A first prototype of a new repository for feature model exchange and knowledge sharing

A i s p o o ype o a new eposi o y o ea u e model exchange and knowledge sha ing Da id Rome o, José Á. Galindo, Jose-Miguel Ho cas and Da id Bena ides Uni e si y o Se ille Se ille, Spain {d o gan idez,jagalindo,jho cas,bena ides}@us.es ABSTRACT Fea u e models a e he “de ac o” s anda d o a iabili y modelling and a e used in bo h academia and indus y. The MODEVAR ini ia- i e ies o es ablish a common ex ual ea u e modelling language ha can be used by di e en communi ies and can allow in o ma- ion sha ing. Fea u e model ela ed esea ches use di e en models o di e en pu poses such as analysis, sampling, es ing, debug- ging, eaching, e c. Those models a e sha ed in p i a e eposi o ies and he e is a isk ha all ha knowledge is sp ead ac oss di e en pla o ms which hinde collabo a ion and knowledge euse. In his pape , we p opose a i s wo king e sion o a new ea u e model eposi o y ha allows o cen alise he knowledge gene a ed in he communi y oge he wi h ad anced capabili ies such as DOI gene a ion, an API, analysis epo s, among o he s. Ou solu ion is a on end in e ace ha uses he popula open science epos- i o y Zenodo as an end poin o ma e ialise he s o age o all he in o ma ion. Zenodo is enhanced wi h cha ac e is ics ha acili a e he managemen o he models. The idea o ou eposi o y is o p o ide exis ing bu also new ea u es ha a e no p esen in o he eposi o ies (e.g., SPLOT). We p opose o popula e ou eposi o y wi h all he exis ing models o many sou ces including SPLOT. CCS CONCEPTS •So wa e and i s enginee ing → So wa e p oduc lines ; Re- qui emen s analysis; So wa e design enginee ing; So wa e imple- men a ion planning;␣ KEYWORDS ea u e model eposi o y, cha ac e is ics, a iabili y, equi emen s 1 INTRODUCTION Fea u e models a e he “de ac o” s anda d o a iabili y modelling and a e used in bo h academia and indus y. Fea u e models we e in oduced in he FODA (Fea u e-O ien ed Domain Analysis) ech- nical epo [ 20 ] in 1990, by he So wa e Enginee ing Ins i u e. Nowadays, ea u e model pu poses anges be ween mul iple ac i i- ies such as analysis [ 7 ], sampling [ 18 ], es ing [ 19 ], debugging [ 15 ], and eaching [21], among o he s. Cu en ly, ea u e model esea che s sha e hei models among hei so wa e a e ac s o imp o e esul s’ eplicabili y and i open-science s anda ds [ 4 ]. Howe e , hose models a e o en sha ed in p i a e eposi o ies and a e commonly sp ead ac oss di e en pla o ms, as o example, Gi Hub. Despi e some ea u e models eposi o ies exis such as SPLOT [ 14 ], ESPLA [ 13 ], o LVAT 1 , hey p esen se e al d awbacks. Fi s , he e is no common empla e o so wa e APIs o access he eposi o ies, making di icul o sea ch o , ob ain me ics, download, o e en au oma e he analyses o he ea u e models. Fo example, SPLOT does no p o ide a REST API se ice o consume he knowledge p o ided. In addi ion, he c ea ion, uploading and subsequen main enance o ea u e mod- els a e igid and wi h limi ed in e ac ion. Second, he e is limi ed knowledge sha ing. Conc e ely, since he e is no common o gani- sa ional s uc u e no s anda d guidelines o he uploaded ea u e models, i is di icul o sha e knowledge in he communi y. To add ess his issue, he MODEVAR [ 1 ] ini ia i e ies o es ablish a common ex ual ea u e modelling language ha can be used by di e en communi ies and enables in e ope abili y. Howe e , he e is a lack o sha ing mechanisms o such ea u e modelling lan- guage. Finally, exis ing eposi o ies a e ou da ed e y soon because o hei echnical deb [ 9 ]. Technology used o he c ea ion o such eposi o ies has been discon inued and has no been adap ed o he echnological equi emen s ha use s demand nowadays, such as a esponsi e-web design [ 10 ], secu e h p p o ocol (h ps), o use load balancing among o he mode n ea u es. Ul ima ely, a new eposi o y can be en isioned i he communi y wan s o easily sha e knowledge in he ollowing yea s. To ease o hose limi a ions, we ha e designed FMRepo, a i s p o o ype o a new eposi o y o ea u e model exchange and knowledge sha ing. FMRepo aims o in eg a e bo h he common a iabili y language and ea u e model analysis capabili ies ha a e in consonance wi h o he MODEVAR ini ia i es, such as he p o- posed uni e sal a iabili y language (UVL) [ 1 ], and he amewo k o au oma ed analysis o ea u e models (AAFM) [ 6 ], as well as adhe ing o open science p inciples [ 16 ]. The main mo i a ion o building up a new eposi o y is o sha e ea u e modelling ela ed 1h ps://gsd.uwa e loo.ca/ ea u e-models-in- he-wild.h ml SPLC ’21, Sep embe 6–11, 2021, Leices e , Uni ed Kingdom Da id Rome o e al a e ac s and enable he ep oducibili y o expe imen s wi h ea u e models. Howe e , he e a e o he ac o s ha p omo e he c ea ion o such eposi o y: Mee open science s anda ds. The e a e nume ous guidelines wi hin he amewo k o Open Science: Open Access, Open Da a, Open Rep oducible Sea ch, Open Science E alua ion, Open Science Policies and Open Science Tools [ 4 ], o name a ew, a e pa o he undamen al pilla s. Ou eposi o y ies o co e all he minimum equi emen s o be classi ied in his axonomy and o collabo a e in he communi y. Teaching and educa ion. Knowledge sha ing o ea u e mod- els allows he signi ican educ ion o a ious academic s ud- ies in ela ion o his echnology. Building he eposi o y wi hin he scope o ee and un es ic ed access can p omo e he use o educa ional and ins uc ional con en . Knowledge agg ega ion. This eposi o y ies o b ing o- ge he he dispe sed knowledge ha exis s abou ea u e models gene a ed in a knowledge base ha se es as suppo o a mo e in-dep h and ex ensi e s udy o his echnology, sa ing ime and e o . In his pape , we p esen he design and implemen a ion o FM- Repo o ul il new demands o esea che s and de elope s. This is a i s wo king e sion o a new ea u e model eposi o y ha allows o cen alise he knowledge gene a ed in he communi y oge he wi h ad anced capabili ies ha do no exis in o he epos- i o ies such as SPLOT [ 14 ] o he ESPLA ca alog [ 13 ]. The main con ibu ions o FMRepo a e: • In eg a ion o Open Science echnologies. FMRepo has been buil on op o open-science echnologies suppo ed by he Eu opean Resea ch Council [ 11 ]. To g an access o uploaded models in he u u e o he sake o eplicabili y. • As pa o he Open Science echnologies, FMRepo p o ides DOI gene a ion o ea u e models h ough he Zenodo epos- i o y [ 16 ]. The gene a ion o DOI can be in e es ing o he wo k o esea ch a icles. Resea che s can upload hei mod- els o FMRepo and di ec ly ob ain a uni e sal iden i ie o be able o in eg a e i in o hei academic wo k. This educes he ime and complexi y o gene a ing new knowledge using a single ool. • An REST API ha allows a se ies o ad an ages o enable he use o FMRepo as a base sys em in u u e p ojec s, p o iding au oma ic access o he ea u e models. • In eg a ion wi h ea u e model analysis echnologies as a se ice. This enables esea che s and de elope s o consume he exis ing models wi hin he eposi o y and alida e hei esea ch. • Popula ion o FMRepo wi h eal-wo ld and academic ea u e models om many sou ces, including exis ing eposi o ies (e.g., SPLOT). The emainde o his pape is o ganised as ollows: Sec ion 2 desc ibes he necessa y concep s o he echnologies men ioned in his pape ; Sec ion 3 lis s he cha ac e is ics and needs ha we co e wi h he unc ionali ies implemen ed in he eposi o y; Sec ion 4, make a gene al summa y o he main modules ha make up he a chi ec u e and he eason o i s p esence; Sec ion 5, de ails he main p oblems o he modules and he solu ion we ha e gi en; and Sec ion 6 whe e we explain he eal eason o c ea ing his eposi o y and why i would be con enien o he communi y o ge in ol ed. 2 OPEN SCIENCE Open Science [ 16 ] is a mo emen p omo ed by he OECD 2 coun ies and p omo ed by he Eu opean Commission. I ad oca es ee access by ci izens o he esul s o scien i ic esea ch, da a, esou ces, esul s, hough s, as well as ha he esul s and disco e ies o scien i ic esea ch a e uni e sally accessible and un es ic ed. Some o he exis en e o s o suppo open science a e he Eu opean OpenAIRE 3 p ojec and he In enio 4 amewo k o build open- science eposi o ies. In enio is an open sou ce p ojec by CERN ha co e s a eposi o y/documen managemen pla o m (In enioRDM), an in eg a ed lib a y sys em (In enioILS) and a code lib a y o build la ge-scale in o ma ion sys ems (such as In enioRDM and In enioILS). FMRepo p omo es he open-science s anda ds h ough he use o Zenodo. Zenodo 5 is an Open Science eposi o y ha has been de eloped on op o he In enio amewo k which p o ides pe sis- ence o he da a uploaded as well as o e s a DOI (Digi al Objec Iden i ie ) o enable easy ci a ion o da a. Zenodo also wo ks as a backup o all he da a uploaded. 3 REPOSITORY CHARACTERISTICS In p e ious wo k [ 5 ], we de ined 12 cha ac e is ics ha a ea u e model eposi o y should ha e. Those cha ac e is ics we e de ined obse ing exis ing eposi o ies and we e p esen ed as use s o ies ollowing he empla e o o “As a [pe sona], I [wan o], [so ha ].”. In his sec ion, we de ail how FMRepo add esses hose cha ac e is ics and discuss he limi a ions o exis ing eposi o ies like SPLOT. (1) As a esea che , I wan o upload models o he eposi o y, so ha my models a e a ailable o o he s. This is one o he mos basic cha ac e is ics, howe e , exis ing eposi o ies do no p o ide simple uploading o models and complica e his op- e a ion. Fo ins ance, while SPLOT equi es o upload models wi hin a URL based on non-secu e h p, ESPLA equi es o c ea e an issue o a pull eques in Gi Hub, in o de o upload he ea u e model. FMRepo allows uploading ea u e models h ough a simple web o m. (2) As a esea che , I wan o syn ax check my models be o e uploading in he eposi o y, so ha my models a e syn ax e - o ee. FMRepo will p o ide suppo o se e al languages and no a ions, including p incipally he uni e sal a iabil- i y modelling language p omo ed by he MODEVAR ini ia- i e [1]. (3) As a esea che o use , I wan o download models om he eposi o y, so ha I can eplica e expe imen s o use exis ing models. The way models a e downloaded could a y om one by only ba ch downloads. Fo ins ance, SPLOT suppo s downloads one by one, o cing esea che s o pe o m ha d- manual ask o o build cus om sc apping sc ip s o download 2h ps://www.oecd.o g/ 3h ps://www.openai e.eu/ 4h ps://in enioso wa e.o g/ 5h ps://zenodo.o g/ all models a once. FMRepo will p o ide suppo o lexible downloading o models, om indi idual models o se o models. (4) As a esea che , I wan o gene a e a uni e sal iden i ie o my models in he eposi o y, so ha my models a e ci able and easily iden i ied. This cha ac e is ic is a new con ibu ion o FMRepo which p o ides DOI gene a ion hanks o Zenodo. The DOI is speci ic o a ea u e model so ha i can be in eg a ed in o he esea ch pape s. (5) As a esea che o use , I wan o manage e sions o my models, so ha I can compa e di e en e sions. Exis ing eposi o ies s o e di e en e sions o he ea u e models as independen (new upload) iles. While his cha ac e is ic could make he eposi o y mo e complex o main ain, ou objec i e is o p o ide e sion con ol in FMRepo. (6) As a use , I wan o sea ch o models in he eposi o y, so ha I can ind he models I am in e es ed in. Only ESPLA allows sea ching i s ca alog using ags. In ac , he in o ma ion is s o ed in a .cs ile. In con as , ou FMRepo p o ides di e en sea ch ypes based on ag, me ada a (e.g., au ho s) o e en me ics (e.g., size o he ea u e model). (7) As a use , I wan o display models in he eposi o y, so ha I can ha e a glance a he model in he case I wan o use i . Exis ing eposi o ies like SPLOT displays he model as a ee isualisa ion along wi h a e y limi ed numbe o me ics. The ESPLA ca alog does no p o ide any isualisa ion o he models. Ou p oposal, in addi ion o displaying a ea u e model h ough a g aphical o ex ual isualisa ion, also p o- ides insigh on ea u e model me ics so ha use s can ha e a quick cha ac e isa ion o he model be o e using. (8) As a use o esea che , I wan o know some indica o s abou he models in he eposi o y such as a ings o numbe o downloads, so ha I can compa e models. FMRepo s o es his in o ma ion as a me ada a o he ea u e model. O he eposi- o ies like SPLOT do no p o ide me ada a in o ma ion such as how many imes a model has been downloaded o a ed. (9) As a de elope o esea che , I wan o ha e an API ha can in e ac wi h he eposi o y, so ha I can p og amma ically access eposi o y’s in o ma ion. The con en s o FMRepo can be accessed p og ama ically in he o m o an API. This allows a de elope o call a me hod o download some models wi h some cha ac e is ics o example a gi en numbe o ea u es. (10) As a use , I wan o ha e ecommenda ions acco ding o my p o ile, so ha I can do be e model selec ions. Since FMRepo s o es me ics o he models acco ding o downloads o use a ings, a ecommenda ion sys em based on use p o iles and a ings can be de eloped o be e selec models using he p e ious in o ma ion s o ed in Zenodo. Howe e , his seems o be a mo e long e m ea u e o be conside ed. (11) As a esea che , I wan o c ea e communi ies, so ha I can ha e a common space o manage my models. The unde lying Zenodo sys em o FMRepo allows managing mul iple use s which can be g ouped a ound a esea ch g oup, uni e si y o communi y. Howe e , his cha ac e is ic has been discussed also in dep h among he au ho s and no consensus has been al eady eached. (12) As a use , I wan o see model’s me ics, so ha I can ha e ex a in o ma ion abou he models. The in eg a ion o ea- u e model analysis echnologies as a se ice wi hin FMRepo allows p o iding implici in o ma ion o ea u e models such as he numbe o ea u es, he numbe o ela ionships, s uc- u al me ics [ 2 ], o o he analysis ope a ions (e.g., dead ea u es, numbe o con igu a ions,...) [3]. 4 ARCHITECTURE OVERVIEW This sec ion p esen s he a chi ec u e o FMRepo as well as he unde line echnology o p o ide suppo o knowledge sha ing and ea u e models exchange. Figu e 1 shows an o e iew o he a chi ec u e di ided in wo main pa s: he in e ace ( op o Figu e 1) and he h ee main modules (bo om o Figu e 1). Figu e 1: FMRepo a chi ec u e. 4.1 FMRepo in e ace The in e ace and communica ion be ween he di e en compo- nen s and he use ha e been de eloped using he La a el amewo k . La a el 6 is a PHP amewo k based on he MVC [ 12 ] a chi ec u e. I p o ides a ious con enience u ili ies such as a empla e engine, a dependency manage o PHP o include hi d-pa y lib a ies in you own p ojec and a command in e ace p og am ha allows you o c ea e p ede ined so wa e a e ac s. This allows he epos- i o y o be ins an ia ed on any web se e wi h a minimum o con igu a ion. La a el se es as he en y poin o he web applica ion and he REST API consump ion managing bo h modules equally bu independen ly so hey can be used sepa a ely. Also, La a el allows us o modula ize [ 8 ] new unc ionali y wi hou needing o e ac o code. FM web applica ion. To in e ac wi h eposi o y ools, we p o ide a web in e ace making he sys em mo e accessible wi hou elying on a command in e ace. In addi ion, he on end allows se e al echnologies o be b ough oge he h ough a con ol panel common o all use s. Figu e 2 illus- a es he web in e ace wi h a lis o models uploaded o he eposi o y, main aining synch onici y wi h Zenodo. Fea u e models a e uploaded using a web o m as shown in Figu e 3. 6h ps://la a el.com/ Figu e 2: Lis ing o models in FMRepo. Figu e 3: Uploading model in FMRepo. FM REST API. FMRepo has a REST ul se ice o p o ide di - e en ope a ions such as consul ing a ea u e model, upda - ing an associa ed ile, o epo ing me ics. The use ulness o his module is o p o ide access and de elopmen o new ools c ea ed in he So wa e P oduc Line amewo k. The eposi o y REST API is e sioned, which allows o inc ease i s unc ionali y wi h new ea u es ha will be included in he sys em in he u u e. 4.2 FMRepo modules This sec ion p esen s he h ee main modules o he a chi ec u e: he Open Science module; he Au oma ed Analysis o Fea u e Models (AAFM) module, and he Da abase module. 4.2.1 Open Science using Zenodo API. To acili a e he in eg a- ion o he Zenodo API wi h he eposi o y, we ha e designed an API w appe . The API w appe is a so wa e a e ac ha wo ks as an in e ace o abs ac he implemen a ion de ails o he se ice, in his case, he Zenodo API . This allows he API o be used as i i was an in e nal eposi o y se ice. The Zenodo API can be ex ended o o he APIs o include mo e Open Science p ojec s. 4.2.2 Au oma ed Analysis o Fea u e Models (AAFM). Mos epos- i o ies wo k only as s a ic s o age. We ha e designed he a chi ec- u e so ha ea u e models a e pos -p ocessed hanks o an in e nal connec ion wi h a Docke ins ance . The AAFM is unning on his machine, which ecei es he use ’s ea u e models and pe o ms a ious analysis ope a ions o ex ac s me ics. Ins ead o using jus one machine, we ha e decided o epli- ca e he AAFM module on mul iple machines. The e o e, a Load Balance dis ibu es he wo k in N ins ances o alle ia e he wo k- load. The pu pose o his module is no only o wo k wi h Docke and AAFM machines unning on hem, bu u u e ea u e model p ocessing echnologies, such as benchma king, can be in eg a ed. The Se ice Poin is in cha ge o p o iding suppo o in eg a - ing new so wa e ha can un in pa allel. 4.2.3 S o age and da abase. The main s o age is delega ed o he Zenodo sys em . To a oid o e -consuming he API, all ea u e models a e s o ed locally once Zenodo has e u ned a posi i e esponse o he upload. The a chi ec u e suppo s di e en ypes o da abases. La a el has an objec - ela ional mapping capable o swi ching be ween di e en da abase echnologys . We ake ad an age o his ea u e o demand di e en se ices. Fo example, he me ada a in o ma ion is s o ed in a non-SQL da abase, in ou case MongoDB, due o he non- ela ional na u e o his in o ma ion. Howe e , he di e en in e nal ansac ions, such as he upload o ea u e models o he esul s gene a ed om he AAFM a e s o ed in a SQL da abase. 5 IMPLEMENTATION DETAILS In his sec ion, we b ie ly desc ibe he implemen a ion de ails o he di e en modules o he FMRepo a chi ec u e. 5.1 Zenodo as abs ac ion laye The o iginal way o wo king wi h Zenodo is h ough a deposi ion. A deposi ion is made up o a se ies o me ada a and a achmen s. To acili a e he implemen a ion o Zenodo in he use o FMRepo, we ha e decided place i as an abs ac ion laye . The impo an hing abou deposi ion is ha hey gene a e a unique and pe manen DOI. Fo p ac ical pu poses, all ea u e models o a use a e uploaded o he same Zenodo accoun o g oup hem o la e usage. FMRepo communica es wi h Zenodo by using i s REST API, which allows he comsump ion o deposi ions and hei a achmen s. Some o he end poin s ha we ha e de ined o consume he Zenodo API a e: • GET /api/deposi /deposi ions. Lis all deposi ions o he cu - en ly au hen ica ed use . • GET /api/deposi /deposi ions/:id/ iles. Lis all deposi ion iles o a gi en deposi ion. • POST /api/deposi /deposi ions. C ea e a new deposi ion e- sou ce. •POST /api/deposi /deposi ions/:id/ iles. Upload a new ile. • POST /api/deposi /deposi ions/:id/ac ions/publish. Publish a deposi ion. Howe e , he Zenodo REST API needs au hen ica ion o be used. To simpli y use expe ience, FMRepo has an in e nal oken s o e ha is au oma ically used o a oid excessi e consump ion o . Zen- odo REST API. 5.2 Running FM analysis Ano he u ili y p o ided by FMRepo is he possibili y o analysing ea u e models uploaded di ec ly om he web applica ion. This comple ely elimina es he need o ins all AAFM so wa e on he use ’s machine. Cu en ly we a e elying on he Py hon amewo k o au oma ed analysis [ 6 ] p esen ed in MODEVAR [ 1 ]. The AAFM ool uns in pa allel on a machine ins an ia ed by Docke 7 o secu i y easons [ 17 ]. We ha e decided o ely on a Docke ins ance o a oid he execu ion o malicious code. This ins ance is capable o analysing ea u e models om a ious web clien s, allowing FMRepo o be unning on se e al se e s and inc easing he demand o esou ces. FMRepo ac s as an in e media y be ween he Docke machine and he uploaded ea u e models so ha end use s ne e ha e access o he in e nal se e s. We ha e also designed a load balance ha allows ha ing mul- iple ins ances o Docke unning. Using a p io i y algo i hm, he eposi o y decides which ins ance has he lowes wo kload and chooses a candida e h oughou he execu ion cycle. To execu e any analysis o e a ea u e model, mos AAFM ools equi e access o iles in a ce ain manne . To sol e his, FMRepo p omo es he use o he same hie a chical ile s uc u e o each upload ha a e willing o p o ide AAFM suppo . This, also gi es us he possibili y o add me ada a o ele an in o ma ion. Then a packaging p ocess is ca ied ou ollowing his s uc u e: • The /model di ec o y s o es he o iginal ile in XML (o simila ) o ma o he ea u e model. • The /me ada a di ec o y s o es he o iginal ile in JSON (o simila ) o ma o he ea u e model me ada a. • The ea u e.json ile (op ional) desc ibes he ea u e model using he isual edi o o he g aphical in e ace ins ead o uploading an XML ile. 5.3 FMRepo API To allow u u e expansions o he sys em, we implemen ed an REST API. FMRepo wo ks as a REST ul se ice o he main unc ionali ies. When a use logs in as a de elope , he sys em p o ides hem wi h a unique access oken. This oken allows consuming he API. Ou API is e sioned and he documen a ion can be consul ed. We we e inspi ed by he s uc u e o Zenodo, acili a ing hus, he usage o he pla o m o use s ha had p e iously wo ked wi h i . The main ac ions ha can be pe o med wi h he API a e: • GET /api/ 1/models. Lis all models o he cu en ly au hen- ica ed use . • GET /api/ 1/models/:model/ iles. Lis all iles o a gi en model. • GET /api/ 1/models/:model/s uc u e. Re ie es he o iginal s uc u e o he uploaded model and i s me ada a. •POST /api/ 1/models. C ea e a new ea u e model. 6 CONCLUSION AND VISION We belie e ha FMRepo is a s a ing poin o c ea e a mo e ac i e communi y in he exchange o knowledge. This new eposi o y is mo e aligned wi h cu en so wa e echnologies and in line wi h he MODEVAR ini ia i es. FMRepo also in eg a es au oma ed anal- ysis being able o use he echnology ha uns in he backg ound 7h ps://www.docke .com/ and encou aging i s use i mo e ad anced ope a ions a e desi ed. The eposi o y a emp s o p omo e he communi y’s use o Open Science and i s ele ance o esea ch. The a chi ec u e o he epos- i o y is modula , which will allow o ex end he unc ionali y in a simple and collabo a i e way in he u u e. Ou ision o he eposi o y is o p o ide a use ul ool o he s udy o ea u e models in e ms o me a-managemen , analysis and sha ing. We appeal o all esea che s and encou age he use o his eposi o y. We belie e ha communi y suppo will a ou he de elopmen , expansion and suppo o u u e cha ac e is ics. To a oid an ou da ed, abandoned and dep eca ed eposi o y, we u ge he communi y o upload hei own ea u e models ha co e all possible scena ios and s udy hei pe o mance as an analysis ool. We belie e ha his will encou age he use o he eposi o y, becoming a p oduc buil on he Open Science guidelines. ACKNOWLEDGEMENTS This wo k was suppo ed by he P ojec (RTI2018-101204-B-C22, OPHELIA), unded by: FEDER/Minis y o Science and Inno a ion - S a e Resea ch Agency; he TASOVA ne wo k (MCIU-AEI TIN2017-90644-REDT); and he Jun a de Andalucia COPERNICA p ojec ; and he Spanish Go e nmen unde Juan de la Cie a—Fo mación 2019 g an . REFERENCES [1] Ma hieu Ache , Philippe Colle , Da id Bena ides, and Rick Rabise . 2020. Thi d In e na ional Wo kshop on Languages o Modelling Va iabili y (MODE- VAR@SPLC 2020). 1–1. h ps://doi.o g/10.1145/3382025.3414948 [2] Eb ahim Baghe i and D agan Gase ic. 2011. Assessing he main ainabili y o so wa e p oduc line ea u e models using s uc u al me ics. So wa e Quali y Jou nal 19, 3 (2011), 579–612. h ps://doi.o g/10.1007/s11219-010-9127-2 [3] Da id Bena ides, Se gio Segu a, and An onio Ruiz-Co és. 2010. Au oma ed analysis o ea u e models 20 yea s la e : A li e a u e e iew. In o ma ion Sys ems 35 (09 2010), 615–636. h ps://doi.o g/10.1016/j.is.2010.01.001 [4] Isabel Be nal. 2018. Open Science: P incipios, polí icas, componen es, buenas p ác icas. (11 2018). h ps://doi.o g/10.20350/12794 [5] José A. Galindo and Da id Bena ides. 2019. Towa ds a New Reposi o y o Fea u e Model Exchange. In P oceedings o he 23 d In e na ional Sys ems and So wa e P oduc Line Con e ence - Volume B (Pa is, F ance) (SPLC ’19). Associa ion o Compu ing Machine y, New Yo k, NY, USA, 170–173. h ps://doi.o g/10. 1145/3307630.3342405 [6] José A. Galindo and Da id Bena ides. 2020. A Py hon amewo k o he au o- ma ed analysis o ea u e models: A i s s ep o in eg a e communi y e o s. In SPLC ’20: 24 h ACM In e na ional Sys ems and So wa e P oduc Line Con- e ence, Mon eal, Quebec, Canada, Oc obe 19-23, 2020, Volume B. ACM, 52–55. h ps://doi.o g/10.1145/3382026.3425773 [7] José Angel Galindo, Da id Bena ides, Pablo T inidad, An onio Manuel Gu ié ez- Fe nández, and An onio Ruiz-Co és. 2019. Au oma ed analysis o ea u e mod- els: Quo adis? Compu ing 101, 5 (2019), 387–433. h ps://doi.o g/10.1007/ s00607-018-0646-1 [8] Jesse G i in. 2021. Modula izing La a el. 237–284. h ps://doi.o g/10.1007/ 978-1-4842-6023-4_8 [9] José-Miguel Ho cas, Mónica Pin o, and Lidia Fuen es. 2019. So wa e p oduc line enginee ing: a p ac ical expe ience. In P oceedings o he 23 d In e na ional Sys ems and So wa e P oduc Line Con e ence, SPLC 2019, Volume A, Pa is, F ance, Sep embe 9-13, 2019. ACM, 25:1–25:13. h ps://doi.o g/10.1145/3336294.3336304 [10] Jason C. Hung and Chun-Chia Wang. 2021. Explo ing he websi e objec layou o esponsi e web design: esul s o eye acking e alua ions. J. Supe compu . 77, 1 (2021), 343–365. h ps://doi.o g/10.1007/s11227-020-03283-1 [11] Sameeksha Ka och, Kowshik Thopalli, Jaya aman J. Thiaga ajan, Pa an K. Tu aga, and And eas Spanias. 2019. In enio: Disco e ing Hidden Rela ionships Be ween Tasks/Domains Using S uc u ed Me a Lea ning. CoRR abs/1911.10600 (2019). a Xi :1911.10600 h p://a xi .o g/abs/1911.10600 [12] Glenn E K asne , S ephen T Pope, e al . 1988. A desc ip ion o he model- iew- con olle use in e ace pa adigm in he small alk-80 sys em. Jou nal o objec o ien ed p og amming 1, 3 (1988), 26–49. [13] Jabie Ma inez, Wesley K. G. Assunção, and Tew ik Ziadi. 2017. ESPLA: A Ca alog o Ex ac i e SPL Adop ion Case S udies. In P oceedings o he 21s In e na ional Sys ems and So wa e P oduc Line Con e ence, SPLC 2017, Volume B, Se illa, Spain, Sep embe 25-29, 2017. ACM, 38–41. h ps://doi.o g/10.1145/3109729.3109748 [14] Ma cílio Mendonça, Moises B anco, and Donald Cowan. 2009. S.P.L.O.T. - So wa e p oduc lines online ools. P oceedings o he Con e ence on Objec - O ien ed P og amming Sys ems, Languages, and Applica ions, OOPSLA, 761–762. h ps://doi.o g/10.1145/1639950.1640002 [15] Mahdi Noo ian, Ali eza Ensan, Eb ahim Baghe i, Ha old Boley, and Ye gen Bile skiy. 2011. Fea u e model debugging based on desc ip ion logic easoning. Collec ion / Collec ion : NRC Publica ions A chi e / A chi es des publica ions du CNRC. [16] Rahul Ramachand an, Kaylin Bugbee, and Ke in Mu phy. 2021. F om Open Da a o Open Science. Ea h and Space Science 8 (05 2021). h ps://doi.o g/10.1029/ 2020EA001562 [17] Edwin Sa mien o. 2020. Docke Images and Con aine s. 69–98. h ps://doi.o g/ 10.1007/978-1-4842-5826-2_5 [18] Se gio Segu a, Da id Bena ides, An onio Ruiz-Co és, and Pablo T inidad. 2007. Au oma ed Me ging o Fea u e Models Using G aph T ans o ma ions. 489–505. h ps://doi.o g/10.1007/978-3-540-88643-3_15 [19] Se gio Segu a, Ana B. Sánchez, and An onio Ruiz-Co és. 2014. Au oma ed Va i- abili y Analysis and Tes ing o an E-Comme ce Si e. An Expe ience Repo . ASE 2014 - P oceedings o he 29 h ACM/IEEE In e na ional Con e ence on Au oma ed So wa e Enginee ing. h ps://doi.o g/10.1145/2642937.2642939 [20] Pe iklis Sochos, Ilka Philippow, and Ma hias Riebisch. 2004. Fea u e-O ien ed De elopmen o So wa e P oduc Lines: Mapping Fea u e Models o he A - chi ec u e. In Objec -O ien ed and In e ne -Based Technologies, Ma hias Weske and Pe e Liggesmeye (Eds.). Sp inge Be lin Heidelbe g, Be lin, Heidelbe g, 138–152. [21] Geo ey I. Webb and Ma k Kuzmycz. 1995. Fea u e Based Modelling: A me hodol- ogy o p oducing cohe en , consis en , dynamically changing models o agen s’ compe encies. Use Modeling and Use -Adap ed In e ac ion 5, 2 (01 Jun 1995), 117–150. h ps://doi.o g/10.1007/BF01099758