METHOD CHUNK FEDERATION ISABELLE MIRBEL EMMSAD 2006

Transcription

1 METHOD CHUNK FEDERATION ISABELLE MIRBEL EMMSAD 2006 LABORATOIRE I3S Route des Lucioles, BP Sophia Antipolis, Cedex France

2 OUTLINE 1. INTRODUCTION 2. MAKING PROJECT-SPECIFIC METHODS FEDERABLE 2.1 METHOD MEANINGFUL ATOMIC PARTS 2.2 THE REUSE FRAME 3. SUPPORTING METHOD CHUNK FEDERATION 3.1 METHOD CHUNK CONTEXT 3.2 METHOD USER SITUATION 3.3 SIMILARITY METRICS 3.4 EXTENDED SIMILARITY METRICS 4. CONCLUSION

3 CONTEXT INTRODUCTION Need for customization Situational Method Engineering Organization-wide standard approaches Merging of project specific methods Capture & understanding of all the project specific methods by method engineers Acceptation & understanding of the organization wide method by method users Method Chunk Federation

4 INTRODUCTION PROJECT SPECIFIC METHOD FEDERATION Make project specific method federable Break down method into meaningful atomic parts Qualify meaningful atomic parts Federate federable method parts Method Chunk Federation

5 OUTLINE 1. INTRODUCTION 2. MAKING PROJECT-SPECIFIC METHODS FEDERABLE 2.1 METHOD MEANINGFUL ATOMIC PARTS 2.2 THE REUSE FRAME 3. SUPPORTING METHOD CHUNK FEDERATION 3.1 METHOD CHUNK CONTEXT 3.2 METHOD USER SITUATION 3.3 SIMILARITY METRICS 3.4 EXTENDED SIMILARITY METRICS 4. CONCLUSION

6 METHOD MEANINGFUL ATOMIC PARTS (1) MAKING PROJECT-SPECIFIC METHODS FEDERABLE Method chunk from J. Ralyté Process & product dimensions Assembly techniques provided METHOD CHUNK: Autonomous & coherent part of a method supporting the realization of some specific Information System & Software Development (ISSD) activities METHOD: Set of loosely coupled method chunks expressed at different levels of granularity Method Chunk Federation

7 METHOD MEANINGFUL ATOMIC PARTS (2) MAKING PROJECT-SPECIFIC METHODS FEDERABLE Input of the chunk Reuse context To extend the contextual view Descriptor Interface: <Situation, Intention> Product Part Product to be delivered Body Process Part Intention/goal the chunk achieves Guidelines allowing to produce the product Simple guidelines Tactical guidelines Strategic guidelines Method Chunk Federation

8 METHOD CHUNK EXAMPLE MAKING PROJECT-SPECIFIC METHODS FEDERABLE Method chunk Behavior Investigation <User Interface Description, Standardize UI business rules behavior> UML notation: - Activity diagram - Class diagram Some business rules may be generic and associated to the whole software while called in several use-cases and windows. To be coherent through the whole UI specification, these business rules have to be isolated in a specific class diagram and explicitly used in the diagrams describing each concerned windows. A typical example of these generic business rules is the control among two dates (to be sure a first date is before the second one, for instance). Give a class diagram and activity diagrams to summarize the general business rules. Taken from the JECKO methodology (I3S/Amadeus) Method Chunk Federation

9 OUTLINE 1. INTRODUCTION 2. MAKING PROJECT-SPECIFIC METHODS FEDERABLE 2.1 METHOD MEANINGFUL ATOMIC PARTS 2.2 THE REUSE FRAME 3. SUPPORTING METHOD CHUNK FEDERATION 3.1 METHOD CHUNK CONTEXT 3.2 METHOD USER SITUATION 3.3 SIMILARITY METRICS 3.4 EXTENDED SIMILARITY METRICS 4. CONCLUSION

10 METHOD CHUNK CLASSIFICATION MAKING PROJECT-SPECIFIC METHODS FEDERABLE Classification & retrieval techniques Based on user intention & structural relationships Method dependent Based on user intention & application domain Domain dependent Based on user intention & ISSD knowledge The REUSE FRAME Method Chunk Federation

11 THE REUSE FRAME MAKING PROJECT-SPECIFIC METHODS FEDERABLE Ontology dedicated to ISSD activities Shared by all projects & project members Method Chunk descriptor: set of at least one keyword taken from the Reuse Frame Reuse Frame content proposal Method Chunk Federation

12 THE REUSE FRAME CONTENT MAKING PROJECT-SPECIFIC METHODS FEDERABLE Universal/critical IS aspects, meaningful reuse keywords Human Organizational Application Domain Expert Analyst Different method users (developer, designer, product definition responsible, test manager,... ) Different levels of expertise (expert, beginner) Contengency factors Legacy System Business Modelling Project characteristics, goals & assumptions System Engineering Activities Application type Source system Application technology Method Chunk Federation

13 THE REUSE FRAME STRUCTURE (1) MAKING PROJECT-SPECIFIC METHODS FEDERABLE Refinement relationships Refinement into more specific aspects Graphical User Interface Application Technology Databases Refinement into more specific and classified aspects Weak Code Reuse 1 Code Reuse 2 Medium Code Reuse Strong Code Reuse 3 Refinement into more specific and exclusive aspects Inter-organization application Application Type Intra-organization application Organization-customer application Method Chunk Federation

14 THE REUSE FRAME STRUCTURE (2) MAKING PROJECT-SPECIFIC METHODS FEDERABLE Refinement into more specific and exclusive aspects Node exc Nodes exc Node exc Node Node Node Node Node ord, seq ord, seq Refinement into more specific aspects Node Node Refinement into more specific and classified aspects Method Chunk Federation

15 EXAMPLE OF REUSE FRAME CONTENT THE REUSE FRAME Contingency factor Organizational Root Application Domain Human exc Node Aspect Classified aspect Exclusive aspect Source System exc User Involvement Real User Virtual User Management Commitment exc Low exc High Application Type Strong Reuse No Source System 3 2 Medium Reuse Code Reuse 1 Weak Reuse Full description available at : Legacy System Interface Reuse Functional Domain Reuse Method Chunk Federation

16 THE REUSE FRAME STRUCTURE (4) MAKING PROJECT-SPECIFIC METHODS FEDERABLE Inclusion between aspects An aspect a1 is included in an aspect a2 if the path from the root node to a1 is a sub-path of the path from the root node to a2. An aspect a1 includes an aspect a2 if the path from the root node to a2 is a sub-path of the path from the root node to a1. Precedence between aspects An aspect a1 is previous an aspect a2 if they have the same father node and cl(a1) < cl(a2) An aspect a1 is next an aspect a2 if they have the same father node and cl(a1) > cl(a2) cl, 1 e d cl, 2 Root Organi- zational e is previous f f Application Domain a b includes a Method Chunk Federation c b

17 THE REUSE FRAME STRUCTURE (5) MAKING PROJECT-SPECIFIC METHODS FEDERABLE Compatibility between aspects Compatible aspects do not share in their path (from the root node) a node with exclusive leaving edges. b & c are not compatible exc b exc a c Root Application Domain exc d e f f & g are compatible g Method Chunk Federation

18 OUTLINE 1. INTRODUCTION 2. MAKING PROJECT-SPECIFIC METHODS FEDERABLE 2.1 METHOD MEANINGFUL ATOMIC PARTS 2.2 THE REUSE FRAME 3. SUPPORTING METHOD CHUNK FEDERATION 3.1 METHOD CHUNK CONTEXT 3.2 METHOD USER SITUATION 3.3 SIMILARITY METRICS 4. CONCLUSION

19 SUPPORTING METHOD CHUNK FEDERATION INTRODUCTION Mechanism(s) to extract meaningful method chunks from the federation Meaningful method chunks for project members To cope with ISSD activities covered by the projectspecific method Alternative guidelines Similarity Metrics to compare method chunks To cope with ISSD activities not (well) covered by the project-specific method Complementary guidelines Similarity Metrics to compare method chunks to user need Method Chunk Federation

20 SUPPORTING METHOD CHUNK FEDERATION METHOD CHUNK CONTEXT A set of at least one aspect to qualify a method chunk Aspects can not be included one in the others All aspects must be compatible among them Method chunks providing general guidelines should be qualified with the help of less refined aspects (i.e. close to the root node). Method chunks providing specific guidelines should be qualified with the help of more refined aspects (i.e. close to the leaf nodes or leaf nodes themselves). Method Chunk Federation

21 SUPPORTING METHOD CHUNK FEDERATION METHOD CHUNK CONTEXT EXAMPLE Behavior Investigation:descriptor - Chunk Name: Behavior Investigation - Chunk Id: DBOA-Inv-behavior - Type: atomic - Objective: Guidelines to help in standardizing UI business rules behavior JeckoFramework:Method :Reuse Intention Description: Standardize UI business rules behavior :Reuse Context Aspects: - High Project Complexity, - High Project Formality, - High Project Size, - Analysis, - Strong Tracing. Method Chunk Federation

22 OUTLINE 1. INTRODUCTION 2. MAKING PROJECT-SPECIFIC METHODS FEDERABLE 2.1 METHOD MEANINGFUL ATOMIC PARTS 2.2 THE REUSE FRAME 3. SUPPORTING METHOD CHUNK FEDERATION 3.1 METHOD CHUNK CONTEXT 3.2 METHOD USER SITUATION 3.3 SIMILARITY METRICS 3.4 EXTENDED SIMILARITY METRICS 4. CONCLUSION

23 SUPPORTING METHOD CHUNK FEDERATION METHOD USER SITUATION (1) A set of at least one aspect to specify necessary aspects All aspects must be compatible among them Searches for general guidelines less refined aspects (close to the root node) Searches for specific guidelines more refined aspects, that is to say aspects close to the leaf nodes or leaf nodes themselves. A set of aspects to specify forbidden aspects All aspects must be compatible among them Method Chunk Federation

24 SUPPORTING METHOD CHUNK FEDERATION METHOD USER SITUATION (2) Constraints between necessary & forbidden aspect sets No common aspects between necessary & forbidden aspects For optimisation purpose (to avoid redundancies) No inclusion between necessary & forbidden aspects It is not possible by definition to find two aspects included one in the other in the same method chunk context No incompatibility between necessary & forbidden aspects It is not possible by definition to find two incompatible aspects in the same methodchunk context Method Chunk Federation

25 METHOD USER SITUATION EXAMPLE : Method User Situation SUPPORTING METHOD CHUNK FEDERATION Necessary Criteria: - Graphical User Interface - Database - High formality - Analyst Forbidden Criteria: - Design Eng. Activity - High time pressure Method Chunk Federation

26 OUTLINE 1. INTRODUCTION 2. MAKING PROJECT-SPECIFIC METHODS FEDERABLE 2.1 METHOD MEANINGFUL ATOMIC PARTS 2.2 THE REUSE FRAME 3. SUPPORTING METHOD CHUNK FEDERATION 3.1 METHOD CHUNK CONTEXT 3.2 METHOD USER SITUATION 3.3 SIMILARITY METRICS 3.4 EXTENDED SIMILARITY METRICS 4. CONCLUSION

27 SUPPORTING METHOD CHUNK FEDERATION SIMILARITY METRICS (1) Between method chunks Number of common aspects 0 < SM < 1 Between method chunk & user situation Number of common aspects / necessary aspects Number of common aspects / forbidden aspects SM < 1 Does not take into account the number of aspects in the context Method Chunk Federation

28 SUPPORTING METHOD CHUNK FEDERATION SIMILARITY METRICS (2) Comparison between method chunks An aspect from the method chunk context Σ i=1..n d(a, CA ) RC c1i c2 sm(c1,c2) = max(card(ca C1 ), card(ca C2 )) Method chunk context 0 < sm(c1,c2) < 1 d(a,a) = 1 if a A, 0 else Method Chunk Federation

29 SIMILARITY METRICS (3) An aspect from the forbidden aspect set (Method user situation) Method chunk context sm(c,s) = Σ i=1..n d(a NA si, CA c ) - Σj=1..m d(a FA sj, CA c ) card(na s ) An aspect from the necessary aspect set (Method user situation) Necessary aspect set (Method user situation) SUPPORTING METHOD CHUNK FEDERATION sm(c,s)>0 sm(c,s)<0 sm(c,s)=1 - card(ca c FA s ) card(na s ) More necessary aspects than forbidden ones More forbidden aspects than necessary ones Perfect adequation between situation & context Worst situation with adequation only on forbidden aspects Method Chunk Federation

30 SUPPORTING METHOD CHUNK FEDERATION SIMILARITY METRICS EXAMPLES Reuse Context Method Chunk 1 Real User Involvement Low Time Pressure Huge Project Size Method Chunk 2 Real User Involvement Low Time Pressure Small Project Size sm(mc1,mc2)=0.66 sm(a,mc1)=1 sm(a,mc2)=0.5 sm(b,mc1)=-0.5 sm(b,mc2)=0 Necessary aspects Forbidden aspects User Situation A Real User Involvement Low Time Pressure Small Project Size Low Project Formality User Situation B Real User Involvement Strong Project Traceability Huge Project Size Low Time Pressure Method Chunk Federation

31 OUTLINE 1. INTRODUCTION 2. MAKING PROJECT-SPECIFIC METHODS FEDERABLE 2.1 METHOD MEANINGFUL ATOMIC PARTS 2.2 THE REUSE FRAME 3. SUPPORTING METHOD CHUNK FEDERATION 3.1 METHOD CHUNK CONTEXT 3.2 METHOD USER SITUATION 3.3 SIMILARITY METRICS 3.4 EXTENDED SIMILARITY METRICS 4. CONCLUSION

32 SUPPORTING METHOD CHUNK FEDERATION CLOSENESS DISTANCE (1) Federation interest : Ability to propose new & meaningful method chunks Partial matching to be considered Method chunk context included in the method user situation Exploitation of the Reuse Frame relationships more generic aspects more specific aspects previous aspects next aspects Method Chunk Federation

33 SUPPORTING METHOD CHUNK FEDERATION EXTENDED SIMILARITY METRICS (1) Extension possibilities Exact Matching Less refined aspects Extended Matching More refined aspects before/after aspects Necessary aspects Forbidden aspects to search for method chunks to avoid method chunks More general chunks to avoid full branches of the Reuse Frame to retrieve more method chunks More specific chunks to retrieve less method chunks to avoid too specific chunks Adjacent method chunks to avoid adjacent / overlapping chunks Method Chunk Federation

34 SUPPORTING METHOD CHUNK FEDERATION CLOSENESS DISTANCE (2) Closeness between aspects Exploitation of the Reuse Frame relationships Aspect extension The extension ext(a) of an aspect a is the set of all the aspects: more generic more specific previous it next it Quantification of the distance with regards to the reference situation (i.e. aspects given by the method user) 1 prev. generic next specific Method Chunk Federation

35 SUPPORTING METHOD CHUNK FEDERATION CLOSENESS DISTANCE (3) Example of closeness distance Closeness distance for more generic aspects root a b cd(a,d) = 0.25 cd(b,d) = 0.5 a c cd(c,d) = 0.75 b d cd(d,d) = 1 c d Method Chunk Federation

36 SUPPORTING METHOD CHUNK FEDERATION EXTENDED SIMILARITY METRICS (2) Σ i=1..n ed(a, CA ) - NA si c Σj=1..m ed(a, CA ) FA sj c sm(c,s) = card(na s ) if ext(a) A =, ed(a,a) = 0 else ed = max cd(y,z) with y ext(a) A z A Method Chunk Federation

37 SUPPORTING METHOD CHUNK FEDERATION EXAMPLE OF EXTENDED SIMILARITY METRICS a root 3 4 MC1 2, 14, 23 MC2 7, 11 ext(a) C mc1 = {2} ed (a,c mc1 ) = cl(2,a) = 0,5 sm (C mc3,c mc1 ) = 0,5 ext(a) C mc2 = ed (a,c mc2 ) = 0 sm (C mc3,c mc2 ) = a US Ext(a) = {root, 2, 5, a, 4, 3, 8, 9, 1} Method Chunk Federation

38 OUTLINE 1. INTRODUCTION 2. MAKING PROJECT-SPECIFIC METHODS FEDERABLE 2.1 METHOD MEANINGFUL ATOMIC PARTS 2.2 THE REUSE FRAME 3. SUPPORTING METHOD CHUNK FEDERATION 3.1 METHOD CHUNK CONTEXT 3.2 METHOD USER SITUATION 3.3 SIMILARITY METRICS 3.4 EXTENDED SIMILARITY METRICS 4. CONCLUSION

39 SUMMARY CONCLUSION Method Chunk Federation Situational Method Engineering Organization-wide standard approaches The Reuse Frame Method Chunk Context Method User Situation Similarity Metrics Method Chunk Federation

40 FUTURE WORK CONCLUSION About projects Project specific view & vocabulary layer on the Reuse Frame Dynamic filtering of new and meaningful method chunks Turning context & situation into profiles To compare user profiles About method chunks Presentation & integration of the retrieved method chunks in the project-specific methods Method chunk comparison (based on guidelines,...) About the Reuse Frame To move the ontology from the federation to Project level Method chunk level Method Chunk Federation