Talk:Meta-Object Facility

From Wikipedia, the free encyclopedia

This article is related to Wikipedia:WikiProject Method engineering.
Please note that the use of Wikipedia to host this project has been questioned. Please read this discussion and, if you wish, contribute your thoughts there.

Contents

[edit] Proposals for addition, requests for completion

[edit] Unresolved issues

  • SPEM
  • powertype
  • ISO/IEC 24744


[edit] QVT and model transformation languages

Removed following section from article (see new tentative 'spoiler' on QVT) -- since it's to broad on transformation languages, less specific on QVT (as part of MOF 2.0).

A very important new standard is QVT. This allows to transform any MOF-based model into another MOF-based model. 
Furthermore, the transformation program itself is considered as another MOF-based model. Current examples of 
Model Transformation Languages (MTLs) are VIATRA, Tefcat, GReAT, AndroMDA, or ATL.

Furthermore, it's Tefkat, not Tefcat. (fbahr. July 26, 2006.)

[edit] strange deletion

I do not agree with the previous removal made by fbahr above. This removal seems very badly motivated. The fact that QVT allows transformation based on MOF-models is an important and clarifying information. It gives more scope and understandibility to the article. I suggest to revert this action of fbahr. Furthermore the last remark made by fbahr about a spelling mistake (Tefkat, not Tefcat) makes me suspicious about all this. Why not simply correcting the mistake on simple Wikipedia ~practices? Strange. Myguest 19:38, 29 September 2006 (UTC)


[edit] Requests for clarification

[edit] OMG's three level modeling stack

This article is badly missing the classical picture of the three-level OMG stack. Who could provide one?

Three-level? ... As far as I did understand, OMG concerns 4 levels (3 "modeling levels" + 1 "real-world level"). Due to critics on the instance_of notion reg. two staged levels (leading to a more generic, (in my words) less concrete conforms_to notion; cf. J. Bézivin from Univ. de Nantes) found in many academic publications on the theme of metamodeling, OMG seems to restrict itself on the class/instance relation of two consecutive levels from the modeling space. You do think in 3 levels as below? [June 29, 2006]
M3: MOF -- the meta-metamodel
M2: MOF model -- a metamodel, such as the UML metamodel
M1: "user model" -- a metamodel instance


[edit] Closed metamodeling architecture

What is a closed metamodeling architecture? MDE 20:35, 10 May 2006 (UTC)


[edit] UML vs. MOF

We absolutely need to clarify on this page the relations between UML and MOF. Who is going to start with this difficult but important issue. I will add a heading and just a line, but we definitively need more.

UML is defined by a metamodel. The UML metamodel conforms to the MOF metametamodel like a UML model conforms to the UML metamodel. MDE 08:41, 27 May 2006 (UTC)
The Meta-Muddle again: "UML('s abstract syntax) is defined by a metamodel. The UML metamodel (which is a MOF model) conforms to the MOF !metamodel! (not metameta-) like a UML model conforms to the UML metamodel."
Additionaly, UML 2 and MOF 2 are founded on the same UML Infrastructure Library. [June 29, 2006]
A sound architecture differentiates the metametamodel (MOF), the library of metamodels like UML and the library of models. This situation has been transformed in a mess by the lobbying of some UML CASE tool vendors that wanted their tools to be able to edit not only UML models, but also MOF metamodels. There is no reason to mention more UML than SPEM or CWM or any other metamodel here. Trying to explain the relations between MOF and UML will only bring more confusion here because the debate is not technical but commercial. I suggest keeping only with the sound idea of MOF being a language to define DSLs and letting the UML article trying to make some sense of this mess. To the best of my knowledge, nobody has been able to come out with a consistent and understandable presentation of UML infrastructure vs. MOF at this time. Cleany 07:23, 2 October 2006 (UTC)

[edit] SMOF

(placeholder)

[edit] Disputed sections

[edit] Clabjects

What about clabjects? Should we keep this part? Does not seem a significant reference

Clabjects again: Seems strange. Look like this is the work of only one or two authors that has creeped through this article. The concept of a Clabject seems completely unknown in OMG literature. So why keeping this reference here? This MOF article needs a lot of cleaning. The Clabject story is typical. MOF is an OMG standard. The concept of Clabject is completely unknown in OMG literature. It seems that one paragraph is completely self-promotion. I suggest that the authors of the concept open a separate Wikipedia entry for Clabject and do not confuse the reader with someting this is absolutely not related to MOF as defined by the OMG. Does this seems reasonable?
Would prefer a deployment of the clabject concept in a separate "metamodeling principles" section; the same for "powerypes" (as analysis pattern?) [June 29, 2006]

[edit] AndroMDA

AndroMDA is not a "transformation language" but a code generator "engine".

Interesting. Could you elaborate on the difference between these? There is already an entry for model transformation language. Is there an entry for code generator engine? MDE 08:44, 27 May 2006 (UTC)
What is the relation between what you call a code generation engine and what OMG calls a Model to Text transformation tool. I would be very interested in knowing. Should we open a Wikipedia entry on Model to Text transformation? 12.192.193.2 20:26, 11 June 2006 (UTC)
OMG, acc. to my understanding, just differentiates between two (main) types of transformations depending on the artifacts involved (and their well-formedness w.r.t. to OMG's notion of this term). Model-2-Model (both defined in terms of their corresponding metamodel(s)) and Model-2-Text, where the latter (artifact) doesn't have to be explicitly captured by a firm metamodel (in practice, most commonly, using a style of template language). [June 29, 2006]

[edit] GReAT

GReAT isn't (directly) related to MOF, though not a "real" QVT-like language -- and should be removed from the listed examples.

Still GReAT is a Model Transformation Language, but not a QVT-like language. One of the main characteristics of a QVT-like language is that it is OCL-based. Do we agree on that? MDE 08:49, 27 May 2006 (UTC)
Essentially, any QVT-like language (as QVT is intended to be part of MOF 2) should comply to MOF 2; this would lead to a discussion on MOF-compliance of ISIS' GME ... w.r.t. to GReAT's primary (not solely) application domain ("semantic anchoring" or "denotation to semantic domains"), as far as I do understand, the languages' specific focuses tend to differ.


[edit] Discussion on formalities and formatting

[edit] Hyphenate or not?

Another question: should we hyphenate or not. I am not sure here if it should be written "Meta Object Facility" or "Meta-object Facility" or be kept as is. Any advice?

There is an ongoing discussion on this under meta-model or metamodel. MDE 08:46, 27 May 2006 (UTC)
Bad enough, that Model-driven architecture is misspelled (should be: "Model Driven Architecture"; cf. www.omg.org/legal. Accord. to the OMG's recommendation it's "Meta-Object Facility", though you also find "MetaObject Facility" as in here www.omg.org/mof.