RJLRef: $PH/06f522/OMGPlenaryUML2MDAConcepts.{htm,txt} 

(PH is http://www.cs.uml.edu/~lechner)                        

 

$PH\OMG\CompuwarePlenary04-02-01Links.htm has links to 3 slide sets,

all from OMG meeting sponsor pep talks on UML2-Model-Driven Arch.

[A new 4th item below is only referenced here not in ...Links.htm.]

 

I outlined some notes below on the second presentation (03-09-03.doc),

which tries to explain automatic translation from abstract models

to actual code. But this seems to open up a new requirement/opportunity

for UML2/MDA tool vendors: 3D visualization of the process.

 

==============================================

 

I. $PH/OMG/CompuwarePlenary04-02-01.ppt : MDA & OptimalJ

(13 simple slides from CompuWare - basic concepts [hype] only)

 

==========================================

 

II. $PH/OMG/OMGBostonCompuwareExample03-09-03.doc

This is a condensed example of automatic 'model'-to-'code'

conversion - from UML TO RDBMS.

 

This '.doc' file has 10 very dense .ppt diagrams but no text.

It covers UML Model to RDBMS Model transformation of

the 'Rosa's Breakfast Service' case study.

 

An interesting aspect of this data model is a M-to-N association

between items in a customized breakfast order and the 'comestibles'

(groceries) in one of the standardized breakfasts on the menu.

'Customization' only specifies departures (e.g. no salt?)

from the standardized menu entry (i.e., a list of comestibles).

 

For more details on Rosa's Breakfast Service, see

Chapters 4-6 of "MDA Explained" (A-W 2003) by

Kleppe, Warmer and Bast [Bast is CompuWare's Optimal-J architect].

In MDA Explained, this model was converted into three distinct

(but trivial) code implementations: Relational, Web-based, and EJB.

(Don't expect state models - order entry doesn't need them.)

 

=============================================

 

III. $PH/OMG/OMGLondon_Compuware03-11-05.ppt - The MDA Reference Model.

 

This 48-slide presentation again uses the Breakfast example,

but in a much grander context (see slides 10-12).

[Don't ask what ORMSC means - I can't find my  bookmark to

OMG's online acronym glossary. :-(] Below are my notes for some slides:

 

-----------------------------------------

 

Slide 5: What is the  MDA Reference Model :

-A definition of the very basic concepts in MDA and their relations

-A class model: defined using only common modeling constructs

        that are shared by UML and MOF

-All MDA examples are instances of the MDA reference model

        where the MDA examples can include all meta levels,

        all modeling languages and all technologies

 

Slides 6-9: how, when, req'ts, who

Slide 9: "The ORMSC is mandated by the AB to develop a draft MDA Reference Model.

Within the ORMSC a colorful palate of different MDA vendors are active...."

 

(Uses the Breakfast Example from the OCL)

 

 Slide 17; Defined Concepts:

 Structure (typed graphs)

 Model

 Model Types

 Model Relations

 Transformation

 Transformation Types

 

 Slide 18: Defined Relation Types:

 Classification / Instantiation

 Usage Relations

 Composition

 Specialization

 Transformation

 

 Slide 20,21 is Node==>Edge graph Ptrn (upside-down)

 Outgoing edges = aggregate; incoming edges = upward-directed M:1 assocn.

 

 Slide 23,24 is map: Node-Edge to Model-ModelRelation.  (TypedGraphs to Models)

 (Edge is associated as an element of (M:N) to ModelRelation.

 Node has a similar M:N assn  from Model, PLUS: Model is_a subclass of Node!

 

 Slide 27: Transformations:

        ModelRelation-to-[Relating and related-to]Model[pair]

 in column 1 are subclassed to

        TransfmtnTracking (resp) Transfmtn in column 2.

 

 Slide 29,30: Typed Graphs: one column of 4 nodes. (3D is needed now)

 

 Node,Edge are subtyped to NodeType, EdgeType.

 [TYPO: Edgeype is shown as subclass of Node - which I think is wrong.]

 

Slide 32,33: makes same mistake? in first column = slide 30.

Column 2 maps/specializes middle two: Node to Model, NodeType to  ModelType,

and M:N associated Edge to ModelRelation, Node to Model,

EdgeType to ModelRelationType, but NOT NodeType to ModelType????.

 

Slide 35,36: same 4x8 array, Left side iscopyof Right side above,

Right side is TransfmtnTracking, Transfmtn, TransfmtnType, TransfmtnPort.

All 4 rows have specialization relations  left to right.

 

Slide  37 (title):

        The simple MDA Example as Instance of the MDA Reference Model

 

Slide 41: In color, Breakfast:ModelType pair and L2R transfmtn at bottom,

        UML and CWM Modelypes and transfmtn in middle,

 

Slide 43: Breakfast class model (3-type vertical chain models M:N assn.)

                This should be: (BrkfastOrder--->part>---comestible)

 

========================================================

 

IV.  $PH/OMG/ADMtutorial_admtf05-05-01.ppt  <--- Note new file name.

(This is from the Athena, Greece OMG meeting in 2005).

 

This 4MB, 109-slide, 3-hour .ppt tutorial is about

Architecture-Driven Modernization (ADM). See http://adm.omg.org/.

 

It presents only two of the seven components of ADM:

Knowledge Discovery Metamodel and Abstract Syntx Tree Metamodel.

 

Slide 7: This is a harder problem than the ORMSC

forward transformation (model to code) problem in III above.

ADM is about reverse engineering (code to model).

Refactoring is a fundamental ingredient of

evolutionary code improvement. ADM is pitched at a

much higher level - the goal is a meta-model 'home'

that would permit standardized interpretation of info

extracted by any technique or tool used for ADM.

 

(If this sounds like pie-in-the-sky, it is now,

but so were lots of HW and SW concepts

that we now take for granted.

So look and see what the future may hold.:-)

 

Slide 90 of 109: Why Marry MOF and parsing Technology ASTs

 

Slide 102: Their toolset JRGEN, JPGEN, JTGEN (very interesting).

Slide 102 refers to TSRI's Janus 5GL Wide-Spectrum Language

as derived from'Refine'. I have seen refs to that.

I don't now how it compares to Martin Ward's WSL.

 

Slide 105: The Relationship of ASTM to Other OMG Specs

 

Slides 101-107 are very dense and printed.