Spl needs an automatic holistic model for software reasoning with feature models

INTERNATIONAL WORKSHOP ON REQUIREMENTS REUSE IN SYSTEM FAMILY ENGINEERING July 5 h, 2004, Ca los III Uni e si y o Mad id, Mad id, Spain Co-loca ed wi h In e na ional Con e ence on So wa e Reuse 8 POSITION PAPERS O ganise s: Juan C. Dueñas, Klaus Schimd Con en s In e na ional Wo kshop on Requi emen s Reuse in Sys em Family Enginee ing………………………………………………………….1 P og am commi ee…………………………………………………………………………………………………………………………………..2 P og amme……………………………………………………………………………………………………………………………………………..3 P oduc Line Requi emen s based on Goals, Fea u es and Use cases……………………………………………………………………..4 B uno González-Baixauli, Miguel A. Laguna, Yania C espo. A Me a-model o Requi emen s Enginee ing in Sys em Family Con ex …………………………………………………………………..8 Rod igo Ce ón, Jose L. A ciniegas, Jose L. Ruiz, Juan C. Dueñas. Va iabili y Desc ip ion in Requi emen s o P oduc Family Suppo ……………………………………………………………………….13 Ra ael Capilla. The Use o Sa is ac ion A gumen s o T aceabili y in Requi emen s Reuse o Sys em Families: Posi ion Pape ………………….18 Ka ina A wood, Tim Kelly, John McDe mid. T aceabili y o P oduc Family Sys ems An XQue y App oach…………………………………………………………………………….22 Wa apo n Ji apan hong, And ea Zisman. SPL Needs an Au oma ic Holis ic Model o So wa e Reasoning wi h ea u e models………………………………………………...27 Da id Bena ides, An onio Ruiz-Co és, Ra ael Co chuelo, and Oc a io Ma ín-Díaz. Tools Requi emen s o Requi emen s Enginee ing in Sys em Family Con ex …………………………………………………………...33 Rod igo Ce ón, F ancisco Valse a, Jose L. A ciniegas, Jose L. Ruiz, Juan C. Dueñas. Expe ience-based App oach o Requi emen s Reuse in P oduc Families wi h DOORS………………………………………………38 An onio Monzón, Juan C. Dueñas. Requi emen s Re-Use in he IT Business P ocess WEB Implemen a ion P oduc Family………………………………………………..44 Law ence E. Day. In e na ional Wo kshop on Requi emen s Reuse in Sys em Family Enginee ing h p://www.di .upm.es/iw eq am July 5, 2004, Ca los III Uni e si y o Mad id, Mad id, Spain Co-loca ed wi h In e na ional Con e ence on So wa e Reuse 8 Goals Requi emen s enginee ing can be nowadays conside ed as a ma u e b anch o so wa e enginee ing ha deals wi h he e icien elici a ion, desc ip ion and managemen o sys ems and so wa e equi emen s. Ano he in e es ing app oach o he de elopmen o complex so wa e sys ems is he sys em amily enginee ing (also called p oduc line enginee ing), ha p oposes echniques and me hods o mass p oduc ion o sys ems, ocusing on he euse aspec s o de elopmen and guiding he c ea ion o p oduc s (applica ion enginee ing) om a la ge so wa e base (domain enginee ing). The join o bo h disciplines leads o he concep o equi emen s enginee ing o sys em amilies; he i s wo kshop speci ically de o ed o he issue was o ganised in 2002 in conjunc ion wi h he In e na ional Requi emen s Enginee ing Con e ence. This wo kshop is he second o ha ack. Family enginee ing adds some axis o complexi y o he adi ional equi emen s enginee ing, especially in he elici a ion o equi emen s, classi ica ion and analysis, which a e he mos ad anced ields o wo k. The managemen o la ge se s o equi emen s, hei in e ac ion wi h s akeholde s in he amily enginee ing con ex , he aceabili y o o he phases in de elopmen appea as p oblems o which esea ch is ye open, especially as ega ds he p o ision o p ac ical me hods and ool suppo . This wo kshop b ings oge he so wa e p oduc line esea che s, equi emen s enginee ing p ac i ione s, ool builde s, indus y and academia, in o de o exchange ideas and expe iences, explo es he s a e o he p ac ice, discuss cu en and eme ging p ac ices and in oduce new concep s in he a ea o equi emen s enginee ing in he con ex o sys em amilies, assis ed wi h ools. A summa y o he applicabili y o ool suppo in he equi emen s enginee ing o p oduc amilies will be ob ained. A lis o echniques applied in p ac ice is also sough . Topics o in e es include -among o he s-: * ools o equi emen s enginee ing in sys em amilies, * a iabili y desc ip ion in equi emen s, * echniques o de i a ion o applica ion equi emen s om domain equi emen s, * decision models in equi emen s enginee ing, * eco e y o equi emen s in sys em amilies, * aceabili y o equi emen s in sys em amily enginee ing. The wo kshop is mainly discussion-o ien ed; b ie p esen a ions by pa icipan s (20 minu es) a e sough ha os e discussions. Second, hi d and ou h sessions will be de o ed o p esen a ions o pa icipan s and discussions and u he discussion on he issues p esen ed. O ganize s * Juan C. Dueñas. Uni e sidad Poli écnica de Mad id, Spain. [email p o ec ed] * Klaus Schmid. F aunho e IESE, Ge many. [email p o ec ed] P og am commi ee Jan Bosch, Uni e si y o G oningen. The Ne he lands. [email p o ec ed] Juan C. Dueñas. Uni e sidad Poli écnica de Mad id, Spain. [email p o ec ed] Julio Cesa Lei e, PUC-RIO, B azil. [email p o ec ed] F ank an de Linden, Philips, The Ne he lands. [email p o ec ed] Bi gi Geppe , A aya, USA. [email p o ec ed] Klaus Schmid. F aunho e IESE, Ge many. [email p o ec ed] P og amme 10:00 In oduc ion and p esen a ion 10:15 Session 1 Requi emen s euse S a e o he a and p ac ice in equi emen s euse in sys em amilies. Juan C. Dueñas. 11:00 Discussion 1 11:30 Session 2 Requi emen s modelling P oduc Line Requi emen s based on Goals, Fea u es and Use cases. B uno González- Baixauli, Miguel A. Laguna, Yania C espo. A Me a-model o Requi emen s Enginee ing in Sys em Family Con ex . Rod igo Ce ón, Jose L. A ciniegas, Jose L. Ruiz, Juan C. Dueñas. Va iabili y Desc ip ion in Requi emen s o P oduc Family Suppo . Ra ael Capilla. 12:30 Discussion 2 13:00 Lunch b eak 14:00 Session 3 T aceabili y The Use o Sa is ac ion A gumen s o T aceabili y in Requi emen s Reuse o Sys em Families: Posi ion Pape . Ka ina A wood, Tim Kelly, John McDe mid. T aceabili y o P oduc Family Sys ems An XQue y App oach. Wa apo n Ji apan hong & And ea Zisman. SPL Needs an Au oma ic Holis ic Model o So wa e Reasoning wi h ea u e models. Da id Bena ides, An onio Ruiz-Co és, Ra ael Co chuelo, and Oc a io Ma ín-Díaz. 15:00 Discussion 3 15:30 B eak 16:00 Session 4 Tool suppo Tools Requi emen s o Requi emen s Enginee ing in Sys em Family Con ex . Rod igo Ce ón, F ancisco Valse a, Jose L. A ciniegas, Jose L. Ruiz, Juan C. Dueñas. Expe ience-based App oach o Requi emen s Reuse in P oduc Families wi h DOORS. An onio Monzón, Juan C. Dueñas. Requi emen s Re-Use in he IT Business P ocess WEB Implemen a ion P oduc Family. Law ence E. Day. 16:30 Discussion 4 16:45 Summa y and end o mee ing P oduc Line Requi emen s based on Goals, Fea u es and Use cases B uno González-Baixauli, Miguel A. Laguna, Yania C espo Depa men o Compu e Science, Uni e si y o Valladolid, Campus M. Delibes, 47011 Valladolid, Spain {bbaixauli, mlaguna, yania}@in o .u a.es Abs ac . T adi ional PL equi emen s app oaches p esen se e al p oblems in equi emen s analysis, mainly in a ian s analysis and selec ion. The eason is undamen ally he di icul o deal wi h non- unc ional equi emen s. These p oblems can be sol ed wi h he in oduc ion o he goal/so goal pa adigm. This pa adigm in oduces in en ionali y ( he “whys”) and allows ela ing unc ional and non- unc ional equi emen s, he basis o he a ian analysis. This p oposal imp o es he PL equi emen s in oducing goal/so goal pa adigm and ela ing i wi h well-known echniques as ea u e modeling, and use case ex end mechanism. 1 In oduc ion So wa e euse has been a e y p omising discipline o many yea s, bu he esul s ha e no been so good as expec ed. Recen ly, P oduc lines (PL) appea as he mo e success ul app oach in he euse ield, as hey combine coa se-g ained componen s, i.e. so wa e a chi ec u es and so wa e componen s, wi h a op-down sys ema ic app oach, whe e he so wa e componen s a e hough a p io i and in eg a ed in a high-le el s uc u e [6]. Conce ning equisi es, he mo e used echniques a e ela ed wi h ea u es. These echniques a e ocused in commonali y and a iabili y o ind he main componen s (common o he en i e PL). Howe e , we hink hese echniques a e oo much design o ien ed, and complex o apply wi hou a mo e equi emen s o ien ed echnique. Recen ly, equi emen s enginee ing (RE) has aken g owing impo ance inside so wa e enginee ing. One o i s mos impo an app oach is he goal-o ien ed RE. I p oposes o model explici ly he in en ionali y o he sys em ( he “whys”). The in en ionali y has been widely ecognized as an impo an poin o he sys em, bu usually i is no modeled. The main ad an ages o he goal-o ien ed app oach a e ha can be used o s udy al e na i es in so wa e equi emen s (i uses AND/OR models ha models he di e en al e na i es) and ha can ela e unc ional and non- unc ional equi emen s in e o lessly. Then , he e a e wo impo an cha ac e is ics o goals ha can be use ul o he PL app oach: i s , hey exp ess he in en ionali y o he sys em, he e o e hey gi e a e y na u al way o ake decisions. We can ake decisions om “wha we wan ”, e sus “wha he sys em do”. Second, hey can model al e na i es in sys em equi emen s, wha can be easily mapped o a ian s in PL. Howe e , his concep mus be linked o PL, mainly wi h PL a chi ec u es. An easy way o do his is o ela e goals wi h adi ional ea u es. He e, he e is a lo o wo k done ela ing goals and use-cases/scena ios and use-cases/scena ios wi h ea u es. Consequen ly, use-cases/scena ios can be a good join poin . In his con ex , we p opose an app oach o PL equi emen s whe e he concep o goal as a guide o he selec ion o a ian s. 2 Using Goals, Fea u es and Use Cases o Requi emen Va iabili y PL Requi emen s de ine he p oduc s de elopable in he PL and hei ea u es. The e a e wo speci ic cha ac e is ics di e en om adi ional equi emen s: he main equi emen s o e e y PL p oduc should be de e mined, e en he equi emen s o he p oduc s no de eloped (mus be o ecas ed); and i is undamen al o know hei commonali y and a iabili y, and he dependencies be ween hem. To ep esen his kind o in o ma ion, he equi emen s a e usually s uc u ed in de ini ion hie a chies [7] o ea u e models as in FORM (Fea u e-O ien ed Reuse Me hod) [4]. Thus, each PL equi emen is a p ominen and dis inc i e concep o cha ac e is ic ha is isible o a ious s akeholde s, o ea u e. These ea u es a e o ganized by a g aphical AND/OR hie a chy diag am, and se i manda o y, op ional, o al e na i e. The main p oblems wi h his echnique a e ha a la ge domain expe ience is needed, and i is also mo e o ien ed o he a chi ec u e de ini ion han o clien p esen a ion o equi emen s de ini ion. Requi emen elici a ion can also be based on use case analysis (usually a mo e amilia echnique). The basis o he use case suppo o he a iabili y is he ex end mechanism [3]. Al hough his echnique is mo e o ien ed o equi emen s, he ex end mechanism is no enough o ep esen complex a iabili y. The common cha ac e is ic o hese wo echniques is hei solu ion-space o ien a ion. I is al eady ecognized ha an in en ional iewpoin ha esponses he “whys” is also necessa y. In o de o add ess his issue, la ge wo k has been done in he ield o goal-o ien ed RE. A goal is an objec i e he sys em unde conside a ion should achie e [8]. The e a e wo ypes o goals: (ha d) goals and so goals: goals sa is ac ion can be es ablished h ough e i ica ion echniques [9], bu sa is ac ion o so goals canno be es ablished in a clea -cu sense (usually used o model non- unc ional cha ac e is ics o he sys em) [8]. The dependences be ween goals and so goals can be es ablished. The NFR amewo k de ines hese co ela ions [1]. I is necessa y o ga he hese h ee iewpoin s: in en ional (goals), ope a ional (use cases) and unc ional ( ea u es). To achie e his, he use case and scena ios a e he na u al common poin . The e a e al eady se e al wo ks ha ela e goals wi h use cases o scena ios in a single sys em en i onmen [5] and use cases wi h ea u es in PL app oaches [2]. The idea is o use he i s echniques whe e goals help o cons uc use cases, and use cases assis in goals disco e y. Then, wi h he use cases, ind ea u es in an easy way wi h he la e echniques. The ea u es a e use ul because hey a e close o he a chi ec u e de ini ion. T aceabili y links goals wi h use cases and use case wi h ea u es. Subsequen ly, he goals (in en ionali y) a e ela ed o ea u es, and hese o PL a chi ec u e and asse s. In addi ion, a di ec ela ionship be ween goals and ea u es can be ound, he goal-o ien ed ask concep ( he means o a aining he goals) can be easily mapped o he mo e implemen a ion ype ea u es. Following his easoning, we a e sepa a ing he ea u es in wo concep s: he goals ha model he capabili ies ea u es (gene al unc ionali y and ope a ions and non- unc ional equi emen s), and he asks, ha model all he o he ea u es ypes (ope a ing en i onmen , domain echnology and implemen a ion echnique). In Fig. 1 he ela ions be ween he di e en PL equi emen enginee ing models a e ou lined. AND OR OR Quali y Analysis Quali y Analysis Implemen a ion Analysis Ope a ionalza ion Analysis Quali y Focus Func ionali y Analysis goals so goals ea u es AND OR OR Quali y Analysis Quali y Analysis Implemen a ion Analysis Ope a ionalza ion Analysis Quali y Focus Func ionali y Analysis AND OR OR AND OR OR Quali y Analysis Quali y Analysis Implemen a ion Analysis Ope a ionalza ion Analysis Quali y Focus Func ionali y Analysis goals so goals ea u es Fig. 1. Rela ionships be ween models. Goals a e ope a ionalized in o use cases, whose a ian s a e ocused o di e en quali y ac o s. Fea u es implemen he use cases ac ions wi h di e en con ibu ions o so goals. Also di ec ela ionships (wi hou use cases) a e possible. This app oach also helps in he PL a ian s selec ion p oblem. In gene al, only he unc ional o ope a ional poin o iew guides he selec ion o a ian s. We hink ha a mo e igo ous mechanism o selec ion is equi ed. In his sense, goal app oach allows o ela e unc ional equi emen s wi h non- unc ional ones and o conduc o mal analysis [1]. The e o e, goal analysis echniques can be used, and he selec ion is done om a mo e na u al iewpoin (in en ionali y). The idea is ha , gi en a se o equi emen s a ian s, encoded in ou model, his echnique allows o selec he be e solu ion (se o ea u es and ela ed asse s) acco ding o a basic unc ional selec ion (ha dgoals) and a gi en quali y c i e ia (so goals). He e, ha dgoals limi he a iabili y space and so goals gi e he c i e ia o ge he bes a ian om he limi ed a iabili y space. Nowadays, we a e ocused in he modeling aspec o he app oach. We a e ex ending UML o allow modeling goals and ea u es wi h wo new models. In addi ion, i is impo an he ela ionships be ween models: he goals co e use cases, he use cases a e desc ibed by ea u es, and ea u es ope a ionalize goals. 3 Conclusions and u u e wo k The main con ibu ion o his app oach is he de ini ion o a gene al model o equi emen a iabili y ha inco po a es goals as guide o he a ian s selec ion. This app oach makes easie and mo e comple e he PL equi emen analysis in oducing an in en ional iewpoin and ela ing h ee well-known echniques. In addi ion, he a iabili y analysis is imp o ed by means o conside ing non- unc ional equi emen s wi h goal analysis. Consequen ly, he p oduc a chi ec s ha e a a ionale o he selec ion o hose cha ac e is ics ha had be e suppo he equi emen s ( unc ional and non- unc ional) o a new p oduc line membe . We a e wo king on a ool (adap a ion om a p e ious one) ha allow he co ela ion o goals, use cases and ea u es models. This ool will ocus on a ian selec ion om he goals, selec ing he desi ed unc ionali y wi h he (ha d) goals and he p e e ed quali y p ope ies by p io i izing so goals. In his way, he ea u es a e hidden, bu hei ela ionships and cons ain s a e used o ge he a ian s. These ela ionships and cons ain s will be also used o ind ela ionships on goals, compa ing he possible a ian s om goal selec ions. The e o e, i is possible o know i one goal equi es o he ( he goal is only achie ed when he o he ), wo goals a e mu ual-exclusi e ( he e is no a ian s ha achie e bo h goals), one hin s o he o he o hey mu ually hinde s. Re e ences 1. Chung, L., Nixon, B., Yu, E. and Mylopoulos, J. Non-Func ional Requi emen s in So wa e Enginee ing. Kluwe Academic Publishe s 2000. 2. G iss, M., Fa a o, J., and d' Alessand o, M. In eg a ing ea u e modeling wi h he RSEB. In P oceedings o he Fi h In e na ional Con e ence on So wa e Reuse, pages 76--85. IEEE Compu e Socie y P ess, 1998. 3. Jacobson I., G iss M., and Jonsson P.: So wa e Reuse. A chi ec u e, P ocess and O ganiza ion o Business Success. ACM P ess. Addison Wesley Longman (1997) 4. Kang, K. C., Kim, S., Lee, J., and Kim, K.: FORM: A Fea u e-O ien ed Reuse Me hod wi h Domain-Speci ic Re e ence A chi ec u es. Annals o So wa e Enginee ing, 5:143- 168 (1998) 5. Ka akli, E., Loucopoulos, P., and Filippidou, D. Using Scena ios o Sys ema ically Suppo Goal-Di ec ed Elabo a ion o In o ma ion Sys em Requi emen s. In P oceedings o IEEE Symposium and Wo kshop on ECBS'96, Ge many:, 1996. pp. 308-314 6. Knaube , P., and Succi, G.: Pe spec i es on So wa e P oduc Lines. ACM So wa e Enginee ing No es, 26(2):29-33 (2001) 7. Kuusela, J., and Sa olainen, J.: Requi emen s Enginee ing o P oduc Families. In P oceedings o 22nd In e na ional Con e ence on So wa e Enginee ing – ICSE 2000. Pages 60-68. ACM P ess (2000) 8. Mylopoulos, J., Chung, L., Yu, E. and Nixon, B.: Rep esen ing and Using Non- unc ional Requi emen s: A P ocess-O ien ed App oach, IEEE T ans. on So wa e Eng, 18(6), June 1992 pp:483-497. 9. an Lamswee de, A. "Goal-O ien ed Requi emen s Enginee ing: A Guided Tou ", P oceedings o he 5 IEEE In . Symp. on Requi emen s Enginee ing, 2001, pp:249-262 Requi emen s Speci ica ions (S d. 830-1998) [6]. One o he mo i a ions o his wo k comes om he expe ience gained in a mid- e m p ojec o de eloping web sys ems o which a ligh weigh p oduc line model was se -up o building he di e en web p oduc s. Each yea wi h he enewal o he p ojec , pa o hese equi emen s change o e ol e while o he s emain s able. In his way, a sui able ep esen a ion o he equi emen s and he a iabili y becomes a need in he p ojec . 2 Modeling Va iabili y in So wa e Requi emen s Fo ep esen ing he equi emen s in he p ojec we used in o mal no a ions bu close o he s anda ds on equi emen s enginee ing. When he ligh weigh p oduc was se - up o he i s yea , he so wa e a chi ec u e was de i ed om he equi emen s and he ini ial web p oduc s we e enginee ed om he co e componen s de eloped unde he p oduc line. As he p ojec e ol es o e ime, pa o he equi emen s change and new ones appea o building new so wa e p oduc s. Thus, he need o ep esen ing and managing his a iabili y [1] [4] becomes no only a design p oblem bu also a challenge o he equi emen s speci ica ion. Ou posi ion in his wo k is o imp o e his si ua ion by ep esen ing a iable issues [4] [7] [8] [9] in he equi emen s desc ip ion, no only in he design phase. To do his we e alua ed he empla es p oposed in he s anda d IEEE 830-1998 as a ecommended p ac ice o equi emen s speci ica ion. F om he eigh empla es p oposed o desc ibing speci ic equi emen s, we hough he o ganiza ion by ea u es (shown in igu e 1) i s be e han o he empla es because many p oposals o ep esen ing he a iabili y o sys ems can be done using a ea u ed model, such as FODA [5]. Fig. 1. Requi emen s o ganized by ea u es based on he s anda d IEEE S d. 830-1998. 1. Specii c equi emen s 1.1 Ex e nal in e ace equi emen s 1.1.1 Use in e ace 1.1.2 Ha dwa e in e ace 1.1.3 So wa e in e ace 1.1.4 Communica ions in e ace 1.2 Sys em Fea u es 1.2.1 Sys em Fea u e 1 1.2.1.1 Pu pose o ea u e 1.2.1.2 S imulus / Response sequence 1.2.1.3 Associa ed unc ional equi emen s 1.2.1.3.1 Func ional equi emen 1 1.3 Pe o mance equi emen s 1.4 Design cons ain s 1.5 So wa e sys em a ibu es 1.6 O he equi emen s A e he equi emen s analysis phases was pe o med, we ound wo main limi a ions no well ep esen ed in he equi emen s empla e selec ed. The i s one e e s o he sepa a ion be ween common equi emen s o he en i e p oduc amily and speci ic equi emen s o pa icula p oduc s. O cou se, his can be sol ed desc ibing he common equi emen s and speci ic p oduc equi emen s in di e en documen s. The second p oblem e e s o he abili y o he empla e o include mo e in o ma ion in he equi emen s ela ed o he u u e a ia ion poin s. In ac , many so wa e sys ems include quan i a i e alues o ange o alues ha a e associa ed o unc ional ope a ional modes. This aspec can be modeled as speci ic a ia ion poin s wi h hei own a ian s. Such as is men ion in [2], ea u ed models don’ ep esen his explici ly bu he same wo k desc ibes an app oach o ex end hese models using a simple no a ion. In o de ep esen he same in o ma ion a he equi emen le el, we p opose in his wo k o ex end he equi emen s empla e shown in igu e 1 o include he ollowing in o ma ion: 1. Common and p oduc speci ic equi emen s: he equi emen s empla e has wo main pa s. One o ep esen ing he common equi emen s o he domain o he en i e p oduc amily and one subsec ion o each p oduc s o p oduc s who ha e pa icula equi emen s. 2. Va ia ion poin s: This ield indica es i he sys em ea u e would ep esen a a ia ion poin . The alues allowed a e: ue o alse. 3. Type o alue ield: o each sys em ea u e o a common o pa icula p oduc equi emen , his ield includes he ype o alue allowed, such as o ins ance: nume ic, s ing, e c. 4. Range o alue ield / Lis o alues: o ce ain ype o alues, anges o lis s o alues can be ep esen ed in his ield. Fo ins ance, he da abase sys ems allowed o connec ing a web sys em wi h da a a e: ms-access, My-SQL and o acle. 5. Max / Min alue ield: he maximum and minimum alues allowed o a speci ic a ian . 6. Repe i ion alue ield: in ce ain cases is necessa y o ep esen he numbe o imes a elemen can occu . 7. Dependencies om common o single pa s: links o es ablishing dependencies o hie a chies be ween amily and p oduc speci ic equi emen s o acing pu poses. 8. G aphical no a ion o dis inguish p oduc amily and speci ic p oduc equi emen s: we p opose a simple g aphical no a ion ha makes easy he dis inc ion o equi emen s o he en i e p oduc line, speci ic p oduc amilies and conc e e p oduc s. The name o a speci ic p oduc amily o pa icula p oduc can be a ached o he no a ion o in he equi emen s lis . This no a ion is shown in able 1. SYMBOL USED TYPE OF PFE REQUIREMENT A equi emen o he en i e p oduc line. A equi emen o a speci ic p oduc amily (PF) A equi emen o a pa icula p oduc in he PF. The in o ma ion p e iously desc ibe can added o equi emen s empla e o achie e a mo e accu a e desc ip ion o he equi emen s o PF as well as o ep esen ing he a iabili y such equi emen s. This p oposals ies o imp o e he dis inc ion be ween common and pa icula p oduc equi emen s in he PF and acili a e he iden i ica ion o he a ia ion poin s in he design phase. Also, aceabili y issues can be a o ed because he ansi ion om he analysis o he design phase esul s easy. 3 T ansi ion om Requi emen s o A chi ec u es Based on he ex ended empla e o desc ibing he equi emen s in p oduc amilies, we p opose a se o simple s eps o pe o m he ansi ion om he equi emen analysis phase o he design phase. These s eps a e he ollowing: 1. Iden i y and ex ac he equi emen s o he p oduc s we wan o build. 2. Iden i y and sepa a e he equi emen s common o he p oduc amily. 3. Fo each p oduc o g oup o simila p oduc s iden i y speci ic p oduc equi emen s a ill a new subsec ion in he equi emen s empla e adding he no a ion o able 1. 4. Fo p oduc o g oup o p oduc s, iden i y which o he equi emen s associa ed o sys em ea u es would be conside ed as ini ial a ia ion poin s. Iden i y ypes o alues, anges and o he ex a in o ma ion sui able o ha e a ia ions in he u u e sys em and ill he app op ia e subsec ions in he empla e. 5. Es ablish dependencies in he equi emen s lis om he p oduc amily o p oduc s. 6. Check inconsis encies among he alues and links al eady speci ied. In addi ion o his, in he Web sys ems domain some o he a ia ion poin s and a ian s a ec o he use p esen a ion laye in he a chi ec u e, which has impac in non unc ional equi emen s such as he usabili y o he sys em. Fo ins ance, he de ini ion o on ypes, size, colo o he alloca ion o he in o ma ion and menu op ions i a Web sys em may a ec he usabili y expe imen ed by he use . In his way, he Web Accessibili y Ini ia i e (WAI) [10] ou lines design p inciples o c ea ing web accessible Web con en s. These ea u es ha ha e impac in non unc ional equi emen s can be also modeled as so wa e sys em a ibu es as shown in igu e 1 and ela ed o speci ic a ia ions such as is men ioned in sec ion 2. The e o e, some Table 1. No a ion o desc ibing ypes o PFE equi emen s a ia ion poin s o he sys em could be associa ed o non unc ional equi emen s ha a ec o he o e all design o he sys em. 4 Conclusions The aim o he posi ion desc ibed in his pape encou ages he inclusion o a iabili y in o ma ion in he equi emen s speci ica ion p ocess. The limi a ions ounded in he de elopmen o web p oduc s unde a p oduc line model ha e shown he lack o a iabili y in o ma ion in he analysis phase. This app oach esul s use ul o se e al easons. Fi s , o ace he esul s om he analysis o he design phase (i.e.: o wa d app oach) and om he design o he analysis phase (i.e.: backwa d app oach). Second, because is no di icul o a CASE ool o implemen he in o ma ion desc ibed in sec ion 2. Thi d, because i ela es unc ional and non unc ional equi emen s connec ed h ough he a ia ion poin s de ined in he a chi ec u e. Fou h, i dis inguishes common equi emen s om speci ic p oduc equi emen s. Finally, all hese easons acili a e he main enance, euse and e olu ion o he equi emen s in p oduc amily app oaches and in pa icula o Web p oduc s bu also opened o o he domains o so wa e sys ems. Re e ences 1. Bosch, J., Design & Use o So wa e A chi ec u es, Addison -Wesley (2000) 2. Capilla, R., Dueñas, J. C.: Modelling Va iabili y wi h Fea u es in Dis ibu ed A chi ec u es. 4 h In e na ional Wo kshop on So wa e P oduc -Family Enginee ing. Lec u e No es in Compu e Science, Vol. 2290. Sp inge -Ve lag, Be lin Heidelbe g New Yo k (2002) 319–329 3. Clemen s, P., No h op, L. So wa e P oduc Lines. P ac ices and Pa e ns. Addison-Wesley (2002) 4. Dob ica, L., Niemelä, E.: Using UML No a ion o Model Va iabili y in P oduc Line A chi ec u es, In e na ional Wo kshop on So wa e Va iabili y Managemen , ICSE’03, Po land, O egon, USA, (2003) 8-13 5. Kang K. C., Cohen S., Hess J. A., No ak W. E., Pe e son A. S. Fea u ed-O ien ed Domain Analysis (FODA) Feasibili y S udy. Technical Repo , CMU/SEI-90-TR-21 ESD-90-TR- 22, So wa e Enginee ing Ins i u e, Ca negie Mellon Uni e si y, Pi sbu gh (1990) 6. IEEE Recommended P ac ice o So wa e Requi emen s Speci ica ion, S d. 830-1998 7. Jacobson, I., G iss, M., Johnsson, P.: So wa e Reuse. A chi ec u e, P ocess and O ganiza ion o Business Success, ACM P ess (1997) 8. T. Myllymäki, “Va iabili y Managemen in So wa e P oduc Lines”, Tampe e Uni e si y o Technology, So wa e Sys ems Labo a o y, ARCHIMEDES, h p://p ac ise.cs. u . i/pub/pape s/Va MgnFinal.pd , 2001 9. Robak, S.: Fea u e Modeling No a ions o Sys em Families, In e na ional Wo kshop on So wa e Va iabili y Managemen , ICSE’03, Po land, O egon, USA, (2003) 58-62 10. Web Accessibili y Ini ia i e (WAI). h p://www.w3.o g/WAI The Use o Sa is ac ion A gumen s o T aceabili y in Requi emen s Reuse o Sys em Families: Posi ion Pape Ka ina A wood, Tim Kelly, John McDe mid Rolls-Royce Uni e si y Technology Cen e in Sys ems and So wa e Enginee ing, Depa men o Compu e Science, Uni e si y o Yo k, Hesling on, YORK. YO10 5DD Uni ed Kingdom {ka ina.a wood, im.kelly, john.mcde mid}@cs.yo k.ac.uk 1 In oduc ion Re inemen o equi emen s in o speci ica ions depends on he concep o equi emen s ‘sa is ac ion’. This is a ecu si e p ocess, in which he ul illmen o lowe -le el equi emen s is seen as a su icien condi ion o he ul illmen o he highe -le el s a emen [1]. T aceabili y s uc u es eco d sa is ac ion ela ionships be ween equi emen s a a ious le els o abs ac ion. I is essen ial o he success ul euse o equi emen s ac oss sys em amilies ha hey con ain su icien in o ma ion abou he ela ionships be ween and con ex o equi emen s om ea ly p oduc s in he amily o assu e de elope s ha he equi emen s a e alid o subsequen p ojec s. Failu e o obse e his p inciple isks he la e, and he e o e cos ly, disco e y o e oneous o unimplemen able equi emen s. In his pape , we o e a c i ique o s anda d aceabili y echniques and p opose a me hod o de eloping aceabili y s uc u es o equi emen s euse. 2 Sa is ac ion A gumen s and ‘Rich T aceabili y’ Za e and Jackson [2] obse e ha sa is ac ion o a equi emen (R) can be demons a ed only by a su icien combina ion o domain knowledge (K) and speci ica ions (S): S, K ├ R. Jackson sugges s ha aceabili y links be ween equi emen s and speci ica ions should be suppo ed by ex ual ‘co ec ness a gumen s’ which explain how he speci ica ions and domain beha iou combine o p o ide assu ance ha he enginee ed sys em sa is ies he equi emen in he applica ion domain [3]. Jackson’s ‘co ec ness’ a gumen s ha e been in eg a ed in o se e al indus ial-s eng h p ocesses o equi emen s enginee ing [ o example, 4]. The ‘Rich T aceabili y’ echnique [5] ep esen s ‘sa is ac ion a gumen s’ using goal-s uc u es cha ing AND/OR decomposi ions, he jus i ica ions o which a e eco ded. Fig. 1 documen s he ‘ ich aceabili y’ ela ionship be ween a op-le el ope a ional equi emen o a ehicle and lowe -le el echnical equi emen s. The ‘sa is ac ion a gumen ’ in oduces domain knowledge and indica es ha a conjunc ion o he h ee speci ica ions sa is ies he op-le el equi emen . Fig. 1. A ‘ ich aceabili y’ s uc u e ( om [5]) 3 Loca ing he Sa is ac ion A gumen The inclusion o a ‘sa is ac ion a gumen ’ in he ‘ ich aceabili y’ s uc u e (Fig.1) p o ides an explici eco d o he s a egy used o decompose equi emen s s a emen s and pe mi s acking o ela ionships and cons ain s be ween equi emen s s a emen s a di e en le els o abs ac ion. In his espec , ‘ ich aceabili y’ o e s conside able ad an ages o e simple aceabili y s uc u es, which simply pos ula e links be ween equi emen s a i ac s, wi h no a emp o jus i y o explain he connec ions be ween hem. We a e conce ned, howe e , ha ‘sa is ac ion a gumen s’ and aceabili y s uc u es o his ype do no ac ually demons a e he sa is ac ion o he op-le el equi emen by he lowe -le el s a emen s. Fig. 1 pu s o wa d no a gumen o con ince he eade ha he e inemen ac ic adop ed will esul in a speci ica ion which sa is ies he equi emen . No e idence is p esen ed o suppo he implied claim ha powe , clea ance and weigh a e he only conce ns which need o be conside ed in assessing he ehicle’s capabili ies in he e ain. No is he e any a gumen o demons a e ha he echnical cons ain s speci ied in he sys em equi emen s a e ele an o he op-le el equi emen and su icien o sa is y i . Wha is o e ed he e is no a ‘sa is ac ion a gumen ’, bu a se ies o p oposi ions which need o be suppo ed by a gumen s o con ince he eade ha he design mee s he equi emen . The eal ‘sa is ac ion a gumen ’, hen, is no o be ound in he decomposi ion s a egy eco ded he e, bu in an unexp essed me a-a gumen s a ing he basis on which his s a egy is p oposed. 4 Sa is ac ion A gumen s and Sa e y Cases P e ious wo k a Yo k has ocused on he de elopmen o a g aphical no a ion me hod o he cons uc ion o sa e y case a gumen s, he Goal S uc u ing No a ion (GSN) [6]. Sa e y cases a e semi- o mal a gumen s which demons a e how a ailable e idence abou a sys em and i s con ex can be used o show ha a sys em is accep ably sa e o ope a e in i s con ex [7]. As applied in equi emen s enginee ing, GSN p o ides a means o eco ding aceabili y links be ween indi idual claims and sub-claims. These claims a e ep esen ed as goals, and equa e o he equi emen s s a emen s and speci ica ions in he ‘ ich aceabili y’ example (Fig. 1). The no a ion also eco ds he s a egies used o decompose he goals. These s a egies map o he ‘sa is ac ion a gumen ’ in Fig. 1, in ha hey seek o p o ide a basis o he ela ionship be ween he goals. As well as p esen ing a clea eco d o he goal-decomposi ion s a egy, howe e , GSN allows his s a egy o be alida ed by he use o an appa a us o jus i ica ions and assump ions a ached o goals and s a egies wi hin he s uc u e, as well as by explici e e ences o a i ac s such as sys em a chi ec u al models o con ex ual in o ma ion. This documen s he me a-a gumen cap u ing he sa is ac ion ela ionships which jus i y he decomposi ion o equi emen s. GSN exp esses jus i ica ions and assump ions as p oposi ions, and he e is no oppo uni y p o ided o hei u he de elopmen . A mo e sophis ica ed ea men o he me a-a gumen can be achie ed by he use o GSN ‘away goals’, by which e e ence is made o a pa allel goal-s uc u e in which he me a-a gumen (ou ‘ eal’ sa is ac ion a gumen ) is elabo a ed. 5 Me a-A gumen s and Requi emen s Reuse o Sys em Families Sa is ac ion a gumen s, exp essed as me a-a gumen s on a equi emen s decomposi ion, p o ide assu ance ha he aceabili y ela ionship be ween equi emen s a di e en le els o abs ac ion is alid wi hin a gi en applica ion domain. I equi emen s a e o be eused success ully be ween p ojec s, i is impo an ha sa is ac ion ela ionships emain alid o equi emen s and speci ica ions in he euse domain. The ollowing example demons a es how GSN can be used o documen sa is ac ion a gumen s, and o indica e whe e changes in design commi men s o con ex s challenge eused equi emen s. Ou example conce ns a sys em amily o ull-au ho i y digi al engine con olle s (FADECs) o ci il ai line s. A FADEC p o ides a compu e -con olled managemen sys em o he engine, which akes inpu s om he cockpi con ols and senso s loca ed on he ai c a and p oduces ou pu s in he o m o digi al signals used o con ol he engine [8]. Engines a e de eloped as a sys em amily, comp ising ‘ma ks’ and ‘ a ian s’. A ‘ma k’ is a speci ic engine in a se ies, such as he Rolls-Royce BR-710 and he BR-715, while a ‘ a ian ’ is a ma k p oduced o he speci ic equi emen s o an Ai ame . Requi emen s o a pa icula ‘ma k’ a e exp essed in e ms o a ‘common co e’, wi h a ia ions associa ed wi h speci ic ‘ a ian s’. Inad e en deploymen o e e se h us while an ai c a is in ligh can esul in se e e yawing and, a wo s , loss o con ol. The common co e equi emen s o a small-engine se ies (we’ll call i ‘Engine A’) ha e ou disc e e check sys ems, all o which mus be ac i ely disengaged be o e e e se h us is deployed: 1. An isola ion al e mus be opened, allowing he low o hyd aulic luid in o he h us e e se sys em. This al e is con olled by he FADEC, and i s de aul posi ion is ‘closed’. 2. The Th us Di ec ional Con ol Val e mus be swi ched o ‘ e e se’ om i s de aul ‘ o wa d’ posi ion. This al e is he join esponsibili y o he FADEC and he ai c a , al hough he ai c a (i.e. he pilo ) has ul ima e o e ide con ol. 3. The mechanical Te ia y Locks holding he h us e e se doo in ‘closed’ posi ion mus be opened. The ai c a has esponsibili y o hese locks. 4. Th o le in e locks p o ide he pilo wi h ac ile eedback conce ning he deg ee o e e se h us , and p e en him om eques ing mo e e e se h us han he equi es. Fig. 2 eco ds a simpli ied decomposi ion, in GSN, o he common co e equi emen s o he h us e e se deploymen p o ec ion sys em in ‘Engine A’. The decomposi ion s a egy p o iding he aceabili y links be ween he checks (R2 - R5) and he op-le el equi emen (R1) is documen ed in he homboid labelled S1. A ich appa a us o con ex ual assump ions and de ini ions is eco ded alongside he decomposi ion. Fig. 2. Decomposi ion o co e equi emen s o he h us e e se deploymen p o ec ion in he ‘Engine A’ amily The decomposi ion in Fig. 2 does no a emp o jus i y he decomposi ion s a egy employed. Ins ead, i con ains an ‘away goal’ e e ence o a jus i ica ion claimed s o ed elsewhe e (highligh ed in ed). This goal is he op-le el claim o he me a-a gumen in Fig. 3, which is he sa is ac ion a gumen jus i ying he equi emen s decomposi ion. The a gumen s a egy is a wo-p onged one: he le -hand side o he goal s uc u e a gues ha he checks, aken oge he , a e su icien o sa is y he op-le el equi emen , while he igh -hand side (no ully de eloped he e) a gues ha all possible ailu e modes ha e been conside ed and a e adequa ely mi iga ed by he checks. The GSN s uc u e makes clea wha e idence is equi ed o demons a e he sa is ac ion o he op-le el equi emen (PSSA). Fig. 3. Sa is ac ion a gumen o co e h us deploymen p o ec ion equi emen s in he ‘Engine A’ amily Fo his o ical and egula o y easons, Ai ame cus ome s o ce ain a ian s in he Engine A amily do no equi e he h o le in e locks ea u e on hei engines. Sa e euse o he co e equi emen s wi h his modi ica ion depends on R1 being sa is ied by he h ee emaining checks. The GSN s uc u e p o ides a s aigh o wa d means o assessing which aspec s o he sa is ac ion ela ionship a e h ea ened by he design change. I R5 ( h o le in e locks) is emo ed om he equi emen s decomposi ion, e e ence o he sa is ac ion a gumen demons a es ha he su iciency a gumen is comp omised. I will be necessa y o consul he PSSA o see whe he he combined ailu e a es o he isola ion al e, he di ec ional con ol al e and he e ia y locks can combine o sa is y he o e all sa e y equi emen o a ailu e a e no g ea e han 1 x 10-9 pe ligh hou . GSN sa is ac ion a gumen s hus allow o he clea eco d o domain in o ma ion and assump ions, and indica e which in o ma ion sou ces a e equi ed o adequa e equi emen s aceabili y. The se e i y o he impac o equi emen s o con ex ual change can be assessed by e e ence o me a- a gumen s on he equi emen s decomposi ion. This is o clea bene i in sys em amily de elopmen en i onmen s, whe e equi emen s euse depends on he assu ance o he alidi y o equi emen s decomposi ions in al e na i e de elopmen con ex s. Re e ences 1. an Lamswee de, A.: Goal-O ien ed Requi emen s Enginee ing: A Guided Tou . P oceedings o he 5 h IEEE In e na ional Symposium on Requi emen s Enginee ing. IEEE CS P ess, To on o (2001) 249-263 2. Za e, P., Jackson, M.: Fou Da k Co ne s o Requi emen s Enginee ing. ACM T ansac ions on So wa e Enginee ing and Me hodology, Vol. 6 No. 1 (1997) 1-30 3. Jackson, M.: P oblem F ames: Analysing and S uc u ing So wa e De elopmen P oblems. Addison-Wesley, London (2001) 4. Hall, A.: A Uni ied App oach o Sys ems and So wa e Requi emen s. P oceedings o he 5 h IEEE In e na ional Symposium on Requi emen s Enginee ing. IEEE CS P ess, To on o (2001) 267 5. Hull, E., Jackson, K., Dick, J.: Requi emen s Enginee ing. Sp inge -Ve lag, London (2002) 6. Wilson, S., Ki kham, P.: SAM Use Manual. Uni e si y o Yo k, Yo k (1995) 7. Kelly, T., McDe mid, J.: Sa e y Case Cons uc ion and Reuse Using Pa e ns. P oceedings o he 16 h In e na ional Con e ence on Compu e Sa e y, Reliabili y and Secu i y. Sp inge -Ve lag, New Yo k (1997) 55- 69 8. Lam, W.: Achie ing Requi emen s Reuse: a Domain-Speci ic App oach om A ionics. Jou nal o Sys ems and So wa e Vol. 38 (1997) 197-209 T aceabili y o P oduc Family Sys ems An XQue y App oach Wa apo n Ji apan hong & And ea Zisman Depa men o Compu ing Ci y Uni e si y No hamp on Squa e, EC1V 0HB, London, UK {w.ji apan hong | a.zisman}@soi.ci y.ac.uk Abs ac . T aceabili y has been p oposed as a mechanism o gua an ee quali y in he de elopmen li e-cycle o a so wa e sys em. In his pape we p esen an app oach o suppo au oma ic gene a ion o aceabili y ela ions be ween equi emen s a e ac s o p oduc amily so wa e sys ems. Ou app oach is ule-based in which he ules a e ep esen ed in XQue y and he a e ac s a e ansla ed in o XML. We p esen eigh di e en ypes o aceabili y ela ions be ween ea u e models, unc ional equi emen s speci ica ions, and class diag am. 1. In oduc ion Requi emen s T aceabili y (RT) has been ecognized as an impo an ac i i y in so wa e sys em de elopmen [9][14][16]. In gene al, aceabili y ela ions can imp o e he quali y o he p oduc being de eloped, and educe he ime and cos associa ed wi h he de elopmen . In pa icula , aceabili y ela ions can suppo e olu ion o so wa e sys ems, euse o pa s o he sys em by compa ing componen s o he new and exis ing sys ems, alida ion ha a sys em mee s i s equi emen s, unde s anding he a ionale o ce ain design and implemen a ion decisions in he sys em, and analysis o he implica ions o changes in he sys em. Au oma ic gene a ion and main enance o aceabili y ela ions is no an easy ask and suppo o aceabili y in so wa e enginee ing en i onmen s and ools is no always adequa e [16]. Some exis ing app oaches assume ha aceabili y ela ions should be es ablished manually [2][11][18], which is e o -p one, di icul , ime consuming, expensi e, complex, and limi ed on exp essi eness. The e o e, despi e i s impo ance, aceabili y is a ely es ablished. In o de o alle ia e his p oblem, mo e ecen ly, o he app oaches ha e been p oposed o suppo semi- o ully au oma ic gene a ion o aceabili y ela ions [1][6][7][13][15]. Howe e , in he majo i y o hese app oaches, he gene a ed aceabili y ela ions do no ha e s ong seman ic meanings necessa y o suppo he bene i s ha can be p o ided by aceabili y. In o de o o e come he di icul ies abo e, in ou p e ious wo k [17][20] we ha e p oposed a ule based app oach o allow au oma ic gene a ion o bi-di ec ional aceabili y ela ions be ween di e en ypes o equi emen s documen s (i.e. cus ome equi emen s speci ica ions, use-case speci ica ions, and analysis objec model). In his p e ious wo k we iden i y h ee di e en ypes o aceabili y ela ions, namely o e lap, equi e and ealize ela ions. In he app oach, aceabili y ules a e used o ma ch syn ac ically ela ed e ms in cus ome equi emen s and use-case speci ica ions wi h seman ically e ms in objec model. Based on his ma ching and using o he se o ules, he app oach also suppo s gene a ion o aceabili y ela ions be ween he cus ome equi emen s and use-cases. The app oach desc ibed in his pape is buil upon ou p e ious wo k. He e, we p opose an app oach o allow au oma ic gene a ion o aceabili y ela ions be ween documen s gene a ed du ing he de elopmen o p oduc amily sys ems, in o de o acili a e iden i ica ion o common and a iable unc ionali y in sys ems composing he amily, and euse o co e asse s ha a e a ailable unde he p oduc amily a chi ec u e. We a e in e es ed in documen s gene a ed in ea u e-based me hodologies due o he ac ha when de eloping p oduc amily sys ems cus ome s and sys em de elope s communica e wi h each o he in e ms o p oduc ea u es. We also belie e ha objec -o ien ed me hodology is impo an o suppo p oduc amily de elopmen . The e o e, we p opose o gene a e aceabili y ela ions be ween documen s p oduced when applying FORM [12] me hodology and some objec o ien ed documen s such as componen , class, and s a e cha diag ams. We decided o use FORM due o i s simplici y, ma u i y, p ac icali y, and ex ensibili y cha ac e is ics. In his pape we concen a e on aceabili y o unc ional equi emen s, ea u e models, and class diag ams. In ou app oach, he documen s a e ep esen ed in XML and he aceabili y ules in XQue y [19]. The a ionale o using XML a e due o se e al eason: (a) XML has become he de ac o language o suppo da a in e change among he e ogeneous sys ems, (b) he exis ence o applica ions ha use XML o ep esen in o ma ion in e nally o as a s anda d expo o ma , and (c) o allow he use o XQue y as a s anda d way o exp essing he aceabili y ules. The use o aceabili y o suppo de elopmen o p oduc amily sys ems is no a new hing. Some app oaches ha e been p oposed be o e [2][3][8][10]. Howe e , o he bes o ou knowledge mos o hese app oaches do no p o ide ways o gene a e aceabili y ela ions au oma ically. The emaining o his pape is s uc u ed as ollows. In sec ion 2, we desc ibe he main documen s used in ou wo k. In sec ion 3 we p esen an o e iew o ou app oach, he di e en ypes o aceabili y ules, and aceabili y ela ions. Finally, in sec ion 4 we summa ise ou app oach and discuss di ec ions o u u e wo k. 2. Types o documen s In his sec ion, we gi e an o e iew o he unc ional equi emen s speci ica ions, ea u e model, and class diag am used in ou app oach, and p esen ex ac s o hese documen s o a p oduc amily applica ion ela ed o mobile phones. Ou discussion assumes a amilia i y o he eade wi h he basic ea u es o XML which, due o space limi a ions, canno be gi en in his pape . (a) (b) Fig. 1. Examples o XML documen s: (a) unc ional equi emen speci ica ion o use case “Send da a”; (b) ea u e model o “Blue oo h” Func ional Requi emen s: Func ional equi emen s speci ica ions a e use-cases de ined acco ding o a empla e p oposed in [4]. These speci ica ions a e ep esen ed in XML, ollowing a DTD ha we ha e c ea ed. A use case con ains i le, s a us, egion, desc ip ion, le el, p econdi ions, pos condi ions, p ima y_ac o s, seconda y_ac o s, low_o _e en , excep ional_e en s, supe o dina e_use_case, and subo dina e_use_case. We p opose o ma k up he wo ds o he na u al language sen ences p esen in he use case desc ip ions by using XML elemen s ha indica e he g amma ical ole o he wo ds in he sen ence. This g amma ical ole is iden i ied by using he pa -o -speech agge called CLAWS [5]. Figu e 1 (a) p esen s an example o a use case desc ip ion o sending da a om a mobile phone, o a membe (MP1) o he p oduc amily. In his example, he wo ds ha appea as he con en o elemen s <Ti le> and <Desc ip ion> a e ma ked-up wi h he g amma ical oles. Fo ins ance, he wo d “Send” is agged wi h <VVB> deno ing ha i is a base o m o a lexical e b, while he wo d “da a” is agged < Fea u eModel > <Fea u e> <Fea u e_name> <NN1>Blue oo h</NN1> </Fea u e_name> <Desc ip ion>…<NN1>Blue oo h</NN1> <NN1>connec ion</NN1> <VM0>can</VM0> <VBI>be</VBI> <VVN>used</VVN> <TO0> o</TO0> <VVI>send</VVI> <NN0>da a</NN0> <AV0>i.e.</AV0> <NN2> ex s</NN2> <NN1>business</NN1> <NN2>ca ds</NN2> <NN1>calenda </NN1> <NN2>no es</NN2> <CJC>o </CJC> <TO0> o</TO0> <VVI>connec </VVI> <AV0>Wi elessly</AV0> <PRP> o</PRP> <NN2>compu e s</NN2>… </Desc ip ion> <Issue_and_decision/> <Type>Capabili y</Type> <Commonali y>Op ional</Commonali y> <Rela ion/> </Fea u e> </Fea u eModel> < Func ionalReqSpec Sys em=” MobilePhone ” P oduc _Membe =”MP1”> <Use_case UseCaseID=”1”> <Ti le> <VVB>Send</VVB> <NN0>da a</NN0> </Ti le> <S a us>Common</S a us> <Region Name = "All"/> <Desc ip ion>…<AJ0>mobile</AJ0> <NN1>phone</NN1> <VM0>can</VM0> <VVI>send</VVI> <NN0>da a</NN0> <VVN>kep </VVN> <PRP>in</PRP> <AT0> he</AT0> <NN1>phone</NN1> <PRP> o</PRP> <DT0>ano he </DT0> <NN1>phone</NN1> <CJC>o </CJC> <NN1>de ice</NN1> <PRP> ia</PRP> <NN1>communica ion</NN1> <NN2>channels</NN2> <AV0>i.e.</AV0> <NN1>Blue oo h</NN1>… </Desc ip ion> <Le el>P ima y Task</Le el> <P econdi ions>…</P econdi ions> <Pos condi ions>…</Pos condi ions> <P ima y_ac o >…</P ima y_ac o > <Seconda y_ac o s>…</Seconda y_ac o s> <Flow_o _e en s>…</Flow_o _e en s> <Excep ional_e en s>…</Excep ional_e en s> <Rela ed_In o ma ion> <Supe o dina e_use_case>…</Supe o dina e_use_case> <Subo dina e_use_case>…</Subo dina e_use_case> </Rela ed_In o ma ion></Use_case>…</Func ionalReqSpec> – High quali y ull p oduc . A p oduc wi h ull unc ionali y and high quali y: high a ailabili y and eliabili y and high cos oo. – Basic quali y ull p oduc . A p oduc wi h ull unc ionali y bu lowe quali y: lowe a ailabili y and eliabili y and lowe cos oo. Un il now, we ha e no ound any p oposal dealing wi h unc ional and ex a– unc ional ea u es in he same model. Howe e , he e a e some wo ks in he li e a- u e sugges ing he need o dealing wi h ex a– unc ional ea u es: Kang e . al ha e been sugges ing he need o ake in o accoun ex a– unc ional ea u es since 1990 [8, pag. 38] when hey depic ed a classi ica ion o ea u es bu hey did no p o ide a way o do i . La e in 1998 Kang e . al [9] made an explici e e ence o wha hey called ’non– unc ional’ ea u es (a possible ype o wha we call ex a– unc ional ea u es). Howe e he au ho s again did no gi e a way o do i . La e in 2001 Kang e . al [3], p oposed some guidelines o ea u e modelling, in [3, pag. 19], he au ho s made again dis inc ion be ween unc ional and quali y ea u es and poin ed ou he need o an spe- ci ic me hod o include ex a– unc ional ea u es, howe e hey did no p o ide such a speci ic way o do i ei he . In [16], he au ho s made dis inc ion be ween unc ional and wha hey called ”pa ame e s” (ano he possible ype o wha we call ex a– unc ional ea u es). The au ho s ma ginally in oduce a way o deal wi h his kind o ea u es. Ne e heless he e is a main d awback in his p oposal: he au ho s p opose o include he alues o he pa ame e s in he ea u e model, his is o say ha ing a pai pa ame- e / alue. This way i is no allowed o ha e anges o se s o alues o e en mo e, i is no possible o ha e a i me ical ela ions be ween a ibu es. In such a way, he possible ins ances o he SPL a e limi ed o he alues speci ied in he ea u e model. Al hough we ag ee ha in a SPL he e a e bo h unc ional and ex a– unc ional ea u es, in ou opinion, he e is no need o sepa a ing unc ional and ex a– unc ional ea u es due ha his sepa a ion may lead o all in manichaean discussions [1, pag.76]. Fo ins ance, an ADSL connec ion can ha e and a ibu e: he bandwid h, his a ibu e can be seen as a quali y ea u e he e o e as an ex a– unc ional ea u e. Ne e heless, depending on whe e he a ibu es akes i s alues i can also be seen as a unc ional ea u e. Thus, i he e a e only wo possible con igu a ions o ADSL connec ions, 128 o 256, e e y o his con igu a ions can be seen as an indi idual unc ional ea u e as illus a ed in igu e 2. We p opose o ha e an Holis ic SPL Model wi h he ollowing cha ac e is ics: – unc ional ea u es and he ela ions desc ibed p e iously: Manda o y, Op ional, Al e na i e and O ea u es. –ex a– unc ional ea u es ela ed o any unc ional ea u e. Howe e , we use he same o malism in he model so we do no make dis inc ion be ween unc ional and ex a– unc ional ea u es. 3.2 Reasoning on Holis ic Fea u e Models The numbe o po en ial single p oduc s in a SPL inc eases wi h he numbe o ea u es he SPL con ains. I ex a– unc ional ea u es a e also conside ed his numbe inc eases s ill mo e. In his con ex , easoning on SPL should be conside ed and some ques ions can be inqui ed o he holis ic model: In e ne Conec ion Powe Line ADSL128 ADSL256 Figu e2. Fea u e model wi h wo possible ADSL connec ions –Is is a alid model?: a alid model would be one on which a leas a single p oduc can be selec ed. –How many po en ial single p oduc s a SPL has?: answe ing his ques ion can gi e an ou look o he lexibili y o he model. G ea e he numbe o po en ial single p oduc s is, mo e lexibili y is suppo ed bu g ea e he complexi y o he SPL is oo. –Is he e any p oduc ul illing he use ’s equi emen s (bo h unc ional and ex a– unc ional)? F om he po en ial single p oduc s o he SPL, a use can equi e some ea u es so he space o he po en ial single p oduc s may be educed o his use . –F om he p oduc s ul illing he use ’s equi emen s which a e he p oduc s ha maximize o minimize an ex a– unc ional ea u e? e.g.: a use need a p oduc whe e he p ice o he numbe o lines o code a e minimized. Excluding a couple o limi ed p oposals [4,11], as a as we know, any o he p o- posal ha e been p esen ed o au oma ically eason on ea u e models un il now. Van Deu sen [4] and Mannion [11] p o ides a somehow o mal suppo o au oma ic ea- soning on ea u e models. Howe e , none o hem deals wi h ex a– unc ional ea u es, so hey a e ocussed on unc ional aspec s o SPLs. We p opose o ha e an holis ic model o malized using cons ain p og amming [12] using he backg ound we ha e [14,13] Re e ences 1. L. Bass, P. Clemen s, and R. Kazman. So wa e A chi ec u e in P ac ice. Addison–Wesley, 1998. 2. M. Be na do, P. Cianca ini, and L. Dona iello. A chi ec ing amilies o so wa e sys ems wi h p ocess algeb as. ACM T ansac ions on So wa e Enginee ing and Me hodology, 11(4):386– 426, 2002. 3. G. Chas ek, P. Donohoe, K.C. Kang, and S. Thiel. P oduc Line Analysis: A P ac ical In oduc ion. Technical Repo CMU/SEI-2001-TR-001, So wa e Enginee ing Ins i u e, Ca negie Mellon Uni e si y, June 2001. 4. A. an Deu sen and P. Klin . Domain–speci ic language design equi es ea u e desc ip ions. Jou nal o Compu ing and In o ma ion Technology, 10(1):1–17, 2002. 5. K. Cza necki U.W. Eisenecke . Gene a i e P og amming: Me hods, Techniques, and Appli- ca ions. Addison–Wesley, may 2000. ISBN 0–201–30977–7. 6. M. G iss, J. Fa a o, and M. d’Alessand o. In eg a ing ea u e modeling wi h he RSEB. In P oceedings o heFi hIn e na ional Con e ence on So wa e Reuse, pages 76–85, Canada, 1998. 7. S. Ja zabek, Wai Chun Ong, and Hongyu Zhang. Handling a ian equi emen s in domain modeling. The Jou nal o Sys ems and So wa e, 68(3):171–182, 2003. 8. K. Kang, S. Cohen, J. Hess, W. No ak, and S. Pe e son. Fea u e–O ien ed Domain Analysis (FODA) Feasibili y S udy. Technical Repo CMU/SEI-90-TR-21, So wa e Enginee ing Ins i u e, Ca negie Mellon Uni e si y, No embe 1990. 9. K.C. Kang, S. Kim, J. Lee, K. Kim, E. Shin, and M. Huh. FORM: A ea u e–o ien ed euse me hod wi h domain–speci ic e e ence a chi ec u es. Annals o So wa e Enginee ing, 5:143–168, 1998. 10. K.C. Kang, J. Lee, and P. Donohoe. Fea u e–O ien ed P oduc Line Enginee ing. IEEE So wa e, 19(4):58–65, July/Augus 2002. 11. Mike Mannion. Using Fi s -O de Logic o P oduc Line Model Valida ion. In P oceedings o he Second So wa e P oduc Line Con e ence (SPLC2), LNCS 2379, pages 176–187, San Diego, CA, 2002. Sp inge . 12. K. Ma io and P.J. S uckey. P og amming wi h Cons ain s: An In oduc ion. The MIT P ess, 1998. 13. O. Ma ´ın-D´ıaz, D. Bena ides, J. Pe na, and M. To o. Un a amien o sensible a la calidad pa a la adquisici´on de se icios web. In VIII Jo nadas de Ingenie ´ıa del So wa e y Bases de Da os JISBD’03, pages 209–220, Alican e, Espa˜na, 2003. 14. O. Ma ´ın-D´ıaz, A. Ruiz-Co ´es, A. Du ´an, D. Bena ides, and M. To o. Au oma ing he p o- cu emen o web se ices. In 1s . In e na ional Con e ence on Se ice O ien ed Compu ing ICSOC’03, pages 91–103, T en o, I aly, 2003. Sp inge –Ve lag LNCS 2910. 15. Ch is ian P eho e . Fea u e-o ien ed p og amming: A new way o objec composi ion. Con- cu ency and Compu a ion: P ac ice and Expe ience, 13(6):465–501, 2001. 16. D. S ei e d , M. Riebisch, and I. Philippow. De ails o o malized ela ions in ea u e mod- els using ocl. In P oceedings o 10 h IEEE In e na ional Con e ence on Enginee ing o Compu e –Based Sys ems (ECBS 2003), Hun s ille, USA. IEEE Compu e Socie y, pages 45–54, 2003. 17. J. an Gu p, J. Bosch, and M. S ahnbe g. On he no ion o a iabili y in so wa e p od- uc lines. In P oceedings o he Wo king IEEE/IFIP Con e ence on So wa e A chi ec u e (WICSA’01), IEEE Compu e Socie y, pages 45–54, 2001. ENAGER – A Tool o Requi emen s Enginee ing in Sys em Family Con ex Rod igo Ce ón1, F ancisco Valse a, Jose L. A ciniegas1, Jose L. Ruiz1, Juan C. Dueñas1 Depa men o Enginee ing o Telema ic Sys ems, Uni e sidad Poli écnica de Mad id, ETSI Telecomunicación, Ciudad Uni e si a ia, s/n, E-28040 Mad id {ce on, jla ci, jl uiz, jcduenas}@di .upm.es [email p o ec ed]om Abs ac . So wa e indus ies a e pu suing he de elopmen o so wa e in ensi e sys ems wi h a highe deg ee o euse, educ ion o cos s, and sho e ime o ma ke . One o he success ul app oaches aken is based on he de elopmen o se s o simila sys ems; in i , de elopmen e o s a e sha ed. This app oach is known as Sys em Families. A weakness in sys em amilies is he lack o exis ence o ools. Requi emen s enginee ing is a ma u e a ea oday. Howe e , i needs some imp o emen s in he sys em amily ools ield. In his posi ion pape , you can ind a s udy o ool suppo o equi emen s in sys em amily con ex . An analysis o ac ual ools was made. Some equi emen s o ool suppo in his ield a e p oposed. In addi ion, some me ics o he ool a e p oposed. A ool we called ENAGER was implemen ed wi h his in o accoun . 1. In oduc ion Fo many yea s, so wa e indus ies ha e been ying o achie e he de elopmen o so wa e in ensi e sys ems wi h a highe deg ee o euse, cos educ ion, and sho en o ime o ma ke . Sys em Families (SF) a e conside ed as one o he mos success ul app oaches o do i . I ocuses on he educ ion o he ime o ma ke and de elopmen cos s by he p o ision o a se o elemen s ha a e common o a numbe o sys ems. The e o e, i is a se o sys ems o se ices; also, i sha es a signi ican pa o hei de elopmen e o . Since i sha es many elemen s, i is sensible o hink ha he concep s behind hem a e also common. The main goal o his posi ion pape is o p esen ool suppo o equi emen s modelling and managemen in SF. Some equi emen s o ool suppo in his ield a e p oposed. In addi ion, some me ics o he ool a e p oposed. A ool we called ENAGER was implemen ed wi h his in o accoun . The basic wo k has been pe o med in he CAFÉ p ojec [1]. The CAFÉ e e ence amewo k (CRF) gi es a guide o classi y he ac i i ies and models ela ed wi h SF de elopmen . In summa y, CRF can be di ided in wo main pa s CAFÉ P ocess Re e ence Model (CAFÉ – PRM) and CAFÉ Asse s Re e ence Model (CAFÉ - ARM). The objec i e o his model is o ep esen majo ac i i ies and me hods ope a ing on he co e asse s, o allow he mapping o he con ibu ion/ ools agains a common e e ence. 1 Rod igo Ce ón and Jose L. A ciniegas a e isi ing p o esso s om Uni e sidad del Cauca, Colombia. Rod igo Ce ón is sponso ed by COLCIENCIAS - Colombia. The wo k pe o med by José L. A ciniegas has been pa ially de eloped in he p ojec TRECOM, g an ed by Spanish Minis y o Science and Technology unde e e ence TIC2002-04123-C03-01. The wo k done by Jose L. Ruiz, and Juan C. Dueñas has been pa ially pe o med in he p ojec CAFE and FAMILIES (Eu eka 2023, ITEA ip00004, ip02009), pa ially suppo ed by he Spanish company Tel en and by he Spanish Minis y o Science and Technology, unde e e ence TIC2002-10373-E. 2. Rela ed Wo k In ou wo k, we analyze and ex ac some conclusions abou ac ual ools. We canno analyse all he ools bu we saw he documen a ion o he mos popula in he so wa e ma ke and in one open sou ce p ojec s. The ools in deep analysed a e hose whe e we can download an e alua ion copy and echnical documen a ion (i.e.: Requisi eP o, In eg al Requisi e Analyze (IRqA), Dis ibu ed Requi emen Enginee ing Sys em (DRES), Compu e Aided Requi emen s Enginee ing (C.A.R.E) and RMT ak). O he ools we e aken in o accoun since we can download echnical documen a ion bu we canno ha e an e alua ion copy. Fo his posi ion pape , we only summa ized he ools in deep analysed. Ra ional Requisi eP o was de eloped and i is main ained by Ra ional (now a di ision o IBM) [2]. They said ha sys ema ic cap u e, con ol and change communica ion is a key poin in equi emen managemen . Fo hem, ano he key poin is he ans o ma ion om he in e nal eposi o y o ma o ex documen s. In his way, equi emen s can be managed and communica ed. Requi emen s a e managed in a p ojec basis. Th ee ypes o ela ionships a e u ilized o associa e equi emen s. They a e called hie a chical, ace and inde ini e. As a conclusion, we ound ha his ool was no designed o SF. IRqA was de eloped and i is main ained by “TCP Sis emas e Ingenie ía” (A Spanish so wa e o ganiza ion). They said ha domain analysis and equi emen managemen a e i s key ea u es. I is based in Me a-IRqA as i s amewo k [3]. The p ojec is he basis o he equi emen s managemen . Inside he p ojec , he use de ines domains. They a e use ul o di ide he complex p oblems. The modelling elemen s a e se ices, equi emen s, concep o en i y and implemen a ion class. These modelling elemen s a e ela ed o he o he s in p ede ined ways. Ano he key poin is he de ini ion o e ms and ace s. One e m is a speci ic ocabula y in he p oblem con ex . One ace g oups se s o e ms. As a conclusion, we ound ha his ool was no designed o SF. Howe e , i gi es us imp o emen s in he way o managemen o equi emen s. DRES is an open sou ce ool o equi emen s managemen [4]. I is s ill in de elopmen and he la es e sion is 0.8. Requi emen s a e s o ed in a adi ional way. The basic idea is o s o e and o e ie e equi emen s. I s s eng h is o e ie e equi emen s om any place using s anda d b owse . P ojec is i s p ima y uni . One p ojec con ains ca ego ies. Requi emen s a e inside one ca ego y. The ool can gene a e documen a ion o i s use s. This ool was no designed o SF. C.A.R.E was de eloped by he Ge man o ganiza ion SOPHIST GROUP [5]. The main elemen is a ex ual documen ; i is called speci ica ion. O he elemen s a e ela ed o his one. Requi emen s and de ini ions a e he o he elemen s de ined by his ool. Requi emen s a e di ided in o unc ional and non- unc ional (use in e ace, con ac , quali y o se ice, e c.). De ini ions a e employed o ep esen concep s. These concep s can be in o he domain space o in he solu ion space. De ini ions a e g ouped in glossa ies. RMT ak was de eloped by RBC Inc [6]. P ojec is i s p ima y uni . P ojec s a e composed o a se ies o documen s. Documen s con ain equi emen s o a p ojec . Requi emen classi ica ion can be om use , ha dwa e o so wa e. Inside a documen , a label iden i ies a equi emen . Requi emen s can be ela ed by means o a a he -son ela ionship. This ool was no designed o SF. DOORS was de eloped by Telelogic AB [7]. I is one o he mos widely used ools. One o i s key ea u es is i s mul i-use suppo en i onmen . The ool eases communica ion among i s use s. Easy na iga ion and unde s anding o i s in o ma ion o use s a e key ea u es. They a e ob ained by means o objec b owse , g aphics, s a is ics, e c. Collabo a i e suppo o i s use s is ano he ea u e. Tools o discussion and change con ol a e inside i . Finally, ace suppo ha eases alida ion and equi emen s acking a e included. RDT and SLATE we e analyzed oo. Fo RDT, SLATE and DOORS we canno ob ain e alua ion copy and we canno do some expe imen s wi h hem. 3. Gene al Issues F om he ool e alua ion, we ound he ollowing gene al issues. • The e is no comme cial o open sou ce ool wi h ull suppo o SF. All o hem a e hough o single p ojec de elopmen . • Domain Enginee ing is no aken in o accoun in all ools. IRqA de ines domain bu o a single p ojec . I hinde s euse in o se e al p ojec s. • Sys em enginee ing suppo is pa ially co e ed. • In e nal ace mechanisms o equi emen s a e de ined in he ools s udied. No ex e nal ace mechanisms can be ound o o he de elopmen phases. • We need o de ine a sys em o classi ica ion o equi emen s in SF. They can be classi ied as unc ional and non- unc ional bu a u he di ision is necessa y. • We need o de ine elemen s ha ease equi emen s managemen in SF. They mus ake in o accoun de elopmen and e olu ion o SF. Con lic s and decisions mus be elemen s o ake in o accoun . • E olu ion o sys em gi es us ano he issue known as e sion con ol. Fo SF, his issue is e y big. Some o hei p oblems a e mul iple sys ems, asse s and equi emen s. • We need aids ha help us in documen a ion gene a ion. • Use s need ways o assess equi emen s s o ed. In he design o ENAGER (ou ool), we y o cope wi h hese gene al issues ound. 4. Requi emen s o ENAGER [8], [9], [10], [11], [12], [13] and [14] gi es us gene al ideas o ool equi emen s in SF con ex . The ollowing equi emen s a e om his poin o iew. • O ganiza ions need o manage mo e han one sys em amily. In gene al, la ge o ganiza ions need suppo o se e al p ojec s hey wo k in. I hey can sha e mos o i s p oduc s equi emen s be ween se e al p ojec s, i can sho en i s p oduc de elopmen cycles. • S akeholde s a e in cha ge o p o iding de ini ion o domain and applica ion e ms. They mus be s o ed in glossa ies. • Glossa ies need o be sha ed ac oss all o ganiza ional uni s. I hey can be sha e ac oss all de elopmen uni s hen a common unde s anding is os e ac oss he o ganiza ion. • Va iabili y managemen o equi emen s need o be sol ed by means o con lic s and decisions. Requi emen s mus ha e been ela ed o o he s by means o se e al si ua ions. A con lic can be used o exp ess a iabili y. A decision can be used o sol e a iabili y. • Func ional equi emen s need o be aced o use cases. I closes he gap be ween equi emen s and design. • Con lic impac need o be aced o use cases. The p oblems de ec ed among equi emen s can in luence ollowing de elopmen phases. I de elope s ha e aids ha help hem in acking hose p oblems o he ollowing phases, hey imp o e hei pe o mance in de elopmen phases. • Decisions mus be s o ed. I is necessa y o know all he decisions aken in o accoun o de i e one p oduc . I helps in change con ol and e olu ion. • Requi emen s need o be classi ied. A hie a chical axonomy mus be used. All ools p oposed a model o hie a chies o i s equi emen . • Func ional equi emen s can be classi ied as goal, ea u e and a omic. Wi h h ee le els and ela ion among hem all, he possibili ies can be co e ed (exclusion, coexis ence, manda o y, al e na i e). • A so wa e quali y model o non- unc ional equi emen is ecommended o he ool (i.e.: ISO 9126). • Each equi emen mus ha e p ope ies. • We need easy na iga ion in o de o keep equi emen managemen unde con ol. These gene al equi emen s we e aken in o accoun o he design o ENAGER. 5. Me ics o ENAGER In o de o asses di e en SF equi emen s ENAGER de ines some me ics. The i s g oup is called F amewo k me ics. They a e in ended o asses he quali y o he o e all speci ica ion s o ed. The second g oup is called SF me ics. They a e in ended o asses he quali y o one o he SF s o ed. F amewo k me ics a e subdi ided in wo g oups called: complexi y and consis ency. The i s one gi es us an idea o how big is he in o ma ion s o ed. The second one gi es us an idea o how cohe en is he in o ma ion s o ed. SF me ics a e subdi ided in i e g oups called: complexi y, consis ency, cos , ma u i y and common/ a iable. The i s wo a e simila o hose ones in amewo k bu o only one SF. The hi d one gi es an idea o he o e all cos o SF. The ou h one gi es an idea on how ma u e is he SF unde analysis. The i h one gi es an idea o he common equi emen s compa ed o he sys em speci ic equi emen s. 6. Conclusions and Fu u e Wo k We ha e pe o med his wo k ying o cope wi h he indus ial equi emen s o ools in SF equi emen s managemen . The posi ion pape gi es a pa ial answe o ool suppo in equi emen s modelling con ex . Fu he esea ch mus be done o sol e i comple ely. In any case, he ool p esen ed he e co e s se e al o he speci ic needs o he SFE as ega ds equi emen s managemen and he aceabili y o o he sys ems- models. Back o he speci ic issues in equi emen s managemen o SF, he ool allows: • The ull na iga ion and aceabili y ma ixes gene a ion o in e nal aces, despi e he complexi y o equi emen s in he SF. The usage o a web na iga ion schema helps in keeping he complexi y unde con ol. • The basic ela ions be ween equi emen s a e co e ed (exclusion, coexis ence, manda o y, al e na i e), as well as ela ions du ing e olu ion ( e ision); hie a chies (gene alisa ion, specializa ion) o equi emen s a e also suppo ed. • The concep o sys em in he sys em amily appea s in his ool as a se o equi emen s. Being hese equi emen s ela ed, he ool allows o he calcula ion o “cohe en ” sys ems, con aining se s o non- con lic ing equi emen s. As well, he p og ess in de elopmen o a ce ain sys em can be ep esen ed by he g owing se o equi emen s co e ed. • Decisions a e ep esen ed explici ly. I is possible o na iga e om each equi emen o he decisions ela ed, and he decisions ela ed o each sys em. Con lic s a e also exp essed, so i s easons and s akeholde s ela ed can be iden i ied and sol ed. • The di e en ia ion be ween common and speci ic equi emen s is no s a ically wi ed; ins ead, as he e is na iga ional capabili y om- o sys ems and each equi emen , i is possible o know in a ce ain poin in ime how many o he sys em in he SF a e a ec ed o one equi emen . Re e ences 1. an de Linden, F.: So wa e P oduc Families in Eu ope: The Esaps & Ca é P ojec s. IEEE So wa e, Vol. 10 No. 4. IEEE Compu e Socie y, Los Alami os, CA (2002) 41-49 2. IBM – Ra ional.: Ra ional Requisi eP o. h p://www-306.ibm.com/so wa e/awd ools/ eqp o/, (2004) 3. TCP Sis emas e Ingenie ía: IRqA in eg al equisi e analyze . h p://www.i qaonline.com/index_i qa.h m, (2004) 4. Kowalczykiewicz, K.: DRES Dis ibu ed Requi emen s Enginee ing Sys em, h p://ophelia.cs.pu .poznan.pl/xd e/, (2003) 5. SOPHIST GROUP: C.A.R.E Compu e Aided Requi emen s Enginee ing, h p://www.sophis .de/sopg oupeng.ns /(ynDK_ amese s)/Main, (2003) 6. RBC P oduc De elopmen : RMT ak, h p://www. m ak.com/, (2003) 7. Telelogic AB: Telelogic DOORS, h p://www. elelogic.com/, (2003) 8. Kuusela, J., Sa olainen J.: Requi emen s Enginee ing o P oduc Lines. In: P oceeding o he 2000 In e na ional Con e ence on So wa e Enginee ing, ACM-IEEE, New Yo k, NY (2000) 60-68 9. Lu ikhuizen P. (ed.): Requi emen s Modelling and T aceabili y. ESAPS, WP3, De i a ion o P oduc s and E olu ion o Sys em Families WP3-0106-01 Technical Repo , (2001) 10. IEEE S anda ds Boa d: IEEE Recommended P ac ice o So wa e Requi emen s Speci ica ions. IEEE s d 830, 1993. Ins i u e o Elec ical and Elec onics Enginee s, New Yo k, NY (1993) 11. Alonso, A., León, G., Dueñas, J.C., de la Puen e, J.A.: F amewo k o Documen ing Design Decisions in P oduc Families De elopmen . In: P oceedings o he Thi d IEEE In e na ional Con e ence on Enginee ing o Complex Compu e Sys ems. IEEE Compu e Socie y, Los Alami os, CA (1997) 206- 211 12. Thaye , R., Do man, M.: So wa e Requi emen s Enginee ing. 2nd edn. IEEE Compu e Socie y P ess, Washing on (1997) 13. IEEE-SA S anda ds Boa d: IEEE Guide o De eloping Sys em Requi emen s Speci ica ions. IEEE s d 1233, 1998. Ins i u e o Elec ical and Elec onics Enginee s, New Yo k, NY (1998) 14. ISO/IEC: So wa e enginee ing -- P oduc quali y -- Pa 1: Quali y model. ISO/IEC 9126-1:2001. In e na ional O ganiza ion o S anda diza ion, Gene e, (2001) Expe ience-based App oach o Requi emen s Reuse in P oduc Families wi h DOORS An onio Monzón, Juan Ca los Dueñas1 1 The wo k done by Juan C. Dueñas has been pa ially pe o med in he p ojec s CAFE and FAMILIES (Eu eka 2023, ITEA ip00004, ip02009), pa ially suppo ed by he Spanish company Tel en and by he Spanish Minis y o Science and Technology, unde e e ence TIC2002-10373-E. Depa amen o de Ingenie ía de Sis emas Telemá icos, Uni e sidad Poli écnica de Mad id Requi emen s euse has e ealed as one o he mos impo an issues in he Re- qui emen s Enginee ing (RE) ield o e he las yea s, as his discipline has a huge impac on sys ems p oduc ion and main enance cos s. Fo his eason i is equi ed o p o ide speci ic solu ions o be di ec ly applied o eal indus ial en i onmen s wi h he help o ools. The objec i e o his pape is o show a p ac ical app oach o equi emen s euse in p oduc amilies suppo ed by a well-known comme cial ool (Telelogic DOORS). Fo illus a ion pu poses he pape ocuses on he pa icula domain o comme cial ai c a p oduc ami- lies. 1 In oduc ion The p ope applica ion o equi emen s enginee ing p inciples is a mus o he suc- cess ul de elopmen o so wa e in ensi e sys ems; as he size and complexi y o hese sys ems g ow, i becomes e iden ha ool suppo is a key. The de elopmen o p oduc amilies –ins ead o indi idual p oduc s- is a s a egic imp o emen ollowed by many companies, pu suing wide euse o any asse o he p oduc amily, includ- ing equi emen s. Many wo ks in he li e a u e ha e al eady been p oduced abou equi emen s euse bu ew o hem p esen p ac ical solu ions o he p oblem wi h he help o comme cial ools. This is he aim o his pape . I is impo an o ema k ha his is a p ac ical expe ience pape wi h no o e y ew di ec heo e ical con ibu ions. The concep o Requi emen handled he e is he exp ession o a need ep esen ed as a classical na u al language desc ip ion, wi h some addi ional a ibu es o managemen pu poses. Nei he o mal app oach no domain modeling echniques a e p oposed. In o de o p o ide a be e unde s anding o he concep s in ol ed in his pape , se e al UML diag ams will be used in o de o ep esen he in o ma ion me amodels behind he p oblem o be sol ed. On one hand, he p ope suppo ing ool basic me amodel will be ep esen ed. On he o he hand, he conc e e equi emen s me a- model used in a pa icula p og am (p oduc amily) will be ep esen ed. And inally, he pa icula app oach o euse echniques applied will also be ep esen ed as an UML diag am. The easons o choosing he a i ac s and he pa icula in o ma ion s uc u es ha e been mainly based on he expe ience o equi emen s enginee ing eams in he ield. Da a s uc u e has been p oposed and e ined o e a couple o decades and i has been ag eed by se e al pa ies. The e o e, i could be said ha he models ha e been alida ed by hei indus ial applica ion du ing yea s. The main in o ma ion sou ces o equi emen s a e he Func ional Requi emen s Documen s (FRD’s). These documen s con ain bo h ex ual desc ip ions o he p ob- lem domain (g aphical and abula in o ma ion included) and also p ope equi e- men s. Requi emen s a e agged h ough unique iden i ie s and clea ly delimi ed by sho ex ual desc ip ions. 2 DOORS Me amodel DOORS is he mos ex ended RE ool in ae ospace a ena. Al hough i is no inal- use o ien ed and i is di icul o use he ool wi hou speci ic aining, i shows an ex emely adap a ion capabili y, e y use ul in echnology-d i en en i onmen s. The main concep s in ol ed in his ool a e ep esen ed in Figu e 1. Fig. 1. DOORS Basic Me amodel I is jus wo h o men ion ha DOORS Objec me a-concep is he in o ma ion elemen sui able o ep esen equi emen s. Bu his is so gene al ha can be also used Requi emen s Reuse Page 3 • Help - FAQ The a ibu es o he objec s a e equi emen s o hem. I u ns ou ha qui e o en, e en hough objec s hemsel es may be qui e di e en , he equi emen o ce ain a ibu es is e y common. Bene i s o Requi emen s Reuse Bene icial esul s o ou equi emen s euse include he ollowing: • Requi emen s Reuse cu new applica ion de elopmen by app oxima ely 80%. • Code om Common Requi emen s imp o emen s quickly mig a ed ac oss si es. • Requi emen s e o s mo e eadily de ec ed by inc eased use ac oss mul iple si es. • Requi emen s e o s a e ixed p io o disco e y by use s on o he si es. • Simila equi emen s a exis ing si es used o demons a e p oposed unc ionali y o new cus ome s. Lessons Lea ned A e de eloping a ce ain le el o capabili y in p ocess implemen a ion equi emen s euse, he p ocess implemen a ion Subjec Ma e Expe s (SMEs) o en end up knowing mo e o he cus ome s equi emen s han he cus ome does. This shows up in equi emen s e e y whe e. The ollowing a e wo examples ha occu ed: 1. Dele e a Reques capabili y: he cus ome wan ed he capabili y o dele e a eques . I acknowledged he equi emen and indica ed ha I would implemen he eques by iden i ying a common equi emen o ag he eques a “Dele ed” bu ac ually lea e i in he da abase. This design equi emen was a well es ablished bes p ac ice equi emen . The cus ome insis ed on he eques being ac ually dele ed. I Requi emen s Reuse Page 4 explained ha i i we e eally dele ed, i would be i eco e able. They s ill insis ed and so I changed he equi emen . La e , when he p ocess was in p oduc ion, managemen complained when eques s ha hey had been acking suddenly disappea ed. 2. “Name” is a Requi ed A ibu e: a cus ome o a Requi emen s T acking p ocess did no iden i y a Requi emen Name as one o he a ibu es o he equi emen . I poin ed ou ha such a ield was needed o display pu poses (pa o he accumula ed bes p ac ice euse equi emen s). They insis ed ha i was no needed. A e going in o p oduc ion, a display showed only he i s ew cha ac e s o he da a in he ield (say 55). Thus hey quickly ealized ha hey needed a “summa y” s a emen o sea ch on and so by, so we added a i le. The ollowing a e some o he Lessons lea ned in dealing wi h cus ome s in implemen ing hei business p ocesses: z The Cus ome is NOT Always Righ : – Dele ing da a – CI # – CI Desc ip ion z The “Real” P ocess is no he one o iginally desc ibed. z The Da a Will Se You F ee z Ha e Some hing To Talk Abou o demons a e equi emen s commonali y z Keep Reco ds o Requi emen s, bu be eady o change (cons an ly!) Acknowledgmen s This pape and accompanying p esen a ion has been suppo ed inancially by The Boeing Company o Chicago, Illinois, USA. Requi emen s Reuse Page 5 Re e ence Conway, William E. Chai man & CEO, Conway Managemen Co. h p://www.conwaymgm .com/ Technical Repo , Uni e sidad Poli écnica de Mad id, Spain, 2004. ISBN: 84-688-7314-4.