A Software Product Line for Digital Libraries

Delﬁna Ramos-Vidal, Alejandro Corti

nas, Miguel R. Luaces,

Oscar Pedreira and Angeles Saavedra Places

Universidade da Coru

na, Centro de Investigaci

on CITIC, Laboratorio de Bases de Datos,

Facultade de Inform

atica, Campus de Elvi

na s/n, 15071, A Coru

na, Spain

Keywords:

Software Product Lines Engineering, Digital Libraries.

Abstract:

Digital libraries (DL) have become very popular nowadays. The most important libraries of the world currently

publish their collections online, properly classiﬁed and, in many cases, with access to the digital version of the

resources. DLs are very useful for the general public and for research groups in many ﬁelds, such as language

and literature. Developing these products is a complex and expensive task, and usually small organizations

have to conform with DLs with poor features or outdated. However, this kind of systems usually share most

of the features, and even the data model, maybe with little modiﬁcations depending on the kind of resources to

manage. In this paper, we analyse the domain of DLs using both the previous knowledge of the research group

and reviewing many existing products. The outcome is an exhaustive feature model that models the variability

of this kind of systems.

1 INTRODUCTION

In the last decades, there have been many efforts to

create digital libraries (DL) to manage and put in the

web digital resources (and their bibliographic infor-

mation), either obtained from digitization processes

of literature and cultural heritage or native digital re-

sources. Non-proﬁt organizations that own valuable

books, journals and pamphlets have seen the DLs as a

means to preserve and disseminate their heritage. Re-

search groups in Digital Humanities have created DLs

as a base to foster research that would be unfeasible

without their navigation and search capabilities.

The Database Lab, the authors’ research group,

has been involved in a number of successful multidis-

ciplinary projects of digital libraries, in collaboration

with experts in different areas of Humanities (Lan-

guage and literature, Arts, Pedagogy, etc.). In the next

section, we describe a selection of them that are rep-

resentative in their areas.

The development of a DL is a complex process

in many ways. Experts from Software Engineering

and Humanities have to ﬁnd a common language to

understand each other. As a result of years of experi-

ence, we can conclude that the development of a DL

from scratch is too expensive both in time and bud-

get that can hardly be faced by non-proﬁt organiza-

tions or research groups. However, there are a set of

characteristics that DLs have always in common and

hence a shared core of assets can be deﬁned. Further-

more, even though each DL has its own particularities

in terms of data model and characteristics, they can be

analysed and deﬁned using a formal language. There-

fore, the domain of DL is suitable for the application

of software product lines engineering (SPLE).

In this paper, we present the analysis of the do-

main of DLs following a speciﬁc methodology for the

application of SPLE, and we deﬁne a complete feature

model for DLs using the knowledge acquired by our

research group. Hence, this paper constitutes a ﬁrst

step to achieve the generation of DLs automatically

from a simple speciﬁcation of its requirements. We

also present an application example on how a partic-

ular DL can be deﬁned from the feature model repre-

senting the variability of this domain.

The rest of the article is structured as follows: In

Section 2 we present background and related work.

In Section 3 we describe the domain and address the

ﬁrst stage of the SPL deﬁnition, which includes prod-

uct planning and variability modelling. In Section 4

we present a case of use of the SPL, describing a real

application by means of features. Finally, Section 5

presents the conclusions of the paper and lines for fu-

ture work.

Ramos-Vidal, D., Cortiñas, A., Luaces, M., Pedreira, O. and Places, A.

A Software Product Line for Digital Libraries.

DOI: 10.5220/0010214903810389

In Proceedings of the 16th International Conference on Web Information Systems and Technologies (WEBIST 2020), pages 381-389

ISBN: 978-989-758-478-7

381

2 BACKGROUND AND RELATED

WORK

2.1 Digital Libraries

The Library of Congress

is the largest DL in the

world, with millions of resources. It stores not only

books or printed material, but also photos, drawings,

newspapers, music, maps, and many other kinds of re-

sources. Its web page is updated frequently with new

collections and news. It provides a simple search en-

gine to search for simple keywords, but the results can

be ﬁltered and sorted by many properties. The user in-

terface is very friendly and modern, and the resources

can be downloaded freely in different formats. This

DL is used by the congress of the US as its main doc-

umentation provider.

Project Gutenberg

is a huge DL that provides

over 60,000 free eBooks for online reading or down-

loading, mostly of old works for which U.S. copyright

has expired. The DL is maintained by volunteers who

digitize the content. Its interface is not very friendly,

and its search engine capabilities are limited (it uses

external search engines to search within the content of

the works), but it provides categorized browsing from

more general categories, or bookshelves, as the web

calls them, to speciﬁc domains. The eBooks are also

provided in different formats to facilitate the readers

its download and posterior reading.

Cervantes Virtual Library

is one of most im-

portant digital libraries in Spain. Its catalogue is clas-

siﬁed into different areas, from literature and lan-

guage to american library or galician literature li-

brary. They support digital resources of many types:

pdf, web, images and eBooks. Also, they provide

metadata in MARC21 format (Library of Congress -

Network Development and MARC Standards Ofﬁce,

2016), the physical libraries standard. The search en-

gine allows searching keywords in the content or the

metadata, and it allows ﬁltering depending on the kind

of document.

Golden Age Digital Library (BIDISO)

is a por-

tal that includes in fact many DLs related to the Span-

ish Golden Age, with collections of four different

kinds: historical emblems, news pamphlets, erudi-

tion resources, and inventories and libraries. The web

page provides a search engine that can query all col-

lections simultaneously but also links to each one of

Library of Congress: https://www.loc.gov/

Project Gutenberg: http://www.gutenberg.org/

Biblioteca Virtual Miguel de Cervantes: http://www.

cervantesvirtual.com/

Biblioteca Digital Siglo de Oro (BIDISO): https://

www.bidiso.es/index.htm

Figure 1: Screenshot of a resource webpage from Symbola.

Figure 2: Screenshot of location management from CB-

DRS.

the related DL so the user can browse through its con-

tent. The user interface varies depending on the DL,

being some of them very old and without modern and

responsive interfaces, but other ones are quite mod-

ern and well design. Some of the DLs related are

SYMBOLA or CBDRS. The former, SYMBOLA

is a DL specialized in historical devices or imprese

(e.g., jousting devices, heraldic badges), from whom

stands out its administration site management, includ-

ing their feature of offering static pages to model ad-

ditional information about the library itself, and the

user interface friendliness and styling (see Figure 1).

The latter, Catalogue and Digital Library of News

Pamphlets (CBDRS)

, is a DL of news pamphlets,

where their advanced search feature, enabling exhaus-

tive and restrictive ﬁltering of the data, stands out, to-

gether with the geographical location management of

publication places, and the inclusion of maps for its

physical representation , visible in Figure 2.

The project Critical Edition

, based on the study

SYMBOLA Divisas o empresas hist

oricas: https://

www.bidiso.es/Symbola/

Cat

alogo y Biblioteca Digital de Relaciones de Sucesos

(CBDRS) https://www.bidiso.es/CBDRS/

Edici

on en Galiza durante a etapa franquista: http:

//ediciongalizafranquista.udc.gal/

APMDWE 2020 - 5th International Special Session on Advanced Practices in Model-Driven Web Engineering

382

of the publishing industry in Galicia during the Fran-

coist period, was also taken into account. Protrudes

their intricate data model, an ambitious approach that

takes into account all different types of literary pub-

lications, them being editions, works, parts of works,

collections of editions, and modelling the different re-

lationships that can happen among them, as well as

their authorities, which can sign each work with a dif-

ferent pseudonym, and the organizations in charge of

executing the edition of each work. In this DL, the

collections do not have digital resources, but external

links to other DLs where the resources are located.

The project from the University of A Coru

na,

Galician Virtual Library (BVG)

, the Galician Vir-

tual Library, is an old fashioned web site with a large

collection of Galician publications, with digital re-

sources in different formats: text, audio, video, and

images. It allows search by content using a strategy

based on Bounded Natural Language. The resources

stored in the database can have comments made by

the users. It also handles editions from the same au-

thor, even when written under different pseudonyms.

2.2 Software Product Lines

Software Product Lines Engineering (SPLE) is a ﬁeld

of Software Engineering that focuses on reusing the

same software artefacts among different software sys-

tems from the same family trying to reduce develop-

ment costs and time to market such systems. This

is, when a software development team has to de-

velop a set of products that share most characteristics,

both functional and non-functional, instead of carry-

ing out the development of each one of them from

scratch, it is efﬁcient to develop a Software Product

Line (SPL) and approach the development of all of

them together (Pohl et al., 2005). This way, the base

code of all the products is the same, and each of the

actual products to sell or ﬁnally publish includes the

share of this code that it needs depending on the ac-

tual functionalities it needs to provide. For exam-

ple, if our company specializes on e-commerce web

applications, some clients may need to have “tags”

on the products, while others do not. Then, “hav-

ing tags on the products of the shop” can be a vari-

ant feature that can be selected or not for a speciﬁc

product. The source code this functionality is im-

plemented will only be included in the ﬁnal product

source code if the feature is selected.

Designing and developing an SPL is a costly task.

As we express in the above paragraph, it is more efﬁ-

cient than developing each product in a one-by-one

basis, but however is also a more complex process

Biblioteca Virtual Galega: http://bvg.udc.es

Figure 3: Methodology for constructing a SPL (Corti

nas

et al., 2017).

since the decisions made affect not only to a partic-

ular product but to the whole set of them. There are

different methodologies to follow when designing a

SPL. The authors of this work deﬁned one in a pre-

vious work (Corti

nas et al., 2017), suitable when the

development team has access to both the source code

of existing products of the family and to experts in

the domain. This methodology divides the deﬁni-

tion process in four stages, as seen in Figure 3: re-

quirements analysis, architecture design, architecture

evaluation, and derivation of a speciﬁc product. Re-

quirements analysis stage is divided into two steps,

domain analysis, which consists in analyzing both the

requirements and the related work of the domain, and

product planning, which consists on studying existing

applications of the domain or family, including prod-

ucts developed from the team who is designing the

SPL and products external to it. Architecture design

focuses on studying relevant product architectures for

the domain and prioritizing components, as well as for

deciding the technology stack for the products. The

output of both these stages is the input of the next one,

architecture evaluation, whose purpose is to validate

the models previously deﬁned and to link the different

assets deﬁned in them to support the future evolution

of the SPL. The last stage is the derivation of a spe-

ciﬁc product, which in fact takes part on a different

context, the application engineering, this is, the team

in charge of conﬁguring a speciﬁc product is the one

in charge of this product, not the whole SPL.

This work focuses on the ﬁrst stage of the method-

ology, requirements analysis, with the objective of

identifying the set of features that the whole possi-

ble products of a family could require in an exhaus-

tive way. The output of this stage is an artefact called

feature model, a tree diagram used to represent fea-

tures of a product or family of products. A feature is

a characteristic of a system relevant for some stake-

holder, functional (e.g., a concrete requirement such

as “user authentication”) or not (e.g., the database

management system to use) (Apel et al., 2013).

A Software Product Line for Digital Libraries

383

3 REQUIREMENTS ANALYSIS

This section illustrates the ﬁrst stage for the deﬁni-

tion of a software product line (SPL) in the domain

of digital libraries following the methodology deﬁned

in (Corti

nas et al., 2017). First of all, we describe the

domain. Then, we detail the two steps of this stage

(see Figure 3). Finally, we show and explain the fea-

ture model deﬁned as the result of the analysis.

3.1 Description of the Domain

DLs are information systems where large collections

of information are stored in digital formats and of-

fered to the users through one or many interfaces.

This generic deﬁnition serves to give an impression

of how wide is the spectrum of information systems

that ﬁt this domain. Usually, these collections of doc-

uments follow the traditional approach for publica-

tions. This is, there are authors, who write works,

published in particular editions. DLs provide an ex-

tensive number of functionalities to manage the col-

lections each one handle, and the different types of

elements that are related such as editions of a work,

collection of editions, parts of works, or standalone

works, as well as the authorities involved, let them

be writers, editors, printers or illustrators, to name a

few. There are DLs that only handle metadata, but

most of them also store digital resources linked to the

elements of their collections, such as eBooks, pdfs,

digitized pages or external links to other libraries.

In order to allow the users to search or navigate

through the information, the metadata of the elements

stored in a DL needs to be extracted and conveniently

classiﬁed. Usually, DLs are managed collaborative,

so there is a set of expert users in charge of handling

the elements and their metadata, and there may be

an internal reviewing process among different team

members to ensure that the data updated to the library

is accurate and of quality. This way, there are different

roles for the users, and there may be variable types of

registration for them, including registration through

social networks. Of course, there are DLs without any

users but a unique administrator who manages every-

thing.

The navigation of the web application can be done

through listings or by querying the database. The lists,

forms and detail views of DLs are much more com-

plex than the standard in web applications since there

is a lot of information from related elements involved

in, for example, an edition. Usually, searches area

also quite complex in DLs since the standard user

is normally trying to ﬁnd something very concrete.

Therefore, the search engine should be able to ﬁlter

by any of the attributes of the elements. For example,

the user may want to ﬁnd an edition of a work writ-

ten by a speciﬁc author, published in a range of years,

with some particular text within its content. At the

same time, the user interface needs to be as friendly

as possible, considering the proﬁle of most users, spe-

cialized in literature. Therefore, the design of these

web pages is highly intuitive and transparent.

OAI-PMH is a protocol to share metadata be-

tween information systems through a harvesting pro-

cess done by a server, the harvester, to many clients,

the data providers. It is used to make libraries inter-

operable, allowing small DLs to aggregate their meta-

data collections into a web portal. There are examples

of these portals at different administration levels, such

as Hispana and Europeana. To provide the data from

a DL into one of these portals, called aggregators, the

library needs to support one or many metadata for-

mats. This is, the metadata for the stored collections

needs to be provided following some convection. Of

course, apart from that, the DLs need to implement

the client-side of the mentioned protocol.

There are some DLs in which the geographic

scope is important. To address this, some features

from geographic information systems may be added.

For example, the elements of the library can be shown

on a map viewer, and the users consulting the library

can do queries with ﬁltering the results by some geo-

graphic boundaries.

As we can see, most of the characteristics of this

libraries are quite common, but there is some space

for variability that depends, ﬁrst of all, on type of the

most important element the DL manages. For exam-

ple, there are libraries whose collections are focused

only on a particular author, like Valle-Incl

an Digital

Archive

, and therefore in these libraries there is no

point on showing and managing the set of authors;

however, there are also libraries that study the authors

of a particular period, such as the DL for the project

Publishing in Galicia during the Franco Era (1939-

1975)

, and then the main page of the library should

be the authors themselves, since they are the most im-

portant element of the library, and not their works.

However, apart from this kind of variations, the at-

tributes to record for the editions, or the authorities,

are mostly the same among DLs, and the components

to view or edit the information are very similar, except

for the particular web style.

Archivo Digital Valle-Incl

an: https://www.

archivodigitalvalleinclan.es/publica/principal.htm

Edici

on en Galiza durante a etapa Franquista: http://

ediciongalizafranquista.udc.gal/

APMDWE 2020 - 5th International Special Session on Advanced Practices in Model-Driven Web Engineering

384

3.2 Domain Analysis

We classiﬁed the requirements of this domain in four

different groups: R1) Library objects, which includes

all the requirements related to which kind of elements

the library should handle (e.g., editions, collections,

kinds of authorities, etc.); R2) Library data expor-

tation, with the requirements regarding the ways the

data may be exported from the library (e.g., supported

formats, exportable elements, etc.); R3) User man-

agement, with requirements regarding the user roles

the application should handle, and the features asso-

ciated to each of these roles (e.g., reporters and col-

laborators, digital ﬁle access by anonymous users, the

possibility of registering in the application, etc.; R4)

Library access, which involves a huge set of vari-

ant requirements that involves the kind of navigation

the DLs should support (e.g., the most important el-

ements are editions, works or authors, the way the

users can use search engines, etc.). Besides these,

there is another requirement: the applications should

be in a particular language, which for the DLs is al-

ways the same language of the content of the library.

We modelled the required features for systems of

the domain of DLs with these requirements in mind.

The result is 162 features that represent most of the

relevant functionalities of these applications. Decid-

ing the grade of the importance of each one of these

features is done in the next step, product planning.

This list of requirements or features needs to be

compared to other SPLs for DLs. To the best of our

knowledge, there is not related work about software

product lines applied on DLs. However, there have

been some attempts to generate the software that im-

plements products from this family using other tech-

niques, so we can study these works as a reference.

Gonc¸alves et. al (Gonc¸alves and Fox, 2002) show

a domain-speciﬁc language (DSL) based a formal

model for DLs, that generates a DL application for

a particular system, MARIAN. Using XML, they de-

scribe all the procedures and elements that the DL to

generate requires, including, for example, the opera-

tions the system should handle, such as a search oper-

ation. This level of detail in the deﬁnition of the DL

was required back in 2002, but nowadays software has

evolved to a point where this is not required. Malizia

et al. (Malizia et al., 2010) use a model-driven ap-

proach to generate DLs as Java desktop applications.

Their models describe, as in the previous work, not

only the elements of the library but also the opera-

tions to a high level of detail. It is difﬁcult to compare

the features these systems provide with our proposal

since both of them are very ﬂexible when deﬁning a

new product, so they can adapt better to ﬁne-grained

speciﬁc requirements of a DL. This ﬂexibility comes

at a great cost since the deﬁnition of a product in these

systems is much more complex than a selection of

features in a software product line.

Another interesting document is the DL reference

model (Candela et al., 2008). The European Com-

mission funded a Network of Excellence on DLs,

called DELOS

, whose main objectives were re-

search and technology transfer, although it was aban-

doned in 2009. On their last period, DELOS was

working on the development of a Digital Library Ref-

erence Model, designed to meet the needs of the next-

generation systems and a globally integrated proto-

type implementation of a Digital Library Manage-

ment System.

The research laboratory of the authors of this

work, besides working on speciﬁc DL as mentioned

in Section 2.1, have previously worked in projects tar-

geting improvements on the software supporting these

systems to simplify their implementation, always in

collaboration with researchers from the Digital Hu-

manities area. For example, one of the projects fo-

cused on enhancing the behaviour of compression and

search algorithms when the text is written in a ro-

mance language and, in addition, publishing the al-

gorithms and tools developed as open source. In an-

other project, major goals were the development of

advanced tools to facilitate the establishment of DLs

on the web, while improving the quality of their ser-

vices from a multidisciplinary point of view. In this

last project, two speciﬁc tools were developed: a ﬁrst

tool for rapid prototyping of DLs, and a tool to feed

these DLs with XML extracted from documents.

3.3 Product Planning

The purpose of this step is to verify the requirements

extracted from the previous step with actual products

in production. Besides that, using these products we

can establish which features are common to all the

family of products, which are variant, what kind of re-

lationship exist between a feature and its sub-features,

and also to determine the constraints between the fea-

tures that are not related with the feature tree itself.

Last goal of this step is to priority the features in or-

der of importance, so the implementation of the base

code for the SPL can be useful to develop basic DLs

in early stages, and the development afterwards can

take advantage of the users of these initial products,

improving the quality of the ﬁnal interfaces and func-

tionalities with their feedback.

DELOS Network of Excellence: http://delosw.isti.cnr.

it/

A Software Product Line for Digital Libraries

385

As products that our laboratory has developed in

the past, and therefore we have access to the software

assets, data model, etc., we have decided to use four

products presented in Section 2.1: SYMBOLA, CB-

DRS, Critical Edition and BVG. These four products

are representatives examples which differ in their na-

ture, features, domain, and in their time of creation.

For example, BVG and Critical Edition have been de-

veloped long time ago, while CBDRS and SYMBOLA

actual versions are very modern, having responsive

interfaces to access from mobile devices. Another

point of variability is the way the content is thought to

be accessed. BVG and SYMBOLA designs are focused

on the navigation through categories and listings, and

particularly in the former there is not a search func-

tionality. Critical Edition and CBDRS main entry are

advanced search engines with many ﬁlters and fea-

tures. There are more points of interest about these

applications, such as the complete and complex data

model in which Critical Edition data is stored, or the

geographic features of CBDRS.

In order to conﬁrm the completeness of the analy-

sis, we need to compare our set of requirements with

external products. We decided to use the Cervantes

Virtual Library as one of the more complete DLs that

we know.

3.4 Feature Model

The resulting feature model is shown in Figure 4.

Each requirement group described in Section 3.2

derived in a high-level feature: LibraryObjects,

LibraryDataExportation, UserManagement,

LibraryAccess, and Language.

LibraryObjects feature (35 sub-features) is

directly related to the kind of data the DL supports.

LO E ElementTypes sub-features includes a feature

to enable works, editions, parts of works and col-

lections (ﬁrst two are mandatory since we have not

found a DL without this kind of elements). There

is also a feature to enable the capability of manag-

ing the locations where editions are published, in-

cluding setting the actual geographical coordinates in

a map or managing older naming for the locations.

LO E SchedulingPublication enables scheduling

elements for publication, instead of showing them the

moment they are created. LO E DigitalResource

includes sub-features to decide which kind of digital

resources are allowed. LO Authority includes sub-

features to deﬁne the possible roles of the authorities

managed, the possibility for authors to have aliases

or pseudonyms, and to store contact information for

the authors. LO Organization, which is optional,

enables managing the organizations, and it has sub-

features to decide if the director role is stored.

LibraryDataExportation feature (17 sub-

features) groups all the features related to exporting

data from the listings of the DL. The sub-features de-

ﬁne the possible formats of the exported information

(pdf, excel, txt, etc.), where is the feature applied

(single elements or multi-element exportation), and

which kind of elements can be exported.

UserManagement feature (19 sub-features) en-

ables the user authentication module, which is manda-

tory since in every DL is required at least one user

with the administration role. There are features to

enable other three roles: reporter, which is the stan-

dard registered user role; collaborators, which is the

people that collaboratively digitizes the content; and

anonymous user. Other relevant features are related

to each role, such as allowing anonymous users to ac-

cess the digital resources or requiring the review of

an administration in order to publish some content.

Allowing using social networks account for authen-

tication is also another feature provided, as well as

allowing the anonymous users to register in the DL as

reporters.

LibraryAccess feature (82 sub-features) is

composed by every characteristic of the DL related

to its user interface. It has two main sub-features,

PublicAccess and ManagementAccess, which en-

compass the features for the anonymous users and re-

porters, and for collaborators and administrators, re-

spectively. The two sub-trees are equivalent, describ-

ing the lists and search engines that the DL provides.

The public interface differentiates between two pos-

sible search engines: a simple one, which is accessed

through the menu bar from all across the DL, and an

advanced one which has its own speciﬁc link, and

which has options to enable searching by different

properties. Besides that, the listings of the DL can

have different types of pagination style (depending

on the number of elements expected), and also they

can be standard listings or grid views with thumbnails

of the elements. Management interface related fea-

tures are more simple, since there is only an advanced

search engine, with speciﬁc options such as ﬁlter ele-

ments not reviewed, or ﬁlter elements which are still

drafts, for example. Besides that, PublicAccess has

some sub-features that deﬁne which is the home page

of the DL, which can be a list of a particular element

kind (editions, authorities or works), the list of latest

publications, or a static page that can be managed by

the administrator (if the feature related to this capa-

bility is selected).

Language feature (3 sub-features) serves to se-

lect the language in which the generated application

user interface labels should be. For the libraries anal-

APMDWE 2020 - 5th International Special Session on Advanced Practices in Model-Driven Web Engineering

386

ysed, three languages need support: English, Spanish

and Galician.

Besides the 162 features, there are 13 cross-tree

constraints that affect 22 features. Some of these con-

straints are natural, since for example the homepage

of the public access cannot be the lists of editions

if listings are disabled (PA N HP Editions implies

PA N L Editions, or “DL public homepage is the list

of editions” implies “DL provides the lists of editions

for public access”).

4 RUNNING EXAMPLE

In order to validate our proposal, we have selected an

additional DL that was developed by our group and

we have tried to deﬁne it by means of our feature

model.

Poliantea

is one of the DL part of the BIDISO

portal. It handles works published between XVI and

XVIII centuries that focus on encyclopedic knowl-

edge. Its target audience are students of literary,

iconographic and cultural production for the men-

tioned centuries, since having these works classiﬁed

and easily accessible is very helpful. The last up-

date of the web application was on 2011, so deﬁnitely

could use an upgrade to a version with modern tech-

nologies. Next, we describe the characteristics of the

DL, and we reference the features that support each

of them.

The main page of the library is the list

of the editions (PA

N HP Editions). There is

only one user who has the administration role

(UM UR Admin), so there is no need for any other

kind of users. However, anonymous users can

access the digital resources provided by the DL

(UM UR AnonymousCanAccessElementFiles). Ad-

ministration navigation is very simple, just a

list with all the editions (PA N L Editions and

PA N L E AsList), without pagination (around 140

editions stored currently).

Regarding the public access navigation, the list

of editions does have, in this case, simple pagi-

nation (PA N L P E StandardPagination). Apart

from that, the user has two different search engines.

The basic one searches by title (PA S S ByTitle),

author (PA S S ByAuthor), year of publication

(PA S S ByYear), and by lasts resources added to

the DL. The latter is the one characteristic of the

DL that is not supported by the set of identiﬁed

features for our SPL, since it was not deﬁned as

Poliantea: Enciclopedias, repertorios de lugares co-

munes y miscel

aneas de erudici

on human

ıstica: https://

www.bidiso.es/Poliantea/

a common requirement for this family of products

in our analysis. The advanced search engine can

search by the same properties, but also by location

(PA S A ByLocation) or by date range instead of

a single date (PA S A ByDateRange). Finally, the

search results are the editions that match the query,

and they are shown in a list (PA SR Edition and

PA SR E F AsList).

The main element stored in the DL is, as we

can see for the characteristics mentioned so far,

the edition (LO E ET Edition), which in the case

of Poliantea can be a publication of a particu-

lar work (LO E ET Work), which is the standard,

or to be modelled with more detail and there-

fore composed by parts of works (LO E ET Part).

The elements can have two kinds of digital re-

sources: pdfs (LO E DR PDF) and digitized pages

(LO E DR DigitalizedPages) (which are collec-

tions (LO E DR Collections) of scanned images of

the work). This way, an edition can be com-

posed of several parts, each one with its own dig-

itized pages. The editions stored in the library

have an author (LO A RoleAuthor), and an editorial

(LO O RoleEditorial) (which in the current version

is a simple text ﬁeld).

Summing up, the Poliantea DL can be deﬁned by

means of features of our proposal, using speciﬁcally

55 features, which are the ones mentioned and the

higher-level features of the tree.

5 CONCLUSIONS AND FUTURE

WORK

In this work, the ﬁrst steps for the deﬁnition of a soft-

ware product line in the ﬁeld of digital libraries are

addressed. We carried out an exhaustive analysis of

existent DLs, extracting common requirements and

identifying the set of features that support this family

of products. Afterwards, we established the relation-

ships between these features, and we identify which

ones are part of every product of the family, and which

ones belong to the variability of the product line.

The work presented in this paper is far from ﬁn-

ished. According to the followed methodology, the

next stage is to design the adequate architecture for

the product line, using reference models or architec-

tures for proved successful DLs. After that, a valida-

tion process will occur in which the decisions made so

far are checked, and the two previous stages may need

to run again to make small adjustments for the consis-

tency of the whole design. Finally, the SPL needs to

be implemented, from the product generation mecha-

nism to the different components that provide for the

A Software Product Line for Digital Libraries

387

Figure 4: Simpliﬁed feature model (many features are collapsed).

features of the product family.

One of the main difﬁculties that will be managed

during the development of the SPL is that many fea-

tures directly affect the data model of the products,

and implementing that kind of variability changes is a

challenge.

The ﬁnal validation of the SPL will be carried out

by using it to generate new versions of the DLs from

the Aracne Network

, a platform that joins many

DLs from the Spanish territory in a joint effort, en-

couraging new initiatives to improve the communica-

tion, convergence and projection of the work in the

ﬁeld of Digital Humanities.

ACKNOWLEDGEMENTS

Partially funded by: MICIU/FEDER-UE, MAGIST:

PID2019-105221RB-C41; Xunta de Galicia/-

GAIN Innovapeme: IN848D-2017-2350417; Xunta

de Galicia/FEDER-UE, ConectaPeme, GEMA:

IN852A 2018/14; Xunta de Galicia/FEDER-UE

GRC: ED431C 2017/58; MICIU/FEDER-UE

STEPS: RTC-2017-5908-7; MICIU/FEDER-UE

BIZDEVOPSGLOBAL (RTI-2018-098309-B-C32);

MINECO-AEI/FEDER-UE Datos 4.0: TIN2016-

78011-C4-1-R; MINECO-AEI/FEDER-UE Velocity:

TIN2016-77158-C4-3-R; CITIC research center

Red Aracne: http://www.red-aracne.es/

funded by XUNTA and EU through the European

Regional Development Fund- Galicia 2014-2020

Program, grant ED431G 2019/01.

REFERENCES

Apel, S., Batory, D., K

astner, C., and Saake, G. (2013).

Feature-Oriented Software Product Lines. Springer.

Candela, L., Castelli, D., Ferro, N., Koutrika, G., Meghini,

C., Pagano, P., Ross, S., Soergel, D., Agosti, M., and

Dobreva, M. (2008). The delos digital library refer-

ence model. foundations for digital libraries (version

0.98).

Corti

nas, A., Luaces, M. R., Pedreira, O., Places,

A. S., and

erez, J. (2017). Web-based geographic information

systems sple: Domain analysis and experience report.

In Proceedings of the 21st International Systems and

Software Product Line Conference-Volume A, pages

190–194. ACM.

Gonc¸alves, M. A. and Fox, E. A. (2002). 5sl: a language

for declarative speciﬁcation and generation of digital

libraries. In Proceedings of the 2nd ACM/IEEE-CS

joint conference on Digital libraries, pages 263–272.

Library of Congress - Network Development

and MARC Standards Ofﬁce (2016).

Marc 21 format for bibliographic data.

https://www.loc.gov/marc/bibliographic/. visited

on 2020-09-21.

Malizia, A., Bottoni, P., and Levialdi, S. (2010). Gen-

erating collaborative systems for digital libraries: A

model-driven approach. Information Technology and

libraries, 29(4):171–186.

APMDWE 2020 - 5th International Special Session on Advanced Practices in Model-Driven Web Engineering

388

Pohl, K., B

ockle, G., and Linden, F. V. D. (2005). Soft-

ware Product Line Engineering: foundations, princi-

ples and techniques, volume 49. Springer Science &

Business Media.

A Software Product Line for Digital Libraries

389