 AM
To: sysml...@googlegroups.com
Subject: [SysML Forum] Re: Lack of complete Instance semantics in SysML
...

BTW, I noticed that your signature indicates you are "IBM's SysML
Representative" and Russell's signature indicates that he is the "OMG
SysML RTF representative for Georgia Tech". Your separate replies to
this topic appear to be inconsistent at best, and contradictory at
worst. (Don't RTF members discuss basic design issues such as this with
each other before proposing minor revisions?) What should we expect to
see in OMG SysML 1.2 to support Instance syntax/semantics? More
"Context-Specific Properties" imagineering, the reuse of UML2 Instance
syntax/semantics, or something else?

...