Jou nal o Web Enginee ing, Vol. 17, No.3&4 (2018) 183-205
© Au ho s
CHALLENGES FOR THE ADOPTION OF MODEL-DRIVEN WEB ENGINEERING
APPROACHES IN INDUSTRY
ESTEBAN ROBLES LUNA
Resea ch and T aining in Ad anced Compu ing Labo a o y (LIFIA), Na ional Uni e si y o La Pla a,
La Pla a (A gen ina)
es eban[email p o ec ed].edu.a
JUAN MIGUEL SÁNCHEZ BEGINES
Web Enginee ing and Ea ly Tes ing G oup (IWT2), Uni e si y o Se ille, Se ille (Spain)
juan.[email p o ec ed]
JOSÉ MATÍAS RIVERO
Resea ch and T aining in Ad anced Compu ing Labo a o y (LIFIA), Na ional Uni e si y o La Pla a,
La Pla a (A gen ina)
m i e o@li ia.in o.unlp.edu.a
LETICIA MORALES
Web Enginee ing and Ea ly Tes ing G oup (IWT2), Uni e si y o Se ille, Se ille (Spain)
le icia.mo[email p o ec ed]
J.G. ENRÍQUEZ
Compu e Languages and Sys ems Depa men . Uni e si y o Se ille. A . Reina Me cedes s/n, 41012,
Se ille, Se ille.
jose.gonza[email p o ec ed]
GUSTAVO ROSSI
Resea ch and T aining in Ad anced Compu ing Labo a o y (LIFIA), Na ional Uni e si y o , La Pla a,
La Pla a (A gen ina)
gu[email p o ec ed].a
Recei ed July 27, 2017
Re ised Ap il 23, 2018
Model-D i en Web Enginee ing app oaches ha e become an a ac i e esea ch and echnology solu ion o
Web applica ion de elopmen . Howe e , o mo e han 20 yea s o de elopmen , he indus y has no
adop ed hem due o he misma ch be ween echnical e sus esea ch equi emen s. In he con ex o his
join wo k be ween academia and indus y, he au ho s conduc a su ey among hund eds o enginee s om
di e en companies a ound he wo ld and, by s a is ical analysis, hey p esen he cu en p oblems o hese
app oaches in scale. Then, a se o guidelines is p o ided o imp o e Model-D i en Web Enginee ing
app oaches in o de o make hem iable indus y solu ions.
Key wo ds: Model-D i en Web Enginee ing, Human-Cen e ed in e aces, MDE, Indus y
184 Challenges o he Adop ion o Model-D i en Web Enginee ing App oaches in Indus y
1 In oduc ion
Model-D i en Web Enginee ing (MDWE) app oaches appea ed 20 yea s ago o ul ill a missing a ea o
Model-D i en De elopmen : Web Applica ion De elopmen [1]. F om ha momen , abou 7 o 8
MDWE solu ions [2] ha e been c ea ed, bu only a ew o hem ha e ended up p o iding ool suppo .
Also, only one has become he mayo playe wi h small company and suppo om Objec Managemen
G oup (OMG) o con e i s language in o a s anda d.
Al hough MDWEs app oaches ha e claimed o imp o e mul iple aspec s o Web Applica ion
De elopmen such as code quali y, de elopmen speed and le el o abs ac ion, we ha e o mula ed he
ollowing main esea ch ques ion: Why has he indus y paid li le a en ion o MDWE?
A ecen s udy [3] has p esen ed he 20 obs acles ha dis u b Web applica ion scalabili y and, hough
hey a e no speci ically a ge ed o he applica ions de i ed om MDWE app oaches, hey a e s ill
a ec ed by hese applica ions.
This s udy p esen s a lis o issues ha hinde s he use o MDWE app oaches in medium o big size
companies. As a consequence, i shows p ecise p ac ical p oblems ha need o be sol ed o suppo he
claims ha he MDWE communi y has made o yea s. Figu e 1 p esen s an agile de elopmen li ecycle
p ocess o a Web applica ion in o de o p o ide a con ex o he issues ha ha e been ound. In his
igu e, he main phases appea in bold ype o ma . Simila ly, he issues ha complica e he usage o
MDWE app oach a e ep esen ed by “pos -i no es” linked o he phase in which each o hem ha e been
loca ed [4].
Figu e 1 Issues ha hinde he usage o he MDWE app oach in medium o big size companies
Es eban Robles Luna, J.M. Sánchez-Begines, Ma ías Ri e o, L. Mo ales-T ujillo, J.G. En íquez, G. Rossi 185
The s udy uses small expe imen s o un a MDWE app oach in a company, in addi ion o an
exhaus i e li e a u e e iew o gi e suppo o i s needs. As many di e en aspec s a e conside ed (no
only echnical) when dealing wi h limi a ions o MDWE app oaches, we ha e c ea ed a ca ego iza ion
o s ess he a ea whe e he issue is loca ed:
● Social [S]: A social issue is ela ed o p oblems among people in ol ed in he p ojec o in
he so wa e a i ac .
● Technical [T]: A echnical issue is associa ed wi h he so wa e elemen s ha cons i u e he
Web applica ion.
● Economical [E]: An economical issue is linked o he p ojec ’s budge o expenses ha he
de elopmen o he applica ion en ails.
Table 1 shows he lis o issues unde hese h ee ca ego ies ha will be p esen ed in he ollowing
subsec ions. I has been ob ained om he in e iews we conduc ed wi h he enginee s. We ha e decided
o include no only echnical aspec s (as all MDWE app oaches ake in o accoun ), bu also social and
economic ones. Mode n Web Enginee ing app oaches in indus y a e a ec ed o a leas 3 o hem.
Each subsec ion explains he conce ns and o e s a inal guideline o sol e hem.
Social
Technical
Economical
Lack o con ol in de elopmen and deploymen
X
X
X
Too ocused in na iga ion
X
Me amodel
suppo and adap a ion o i s ools
X
X
T aceabili y and debugabili y
X
Lack o ools
X
Rigid a chi ec u e
X
Technological aspec s
X
Communi y
X
X
Licensing
X
Table 1. Lis o issues by ca ego y
The expe imen is based on a su ey ha e alua es bo h he cha ac e is ics o he companies
pa icipa ing in he p ojec and he guide o sol e he iden i ied p oblems. This su ey ies o add ess all
he issues om he social, echnical and economic poin o iew.
The aim o he su ey is o ex ac help ul in o ma ion om hese companies in o de o alida e o
e u e ou hypo heses in some way. The esul s ob ained will be conside ed eliable, since all he
companies ha ha e pa icipa ed in his expe imen a e o a ele an size and epu a ion in he so wa e
wo ld. Each o he companies has p o ided specialis s in he ield such as p ojec manage s, analys s o
senio p og amme s who a e amilia wi h he echnical p oblems hey ackle.
A e he execu ion o he expe imen , we will s a e whe he he hypo heses a e alid, ha is, whe he
hey sol e he p oblem o each o he phases o Web de elopmen , o in con as , i is necessa y o
e o mula e a new solu ion guide. This can open he ield o MDWE app oaches o achie e a amewo k
o s anda d ha may co e he needs ha companies demand om his me hodology and his way, he
indus y will pay mo e a en ion o hem.
In his pape , we will i s in oduce he MDWE app oach and i s p oblems o be used. In sec ion 2,
we will p esen he wo k ela ed o hese p oblems. In sec ion 3, we will desc ibe he expe imen a ion
phase and in sec ion 4 we will p opose a se ies o challenges o sol e some o he p oblems iden i ied
186 Challenges o he Adop ion o Model-D i en Web Enginee ing App oaches in Indus y
oge he wi h a su ey. Nex , da a ob ained in he su ey will be analyzed and alida ed in sec ions 5
and 6, espec i ely. To end up, sec ion 7 will s a e a se o conclusions and i will ad ance u u e lines
o wo k.
2 Rela ed Wo k
Since he beginning o he 21s cen u y, he so wa e communi y has widely accep ed modeling
languages o Web Applica ions De elopmen . The MDWE app oach has ied o inco po a e hese
languages in o he whole de elopmen li ecycle in hei se e al s ages.
F om he onse o he MDWE app oach, se e al solu ions ha e eme ged ha ha e a emp ed o
suppo di e en Web echnologies and a chi ec u es (such as Rich In e ne Applica ion (RIA), Seman ic
Web o Mobile Web, among o he s) and ha e ied o imp o e he co e age o he whole de elopmen
li ecycle o applica ions acco ding o he speci ica ion o equi emen s, es ing and main enance [5].
Va ious solu ions ha e been p oposed conce ning he nega i e aspec o ha ing a my iad o design
app oaches and anno a ions in he MDWE. The i s one is o uni y he di e en me hods by ex ac ing
he bes pa s om each o hem [7]. The second p oposes o make all app oaches in e ope a e in one
way o ano he by allowing be e po abili y among design models and acili a ing hei connec ions [8].
The easibili y o c ea ing a s anda d has been s udied and inally he OMG has adop ed he In e ac ion
Flow Modeling Language (IFML) s anda d as a solu ion o ha p oblem.
Ano he impo an aspec ha he MDWE app oach unco e s is o p o ide suppo ools and
echniques ha would play an impo an ole in simpli ying he design, de elopmen and e olu ion
phases o applica ions. Compu e -Aided So wa e Enginee ing (CASE) ools play a key ole o so wa e
enginee s o inc ease p oduc i i y. The e o e, hese ools a e e y signi ican when adop ing he MDWE
app oach, as hey will ha e a g ea impac on he imp o emen o so wa e indus y ha eeds on such
ools [5].
The impac o MDWE app oaches on indus y has no been su icien ly s udied, al hough we can
a gue ha he applica ion o Model-D i en So wa e Enginee ing (MDSE) is widely adop ed in a ious
domains and he e is a la ge numbe o ools ha ha e al eady assimila ed his pa adigm.
MDSE bases i s de elopmen on speci ic domain models a he han on mo e gene al and
s anda dized languages like Uni ied Modeling Language (UML) [9,10]. The e a e many nega i e aspec s
ound in companies when adop ing MDSE as, o ins ance, he e o made in aining people o use
MDSE echniques. As we will analyze la e , ools no suppo ing he pa adigm p ope ly, among o he
aspec s such as imma u i y, complexi y and usabili y, p o oke inc eases in cos s when applying MDSE
me hodologies [11].
Ano he poin o deba e will be he need o acqui e p op ie a y so wa e o o he wise o c ea e a
la ge communi y ha de elops ee ools o suppo MDWE. Eclipse Modeling Tools sui e is he p emie
open sou ce echnology o deploying MDE used by IBM and O acle. The e a e o he applica ions, like
Je B ains Me a P og amming Sys em, ha a e p op ie a y echnologies. In con as , ools can be di e en
depending on he modeling language: some o hem use UML language and some o he s a speci ic one.
Finally, i is wo h men ioning he e o ha academic communi ies ha e made in o de o de elop
ools o suppo MDWE pa adigms such as NDT-Sui e [12], QuEF [13] and WebRa io [14]. NDT-Sui e,
o ins ance, was submi ed o a su ey in di e en companies whe e i ob ained sa is ac o y esul s in
e ms o unc ionali y, usabili y and u ili y [15].
Ano he issue ha we will add ess in he su ey is he impo ance o ha ing a b oad communi y ha
assis s and compiles in o ma ion o in e es o a be e implemen a ion o he MDWE pa adigm. The
Es eban Robles Luna, J.M. Sánchez-Begines, Ma ías Ri e o, L. Mo ales-T ujillo, J.G. En íquez, G. Rossi 187
MDWE communi y g ew a a apid pace hanks o he i s Web Enginee ing wo kshop held a he Wo ld
Wide Web (WWW) con e ence in 1998 ha ocused on Web-based modeling echniques. This enhanced
a g ea e unde s anding o he communi y o he p oblems in ol ed in he de elopmen d i en by Web
applica ion models, hus o ming a la ge olume o knowledge ha was published la e in di e en
places. This laid an excellen ounda ion in he MDWE pa adigm o p o essionals in he a ea [5].
All hese MDWE issues will be add essed in his pape om he pe spec i e o use s in companies
ha will e alua e hem acco ding o hei c i e ia. A se ies o solu ions o imp o emen s ela ed o he
a o emen ioned a gumen s will be p oposed and alida ed based on a su ey ca ied ou by use s.
3 Expe imen a ion
I is necessa y o know he poin s o iew o di e en p o iles ela ed o he objec o analysis so ha he
MDWE app oach can be e alua ed. Tha is why we ha e conduc ed his s udy h ough a su ey.
In So wa e Enginee ing (SE), su eys a e one o he mos widely used esea ch me hods o ca y
ou empi ical esea ch s udies. The SE esea ch di e s om o he disciplines, pa icula ly wi h espec
o i s objec i es and opics. Objec i es a e o en associa ed ei he wi h imp o ing he de elopmen
p ocess o explo ing new echniques and me hods. The subjec s o such s udy usually in ol e well-
quali ied p o essionals wi h clea opinions and expe ience in hei a eas o in e es [16].
Following his me hod, we ha e in ended o collec and assess e idence o a sample o in e es in
he opic o deal wi h. Taking hem as a basis, we will alida e i he solu ions p o ided coincide wi h
he e alua ions o he pa icipan s and, his way, hey will co obo a e ou con ibu ion o he MDWE
pa adigm.
We ha e decided o apply his expe imen a ion echnique due o he numbe o use s pa icipa ing
in i and hus, ano he mo e open me hod will esul in a long and expensi e s udy. Thanks he possibili y
o doing i on he In e ne , we ha e p e e ed his me hod a he han some mo e complex and du able
[17] ones.
4 Challenges
4.1 Lack o con ol in de elopmen and deploymen [S, T, E]
C ea ing a Web applica ion is a complex p ocess ha in ol es no only coding/modeling, bu also ha ing
mee ings wi h s akeholde s and debugging o ix p oduc ion p oblems, o ins ance. In pa icula , when
he applica ion is deployed, aspec s such as moni o ing, logging and p o iling become mo e impo an
[3]. The e o e, he enginee ing eams ha de elop Web applica ions mus ha e he owne ship and
esponsibili y o he deli e ables. Consequen ly, hey need o con ol he comple e p ocess ( om
de elopmen o deploymen ).
Despi e he bene i s o model-d i en echnologies, hey also add an ex a le el o complexi y as he
de i a ion p ocess is o en seen as a “magic wand” ha ob ains an applica ion om a se o models. Mos
MDWE ools hide and make his p ocess “close sou ce” c ea ing a dependency be ween he MDWE ool
and de elopmen eams. This dependency does no ma ch ime cons ain s and he ac o no ha ing he
sou ce code a ailable o e e yone will make hings e en wo s . Addi ionally, unless he MDWE ooling
is open sou ce and i has a big communi y, which ha dly e e happens, i is likely ha he commi men
om he MDWE ooling eam will equi e paying ees in he o m o licenses, wha inc eases he cos
o de elopmen .

188 Challenges o he Adop ion o Model-D i en Web Enginee ing App oaches in Indus y
4.1.1 Guideline
MDWE ools need o p o ide he ollowing aspec s in o de o deeply con ol he ac ual
de elopmen :
● Ways o hook in he modeling and ans o ma ion phases, so ha de elope s can add and
imp o e he de elopmen p ocess based on he applica ion hey a e building. This could be done
i he ans o ma ion algo i hm wo ks using he empla e me hod design pa e n.
● Make he code open sou ce, so ha he de elopmen eam will debug he ool and he gene a ed
applica ions easily in case ha hese p oblems a ise du ing he de elopmen o main enance
p ocess.
4.2 Too ocused in na iga ion [T]
O iginal hype media-based Web applica ions ha we e de eloped 10 yea s ago a e a he di e en om
he cu en in eg a ion Web pa adigm, whe e na iga ion is one iny conce n. Aspec s such as Rich
In e ne Applica ions (RIA), in eg a ion wi h di e en sys ems, sea ch capabili ies, pe sonaliza ion and
ecommenda ion ha e become mo e complex and ele an poin s han na iga ion i sel .
In he ea ly s ages, MDWE app oaches we e c ea ed o adap he na iga ion pa adigm om
hype media o he Web. Many new concep s we e in oduced and al hough mos o hese enhancemen s
ha e been epo ed in he li e a u e [18], only a ew o hem ha e been ac ually implemen ed. The e o e,
oday he main model o indus y leading WebRa io ool is he na iga ional model ha desc ibes he
na iga ional pa hs ha a use can ollow. Opposi e o wha many esea ches ha e shown abou he
impo ance o Web applica ion aspec s (e.g. in eg a ion), na iga ion is s ill he mos impo an one o
MDWE app oaches.
4.2.1 Guideline
MDWE app oaches need o de ach om being ocused on na iga ion o suppo i as one conce n o
many p esen in he Web applica ion li ecycle and a chi ec u e. Fu he mo e, he e is a special
equi emen o p o ide a eal ool suppo o he ea u es ha MDWE app oaches claim o ha e and, in
case ha hey a e no suppo ed, p o ide hooks o pe o m manual coding o hese ea u es.
4.3 Me amodel suppo and adap a ion o i s ools [T, S]
The inc easing amoun o echnologies ha a e being de eloped e e y day in addi ion o cus ome ’s ime
cons ain s pose mul iple challenges o Web applica ion de elopmen . Al hough hese asks can be
pe o med manually (in code-based en i onmen s) and may be ime consuming, hey can be easily
pe o med by ex ending hem in o he so-called “ho spo s” ha code-based amewo ks p o ide.
In MDWE app oaches, his kind o ex ensions may be ei he ca ied ou by ins an ia ing some p e-
exis ing me amodel classes o , i he unc ionali y is no suppo ed, by me amodel ex ensions. Ex ending
a me amodel does no only equi e o add new classes and ans o ma ions, bu also o adap ools in a
imely manne . Some app oaches, such as NDT [19] o UWE [20], a e ex ensions o he UML p o ile,
hus, hese app oaches can be adap ed mo e easily han o he s, such as IFML [21].
Es eban Robles Luna, J.M. Sánchez-Begines, Ma ías Ri e o, L. Mo ales-T ujillo, J.G. En íquez, G. Rossi 189
4.3.1 Guideline
Due o he exis ing echnology and economic ime cons ain s, me amodels and ools need o be adap ed
in a ew days o cope wi h cus ome s’ demands. Cu en ly, changing me amodels and ools can ake a
ew weeks and, hus, hey canno be accep ed by he cus ome ’s imeline, specially in he con ex o use
o Agile me hodologies.
4.4 T aceabili y and debugabili y [T]
One o he mos impo an aspec s o so wa e de elopmen is he abili y o in ospec , change and
moni o he “li e” applica ion unde de elopmen o quickly ix he p oblems; hese ac ions a e
conside ed as “debugging” an applica ion. Code-based de elopmen in o he languages such as
Small alk, Ja a and .NET has hese ea u es om he beginning, making easy o i e a e hem along he
de elopmen p ocess.
In MDWE ools, whe e code is gene a ed om models, he abili y o debug is ela ed o he
de i a ion o aceabili y links be ween models and he gene a ed code. Nowadays, only WebRa io
pa ially suppo s his schema by allowing debugging he applica ion unde de elopmen [22].
None heless, WebRa io needs u he wo k o help ace back he p oblems while he applica ion is
unning in p oduc ion. Un o una ely, excep ions occu as no suppo is p o ided in his case. All hese
aspec s make co e enginee s a oid adop ing MDWE ools as hey lack con ol o e he sys em unde
de elopmen .
4.4.1 Guideline
As high-le el languages (e.g. Ja a) p o ide ways o ace back p oblems o concep s o he language
(e.g. classes and line numbe s in excep ion s ack aces), he MDWE app oach should p o ide hose
ea u es in o de o de ec he oo causes o he p oblems. Addi ionally, u he debugging suppo mus
be gi en o debug, e alua e and al e he applica ion while i is unning in a de elopmen en i onmen .
4.5 Lack o ools [T]
Building a Web applica ion equi es a se o ools ha eases he p ocess o de elopmen , deploymen
and moni o ing. Fo example, a ypical JEE Web applica ion can be de eloped using Ma en [23] as a
build sys em, Jenkins [24] o con inuous in eg a ion and a a ie y o amewo ks o ac ually build and
implemen he applica ion (Sp ing, Hibe na e o Angula JS, among o he s). A se o ools ha au oma e
and con ol he p ocess om he momen he applica ion is buil o he ins an ia ion o he se e s and
applica ion deploymen , as well as i s ini ial moni o ing is p o ided in o de o pe o m he ac ual
deploymen . All hese ools, despi e coming om a lowe le el o abs ac ion (i we compa e i wi h he
MDWE app oach), clea ly help cons uc and deploy an applica ion.
Howe e , he ac ual de elopmen p ocess o applica ions om high-le el models has been
speci ically poin ed ou in he MDWE a ea. Al hough his is co ec in e ms o he MDWE philosophy,
i inc eases he e o made in de elopmen and moni o ing o many o he opics discussed in his wo k
( echnological ea u es, aceabili y, and debugabili y). Below, we p o ide conc e e examples ha
illus a e his poin :
● The e a e no ools o link s ack ace excep ions back o he model elemen s. This aspec
complica es he ask o co ec ing e o s o igina ed in he applica ion deployed.
● The e a e no ools o suppo he moni o ing p obes o he model elemen s and he e o e,
de ec ing pe o mance issues is also a di icul ac i i y. By using moni o ing echnologies like
190 Challenges o he Adop ion o Model-D i en Web Enginee ing App oaches in Indus y
New Relic [25], we can iden i y some o hese poin s, bu hese would be classes de i ed om
he MDWE ool ha may co espond o mul iple model elemen s.
4.5.1 Guideline
MDWE app oaches need o suppo a hand ul se o ools ha help in he comple e applica ion li ecycle,
in o de o p o ide alid app oach o indus y. Some o he a o emen ioned ools (e.g. Ma en, Jenkins
o New Relic) a e ully ex ensible, so building some o hem can be ai ly simple by ha ing good
aceabili y links and ex ending hem wi h he igh in o ma ion. Tools a ound Web applica ions a e one
o he mos c i ical aspec s o keep he applica ion unning 24x7.
4.6 Rigid a chi ec u e [T]
C ea ing a Web applica ion o any size may equi e small o big changes in a s anda d 3- ie Web
applica ion. Some elemen s ha need o be conside ed may in ol e in eg a ion wi h ex e nal se ices,
p ocessing queued in o ma ion in an asynch onous manne , exposu e o Rep esen a ional S a e T ans e
(REST) se ices o ex e nal use s o in e nal mobile applica ions, among o he s issues. Hence, being
able o adap a chi ec u e o suppo any o hese ypes o equi emen s is ex emely impo an .
In cu en MDWE ools, a chi ec u e is no modeled a all and, as a consequence, hey de i e a
simplis ic 1- ie Web applica ion [26] ha can only handle a ew se s o use cases and does no allow he
de elopmen eam o be able o adap o u u e needs. A ecen pape [27] has shown he necessi y o
model hese aspec s in some way, so ha hey can be conside ed h ough he de i a ion p ocess. In i s
ac ual s a e, MDWE ools can lead o simple applica ions ha may no scale p ope ly, hus making i
ha de o be adop ed in medium o big size companies.
4.6.1 Guideline
Adap abili y o mo e complex equi emen s ha may in ol e unc ional (e.g. p ocessing o line da a) o
non- unc ional (e.g. pe o mance and scalabili y issues) equi emen s en ails ha he MDWE app oach
needs o model a chi ec u e in such a way ha de elopmen eams can decide which app oach should be
used. We mus s ess ha al hough some a chi ec u es (e.g. 3- ie Web app) can be p e-con igu ed, i is
c ucial o model a chi ec u e p imi i es and le de elopmen eams abs ac highe -le el concep s om
hem, such as he 3- ie Web applica ion, ins ead o ha dcoding i .
4.7 Technological aspec s [T]
Logging, caching, load balancing and p o iling a e some o he aspec s ha enginee s need o build high-
scalable Web applica ions [3]. The lack o any o hem poses some limi a ions on he ype o applica ion
ha can be p oduced. Fo ins ance, he absence o a caching s a egy o ces he applica ion o compu e
o e ch in o ma ion o e e y eques , hus limi ing i s g ow h. Addi ionally, i may add he ollowing
p oblems:
● Run ou o DB connec ions: No being able o cache in o ma ion s o ed in a DB equi es he
usage o a DB connec ion o e e y eques . Thus, because o a DB limi a ion issue, he
maximum numbe o use s able o access he Web applica ion is equal o he numbe o DB
connec ions and as a esul , new use s will no be able o access he DB.
Es eban Robles Luna, J.M. Sánchez-Begines, Ma ías Ri e o, L. Mo ales-T ujillo, J.G. En íquez, G. Rossi 191
● Inc easing esponse ime: I we canno s o e ex e nal se ice calls, hey need o happen e e y
ime, hus inc easing he o e all esponse ime o he eques ed Web page.
● Inc ease in ha dwa e needed: I no cache is p o ided, we may need o use mo e ha dwa e o e-
compu e alues ha we e compu ed be o e.
E en hough caching is a ai ly simple aspec in insic o applica ion de elopmen , MDWE
app oaches conside i oge he wi h he a o emen ioned aspec s as a “ echnological” elemen . Being
pa o his ca ego y means ha li le impo ance has being paid o model, hus enginee s will ha e o
weak hem in he gene a ed code. As none o he MDWE ools p o ide a ound ip be ween he
gene a ed code and models, hese “ echnological” weaks ha e o be adjus ed e e y ime he applica ion
is de i ed.
4.7.1 Guideline
“Technological” aspec s need o be conside ed in some way inside he model-d i en de elopmen . I he
MDWE app oach in ends o conside and model hem, i will p o ide a g ea bene i o he size and
quali y o he applica ion ha can be buil wi h MDWE ools. A he same ime, he mo e esponse ime
is educed, bigge amoun o wo k is handled by he same amoun o ha dwa e, wha will clea ly show
he bene i s o using a model-based solu ion.
4.8 Communi y [S, E]
In he MDWE esea ch a ea, he e is a good ini ia i e like he MDWEne [28], whose main esea ch
ocus a e me a-modeling and model ans o ma ions. I was c ea ed wi h he aims o imp o ing
in e ope abili y be ween exis ing MDWE app oaches and hei ools, so as o o e be e me hods and
solu ions o indus y. Today, mos app oaches ha e lo s o aspec s in which hey di e [29]: di e ences
be ween me a-models and models, unlike ways o implemen ans o ma ions o di e se ools and used
echnology, among o he hings. Then, he e is a lack o consensus and documen a ion among app oach
designe s and his en i e con ex is causing di e en si ua ions:
● On he one hand, o ganiza ions do no know how hey can ake ad an age o hese app oaches
and how hey can be op imized in hei pa icula con ex , due o he di e si y se o
cha ac e is ics hey o e and he global he e ogenei y associa ed wi h speci ic aspec s o ideas
p ocessed by each app oach.
● On he o he hand, unde his si ua ion i is e y complica ed o app oach designe s o iden i y
he eal equi emen s and demands o he o ganiza ion in o de o imp o e hei app oaches o
design new ones.
Ne e heless, wi hin he con ex o he MDWE app oach, WebRa io is an excep ion. This ool
suppo p o ides a big communi y wi h a big a ie y o u o ials, webina s, use guides, assis ance, o a
and di e en ypes o ool ce i ica ions o use s. Nowadays, he e is no doub ha WebRa io is he
leade ool in he ma ke wi hin he con ex o MDWE app oaches, al hough nei he his ool no his
communi y can be compa ed o o he communi ies in he wo ld ela ed o Web de elopmen .
Unlike MDWE app oach, he e a e exis ing and e y ex ended Web de elopmen amewo ks
a ound he wo ld like Ruby on Rails [30], Django [31], G ails [32] o Codeigni e [33], among o he s.
All o hem in ol e big communi ies o de elope s ha p o ide lo s o documen a ion, u o ials, use
guides, o a and assis ance, o ins ance. Mo eo e , hese amewo ks sha e many ea u es and
componen s ha le de elope s compa e hem.
198 Challenges o he Adop ion o Model-D i en Web Enginee ing App oaches in Indus y
Figu e 7: How impo an is o you o de ine he design o you Web applica ion?
As we can obse e in Figu e 7, he se comp ised in 92.1% o he sample conside s ha i is key o
know and con ol absolu ely Web applica ion a chi ec u e, highligh ing i as he mos chosen and
ele an op ion.
The nex ques ion o he pa icipan s is o o de a se ies o phases and ac i i ies acco ding o he
e o hey need o be add essed. The indica o s ange om ac i i ies o phases ha equi e g ea e e o ,
gi en alue 1, o ac i i ies o phases ha demand less e o , gi en alue 6.
Figu e 8: Please, ank he ollowing aspec s acco ding o he e o made on add essing each de elopmen i e a ion
Fi s , we ha e analyzed he e o made in debugging memo y o pe o mance p oblems. Figu e 8
shows ha we a e acing an ac i i y ha needs a g ea e o o pa icipan s o be add essed, since he
highes pe cen age o esul s is ound in he i s indica o s.
Second, ano he ac i i y o be e alua ed is he implemen a ion o caching s a egies. The as
majo i y o he esul s a e concen a ed in he lowes indica o s, wha poin s ou ha his ask is no oo
expensi e o use s.
Thi d, we p esen he e o ha main aining he applica ion supposes o pa icipan s. Figu e 8
con i ms ha he esul s a e concen a ed in he in e media e indica o s, whe e indica o 4 poin s ou ha
he e o o his ask is no ex emely expensi e.

Es eban Robles Luna, J.M. Sánchez-Begines, Ma ías Ri e o, L. Mo ales-T ujillo, J.G. En íquez, G. Rossi 199
Fou h, we show now he e o made in he de elopmen o he applica ion. The esul s o his cha
a e equi able, highligh ing indica o 5 as he highes alue, ollowed by indica o 1, as he second mos
selec ed alue.
Fi h, ano he impo an aspec o look a is he imp o emen o scalabili y in he applica ion. The
esul s a e clea ly loca ed in he middle o he cha , wha is a symp om ha o pa icipan s i is a ask
ha in ol es medium-high e o .
Following he p esen a ion o esul s ob ained in he su eys, we desc ibe he e o ha he
pa icipan s es ima e he p o iling ask o analysis o he applica ion pe o mance in ol es. In Figu e 8,
he esul s a e concen a ed in he ini ial zone whe e indica o numbe 2 s ands ou .
Finally, we show he esul s ob ained o one o he key phases in so wa e de elopmen : es ing.
This cha p esen s equi able alues whe e i is no possible o cla i y a esponse among all he indica o s
ob ained.
A e analyzing he e o o he de elopmen phases in companies, i will be e i ied he impo ance
ha he pa icipan s gi e o he exis ence o communi ies and u o ials abou ools and amewo ks ha
a e usually used in so wa e de elopmen .
Figu e 9: How impo an is he communi y and u o ials a ound he ools and amewo ks you use o de elopmen ?
Figu e 9 ep esen s ha he alues a e concen a ed in he highes op ions, wi h 89.7% o he op ions
being selec ed o alues "Ve y Impo an " and "Ex emely Impo an ".
To inish wi h he p esen a ion o da a, he e is a possibili y ha conce ns whe he he o ganiza ion
would be willing o pay o a license o a amewo k o ool ha may add some alue o he de elopmen .
I is clea ly obse ed ha he pe cen ages displayed in Figu e 10 indica e ha all companies p e en
om in es ing oo much in ools, especially in so wa e, since hey all in o disuse oo ea ly. Fo his
eason, 73% o he pa icipan s condi ion he pu chase o such license o he alue ha con ibu es he
equally as well as i s p o i abili y.
200 Challenges o he Adop ion o Model-D i en Web Enginee ing App oaches in Indus y
Figu e 10: Is you o ganiza ion/cus ome willing o pay a license o use a amewo k o ool ha p o ides some alue o
de elopmen ?
6 Valida ion o esul s
Once he esul s o he su eys ha e been assessed and p esen ed, we will compa e he conclusions
ob ained wi h he hypo heses p esen ed in sec ion 3 (Challenges).
6.1 Model ex ensibili y h ough hooks
To begin wi h, we will compa e conclusions om he su ey wi h he hypo hesis dealing wi h he use o
design pa e ns by means o he empla e me hod, as well as he con e sion o code in o an open sou ce.
Fo his pu pose, we will look speci ically a he cha s o Au oma ion o he de elopmen p ocess whe e
i is con i med ha he e o made in his ask is ela i ely signi ican . We will also ake in o accoun
he esul s ob ained in he cha s o code gene a ion om o he elemen s such as mockups, whe e i is
obse ed ha he esul o he su ey gi es his ac a conside able impo ance.
6.2 Open sou ce
As o open sou ce, i is c i ical o highligh he cha Communi ies and u o ials on ools and amewo ks
in de elopmen , whe e 80% o esul s s a e ha hey a e " e y impo an ". To conclude his hypo hesis,
i should be added ha o ganiza ions p e e no o pay o a license o a ool unless he alue i o e s
will gene a e p o i able ou comes, as indica ed by 73% o he esul s in he las cha .
6.3 Need o include na iga ion
Con inuing wi h he alida ion o he hypo hesis, we p oceed o e i y he need o inco po a e na iga ion
in he Web li ecycle, as well as a ool ha suppo s i .
In his case, i has no been possible o ob ain ele an da a indica ing he need o include his aspec
in he li ecycle.
6.4 Time o he clien ’s demand
The nex hypo hesis ha is going o be checked e e s o he ime o he clien ’s demands and how hey
can in luence ools such as me amodels. This can be no iced in he cha o Reques o change o a
Es eban Robles Luna, J.M. Sánchez-Begines, Ma ías Ri e o, L. Mo ales-T ujillo, J.G. En íquez, G. Rossi 201
equi emen du ing de elopmen , whe e he highes esul (44.4%) is gi en o “7 business days” as he
maximum pe iod o espond o a eques .
Then, e o de ec ion ea u es a e included in he MDWE app oach by means o he ollowing cha s:
● Cha s ega ding T aceabili y o he elemen s ha could cause an incidence ha le us p o e
ha he majo i y o pa icipan s (81.8%) conside i as a " e y impo an " aspec .
● Moni o ing o Web applica ion h ough logs o dashboa ds ha is o a e age impo ance among
pa icipan s.
6.5 Tool o suppo he comple e li ecycle
Nex , we will check he need o ind a ool capable o being modi ied in o de o suppo he comple e
li ecycle.
6.6 Adap i e a chi ec u e
La e , he need o ha e an adap i e a chi ec u e, ha allows he de elopmen eam o abs ac high-le el
concep s h ough g aphs ha con empla e he impo ance o unde s anding and con olling he
a chi ec u e o an applica ion, will be co obo a ed. This a chi ec u e mo i a es ha mos o he use s se
ou he idea o unde s anding his concep as an impo an aspec .
6.7 Technical aspec s ha he MDWE app oach suppo s
Subsequen ly, he echnical aspec s ha he MDWE app oach suppo s a e e i ied. Fo ha pu pose,
some cha s co esponding o he measu emen o he e o o he echnical ac i i ies a e used. They
emphasize ha all he cha s cen e hei highes alues on hei middle zone.
6.8 Need o communi ies and u o ials
One o he las alida ions will con i m ha communi ies and u o ials eally need ools and amewo ks
as indica ed in he cha on his opic, whe e mos esponden s ag ee ha i is a c ucial and essen ial
aspec .
6.9 Licenses
Finally, he opic o alida ions is comple ed wi h he hypo hesis on he subjec o licenses and he alue
hey p o ide. The las cha exp esses u h ully, as 73% o pa icipan s chose his op ion, ha
o ganiza ions p e e no o pay, unless he alue o e ed by such ools is p o i able enough.
Finally, Table 2 p esen s a summa y o his su ey wi h he esul s o each o he p oposals.
Challenges Resul s
Model ex ensibili y h ough
hooks
The e o made in his ask is ela i ely big and impo an . Code
gene a ion is highly
ele an
Open sou ce
O ganiza ions p e e no o pay o a license o a ool
Need o include na iga ion
I has no been possible o ob ain ele an da a
Time o he clien ’s demands
Reques o changing a equi emen du ing de elopmen . 7 labo days as
he maximum e m o espond o ha eques . Wi h ega d o alida ion
o he inclusion o he e o de ec ion, he mos impo an aspec s conce n
aceabili y and
Web
moni o ing wi h logs
202 Challenges o he Adop ion o Model-D i en Web Enginee ing App oaches in Indus y
Tool o suppo he comple e
li ecycle
A ool capable o being modi ied in o de o suppo he comple e
li ecycle
Adap i e a chi ec u e The majo i y o use s conside ha i is e y impo an unde s anding
his concep
Technical aspec s ha he
MDWE app oach
suppo s
The highes alues a e in e media e- alue e o
Need o communi ies and
u o ials
Necessa y and impo an
Licenses
P e e no o pay o hem
Table 2. Summa y o Su ey
6.10 Th ea s o alidi y
The e a e se e al h ea s o alidi y ha a e conside ed du ing he expe imen design. This esea ch
p esen s a p elimina y esul and, o space limi a ions, we only include he mos impo an se o
iden i ied h ea s.
Cons uc alidi y. The expe imen was designed o es a se o hypo hesis based on a well-de ined
su ey. In o de o educe he expe imen ’s complexi y and bias in oduc ion possibili y, we de ined he
su ey as he only a iable. Ques ions we e non-ambiguous and a classic 5-poin scale answe wi h no
ee con en was speci ied o mi iga e any misunde s anding.
In e nal alidi y. To mi iga e hese kinds o h ea s, a homogeneous se o p o essionals we e chosen
and il e ed in o de o a oid any ex e nal ac o ha may in luence he answe s. We p e-checked ha
all pa icipan s had se e al yea s o expe ise in he sec o and also ensu ed ha hey had aluable
indus ial posi ions a ha momen . Besides, se e al di e en companies we e chosen in o de o espond
o di e si y coming om di e se wo kplaces wi h in e nal p ac ices ha may a y om s anda ds.
Ex e nal alidi y. MDWE me hodologies a e 15 yea s old and, as model-d i en pa adigm in gene al,
hey ha e demons a ed ha he e exis cases in which models can inc ease p oduc i i y signi ican ly.
Howe e , MDWE me hods ha e no eached as big audiences as o he s, o ins ance, Agile. While his
may comp omise gene aliza ion o he esul s, his pape in ends o answe a bigge ques ion ha is
gene al in he Compu e Science ield: how much and in which con ex s abs ac ion helps inc ease
p oduc i i y in he di e en phases o he de elopmen cycle? This pape has he in en ion o answe his
ques ion, which is a gene al one, in he con ex o Web Enginee ing. To lessen his kind o issues and
ensu e gene aliza ion o his su ey, ou ques ions we e ca e ully elabo a ed o include basic and co e
aspec s o he Web de elopmen p ocess like a chi ec u e, logging, main enance o p o iling, ha a e
ex emely impo an in en e p ise con ex s.
Conclusion alidi y. E e y hypo hesis desc ibed in Sec ion 3 was backed up by conc e e esul s o
he su ey, whose ques ions we e ca e ully designed o p o e hem in a Goal-Ques ion-Me ic-like
ashion. S a is ical signi icance was no a p oblem in his case since no s a is ical es s was equi ed o
his wo k.
7 Conclusion and Fu u e Wo ks
To conclude, i has been possible o success ully e i y many o he p oposals aised in sec ion 2 hanks
o he pa icipan s o he su eys ha sha e he p esen ed needs. In spi e o his, no all he p oposals
ha e had he expec ed accep ance, since he needs o he esponden s ha e no eally cla i ied whe he
he p oposed solu ion i s each o hem o i is he same o all. Ne e heless, a e ca ying ou he
su eys and due o he possibili y o imp o ing some o he phases o he MDWE app oaches wi h he
Es eban Robles Luna, J.M. Sánchez-Begines, Ma ías Ri e o, L. Mo ales-T ujillo, J.G. En íquez, G. Rossi 203
p oposals p esen ed in his a icle, i can be e i ied ha he accep ance and ma gin o imp o emen o
his pa adigm can be qui e b oad.
A e he esul s ob ained om he hypo heses posed, a new pa h can be de ined in he u u e o he
MDWE app oach as a u u e line o wo k. In his sense, he idea would be ei he o implemen such
solu ions so ha hey can be in eg a ed in o he MDWE me hodology by means o some ool o o expand
he exis ing ones. Apa om his, i should be no iced ha he solu ions p oposed ha e no ma ched he
esul s o he su eys, hus i would be use ul o p o ide new solu ions o o e come iden i ied
de iciencies o e en o e alua e again whe he o no such de iciencies eally exis .
As we ha e ealized h oughou he pape , he e a e many ields ha can be implemen ed wi hin his
app oach, al hough we ha e jus ocused on some o hem. As possible u u e wo k, i would be e y
in e es ing o co e o he ields such as secu i y, usabili y o suppo o es ing among o he s, which
would u he inc ease he bene i s o his app oach.
Acknowledgemen s
This esea ch has been suppo ed by POLOLAS p ojec (TIN2016-76956-C3-2-R) o he Spanish
Minis y o Economy and Compe i i eness and by he Fi h In e nal Resea ch Plan (VPPI) o he
Uni e si y o Se ille.
Re e ences
[1] B. Selic, The p agma ics o model-d i en de elopmen , IEEE So w. 20 (2003) 19–25.
doi:10.1109/MS.2003.1231146.
[2] G. Rossi, O. Pas o , D. Schwabe, L. Olsina, Web enginee ing: modelling and implemen ing web
applica ions, Sp inge Science & Business Media, 2007.
[3] S. Hull, 20 obs acles o scalabili y, Commun. ACM. 56 (2013) 54.
doi:10.1145/2500468.2500475.
[4] F.J. Domínguez-Mayo, M.J. Escalona, M. Mejías, Quali y issues on model-d i en web
enginee ing me hodologies, in: In . Sys . De . Asian Exp., 2011: pp. 295–306. doi:10.1007/978-
1-4419-7355-9_25.
[5] M. U bie a, D. Dis an e, J.M. Ri e o, S. Fi menich, 25 Yea s o Model-D i en Web
Enginee ing: Wha we achie ed, Wha is missing, CLEI Elec on. J. 19 (2016) 1:1-1:29.
doi:10.19153/cleiej.19.3.1.
[6] S. Cas eleyn, W. Van Woensel, K. Van De Sluijs, G.J. Houben, Aspec -o ien ed adap a ion
speci ica ion in web in o ma ion sys ems: A seman ics-based app oach, New Re . Hype media
Mul imed. 15 (2009) 39–71. doi:10.1080/13614560902818297.
[7] I. Jacobson, S. Bylund, The oad o he uni ied so wa e de elopmen p ocess, Camb idge
Uni e si y P ess, 2000.
[8] N. Mo eno, A. Vallecillo, Towa ds in e ope able Web enginee ing me hods, J. Am. Soc. In .
Sci. Technol. 59 (2008) 1073–1092. doi:10.1002/asi.20811.
[9] J. Whi le, J. Hu chinson, M. Rounce ield, The s a e o p ac ice in model-d i en enginee ing,
IEEE So w. 31 (2014) 79–85. doi:10.1109/MS.2013.65.
[10] F.J. Domínguez-Mayo, M.J. Escalona, M. Mejías, M. Ross, G. S aples, Towa ds a homogeneous
cha ac e iza ion o he model-d i en web de elopmen me hodologies, (2014).
h p://www.scopus.com/inwa d/ eco d.u l?eid=2-s2.0-
84899720764&pa ne ID=MN8TOARS.
[11] J. Whi le, J. Hu chinson, M. Rounce ield, H. Bu den, R. Heldal, Indus ial adop ion o model-

204 Challenges o he Adop ion o Model-D i en Web Enginee ing App oaches in Indus y
d i en enginee ing: A e he ools eally he p oblem?, in: Lec . No es Compu . Sci. (Including
Subse . Lec . No es A i . In ell. Lec . No es Bioin o ma ics), 2013: pp. 1–17. doi:10.1007/978-
3-642-41533-3_1.
[12] J.A. Ga cía-Ga cía, M.J. Escalona, F.J. Domínguez-Mayo, A. Salido, NDT-Sui e: A
Me hodological Tool Solu ion in he Model-D i en Enginee ing Pa adigm, J. So w. Eng. Appl.
7 (2014) 206–217. doi:10.4236/jsea.2014.74022.
[13] F.J. Domínguez-Mayo, M.J. Escalona, M. Mejías, QuEF (Quali y E alua ion F amewo k) o
model-d i en web me hodologies, in: Lec . No es Compu . Sci. (Including Subse . Lec . No es
A i . In ell. Lec . No es Bioin o ma ics), 2010: pp. 571–575. doi:10.1007/978-3-642-16985-
4_57.
[14] R. Ace bis, A. Bongio, M. B ambilla, S. Bu i, WebRa io 5: An Eclipse-based CASE ool o
enginee ing Web applica ions, Web Eng. (2007) 501–505. doi:10.1007/978-3-540-73597-7_44.
[15] M.J. Escalona, G. Lopez, S. Vegas, L. Ga ccia-Bo goñon, J.A. Ga cia-Ga cia, N. Ju is o, A
So wa e Enginee ing Expe imen s o alue MDE in es ing. Lea ning Lessons, (n.d.).
[16] J.S. Molle i, K. Pe e sen, E. Mendes, Su ey Guidelines in So wa e Enginee ing, in: P oc. 10 h
ACM/IEEE In . Symp. Empi . So w. Eng. Meas. - ESEM ’16, 2016: pp. 1–6.
doi:10.1145/2961111.2962619.
[17] K.B. W igh , Resea ching In e ne -Based Popula ions: Ad an ages and Disad an ages o Online
Su ey Resea ch, Online Ques ionnai e Au ho ing So wa e Packages, and Web Su ey
Se ices, J. Compu . Commun. 10 (2006) 00–00. doi:10.1111/j.1083-6101.2005. b00259.x.
[18] E. Luna Robles, G. Rossi, I. Ga igós, WebSpec: A isual language o speci ying in e ac ion
and na iga ion equi emen s in web applica ions, Requi . Eng. 16 (2011) 297–321.
doi:10.1007/s00766-011-0124-1.
[19] M.J. Escalona, G. A agón, NDT. A model-d i en app oach o web equi emen s, IEEE T ans.
So w. Eng. 34 (2008) 377–394. doi:10.1109/TSE.2008.27.
[20] M. Busch, N. Koch, MagicUWE - A case ool plugin o modeling web applica ions, in: Lec .
No es Compu . Sci. (Including Subse . Lec . No es A i . In ell. Lec . No es Bioin o ma ics),
2009: pp. 505–508. doi:10.1007/978-3-642-02818-2_49.
[21] M. B ambilla, IFML: Building he on -end o web and mobile applica ions wi h omg’s
in e ac ion low modeling language, 2014.
[22] P. F a e nali, M. Tisi, Using aceabili y links and highe -o de ans o ma ions o easing
eg ession es ing o web applica ions, J. Web Eng. 10 (2011) 1–20.
[23] Ma en, h p://ma en.apache.o g, in: Las Access Feb ., 2018.
[24] Jenkins, h p://jenkins-ci.o g, in: Las Access Feb ., 2018.
[25] NewRelic, h p://new elic.com, in: Las Access Feb ., 2018.
[26] A.O. Rami ez, Th ee-Tie A chi ec u e, Linux J. 2000 (2000) 1–4.
[27] G. To e i, Web enginee ing o Cloud compu ing, in: Cu . T ends Web Eng., 2012: pp. 5–19.
doi:10.1007/978-3-642-35623-0_2.
[28] MDWEn , h p://www.iswe-e .de/ac i i ies/2007/mdwe/, in: Las Access Feb ., 2018.
[29] F.J. Domínguez-Mayo, M.J. Escalona, M. Mejías, M. Ross, G. S aples, Quali y e alua ion o
Model-D i en Web Enginee ing me hodologies, In . So w. Technol. 54 (2012) 1265–1282.
doi:10.1016/j.in so .2012.06.007.
[30] RubyOnRails, h p:// ubyon ails.o g, in: Las Access Feb ., 2018.
[31] DJango, h p://djangop ojec .com, in: Las Access Feb ., 2018.
[32] G ails, h p://g ails.o g, in: Las Access Feb ., 2018.
[33] Codeigni e , h p://ellislab.com/codeigni e , in: Las Access Feb ., 2018.
[34] F.J. Dominguez-Mayo, M.J. Escalona, M. Mejias, a. H. To es, A Quali y Model in a Quali y
E alua ion F amewo k o MDWE me hodologies, Res. Challenges In . Sci. (RCIS), 2010
Es eban Robles Luna, J.M. Sánchez-Begines, Ma ías Ri e o, L. Mo ales-T ujillo, J.G. En íquez, G. Rossi 205
Fou h In . Con . (2010). doi:10.1109/RCIS.2010.5507323.
[35] J.A. Ga cía-Ga cía, J. Vic o io, L. Ga cía-Bo goñón, M.A. Ba celona, F.J. Domínguez-Mayo,
M.J. Escalona, A Fo mal Demons a ion o NDT-Quali y: A Tool o Measu ing he Quali y
using NDT Me hodology, in: 21s Annu. So w. Qual. Manag. Con ., 2013.
[36] En e p iseA chi ec , h p://www.spa xsys ems.com, in: Las Access Feb ., 2018.
[37] MagicD aw, h p://nomagic.com/p oduc s/magicd aw.h ml, in: Las Access Feb ., 2018.