Real-world information, knowledge, and procedures afterwhich information systems are modeled are generally ofdynamic nature and subject to changes, due to the emer-gence of new requirements or revisions to initial specifi-cations. E-government information systems (eGIS) presenta higher degree of volatility in their environment, sincerequirement changes may stem from multiple sources,including legislation changes, organizational reforms,end-user needs, interoperability, and distribution con-cerns, etc. (Jansen, 2004; Prisma Project, 2002; Scholl,Klischewski, & Moon, 2005. To this end, the design andimplementation of eGIS must adhere to paradigms andpractices that facilitate the accommodation of changes tothe eGIS as they occur in the real world. Object-orientedtechnologies have been extensively used to encapsulatereusable, tailorable software architectures as a collectionof collaborating, extensible object classes; however theinherent conflict between software reuse and tailorabilityhas inhibited the development of frameworks and modelsthat would effectively support all requirements exposedby eGIS (Demeyer, Meijler, Nierstrasz, & Steyaert, 1997).The lack of such frameworks has lead to eGIS that cannoteasily be adapted to the new requirements, mainly be-cause only the predetermined specifications are takeninto account and design decisions are fixed during theimplementation phase (Stamoulis, Theotokis, Martakos,& Gyftodimos, 2003).
A key issue to a viable solution eGIS modeling is theprovision of the ability to multiple public authorities(PAs) to represent different aspects of the same real-world entity, while maintaining at the same time informa-tion consistency. Aspect representation is not only lim-ited to data elements that describe the particular entity,but may extend to behavior alterations, when the entity isexamined in different contexts. For example, an entity
representing the citizen is expected to assume the behav-ior of beneficiary, when used in the context of the Minis-try of Social Security, and the behavior of taxpayer, whenaccessed from the Ministry of Finance’s eGIS. Distinctbehaviors may rely on different data representations and/or respond differently in requests. In this work we presenta role-based modeling and implementation framework,which can be used for building eGIS and we argue that thismodel promotes the tailorability and maintainability ofeGIS.
BACKGROUND
Currently, the representation of different real-world entityaspects is mainly achieved through the use of multiple,totally independent representations of the real-worldentities, one per PA eGIS. Each representation encom-passes the data elements, and models the behavior perti-nent to the specific organization. Note that these data andbehavior may include portions administratively regulatedby other PAs, (e.g., the Ministry of Transport is adminis-tratively responsible for defining vehicle ownership-re-lated data), however the Ministry of Finance eGIS shouldinclude such data, for taxation purposes.
At the other extreme of each eGIS having its own real-world entity representations, a single, centralized reposi-tory can be employed. According to this approach, somePA develops and maintains an eGIS, which is the authori-tative source for both defining the schema and storing alldata values for real-world entities. The schema mustconsolidate all data-related requirements of all PAs, whilesecurity rules can limit the access of any PA to the schemaelements pertinent to its task(s).
Between the fully replicated and the fully centralizedapproach, a federated database approach (Chorafas &Steinmann, 1993) could be adopted, according to which
1065
Tailorable E-Government Information Systems
�each PA eGIS defines a schema portion, which is ex-ported. PA eGISs may also import schema portions thathave been exported by other “federation members”, witheach import augmenting the locally defined schema. Thefederated approach decentralizes schema maintenancetasks, with each PA updating the global schema portionfor which it is responsible.
A methodology towards meeting both the require-ments of multi-aspect modeling and context-specific be-havior, while enhancing the overall system maintainabilityis the adoption of two base constructs for entity modeling:the first construct models fundamental behavioral systemblocks, providing only the essential behavioral elements ofthe most abstract version of the modeled entity; thisconstruct is termed ATOMA (Theotokis, 1997) and realizesthe most basic collaboration and reuse contracts (Codenie,De Hondt, Steyaert, & Vercammen, 1997). Enhanced andcontext-specific behaviors are modeled using a secondconstruct, namely roles, which are attached to atoma formodeling functional behavior related to a basic entity. Roleattachment and removal can be performed dynamically, andmultiple roles can be attached to a single entity, effectivelymodeling facets of this entity. Roles also implement theirown collaboration and reuse contracts, through whichoperations are requested in the context of eGIS.
The Atoma framework is based on the concept ofseparation of concerns (Theotokis, 2003), which is a keyconcept in realizing deferred design decisions as it facili-tates the notion of “injectable” behavioral adjustments inexisting operational eGIS. The ATOMA model allows ob-ject-oriented design and code to be decomposed intounits, describing basic behavior, as this is captured dur-ing the initial design phase from the contractual require-ments, and units that specify either variations or changesto these requirements, or new requirements, as theseemerge in time. Both at design and implementation level,the former are represented as standard object-orientedclasses, while the later are roles that, when composed withclasses, realize the ever-emerging requirements. Each rolecan therefore be refined separately to a code artifact, andthe details of the code composition can be derived fromthose of the design composition.
THE ROLE-BASED MODEL FORTAILORABLE E-GOVERNMENTSYSTEMS
eGIS Modeling Using the AtomaFramework
According to the atoma framework, for each real-worldentity, a single atoma construct is defined, which encap-
sulates the fundamental data and functionality needed formanaging this entity. The PA eGIS within which theconstruct is defined may be chosen on the basis ofvarious criteria, such as administrative responsibility,technical know-how, expected access patterns and soforth. Once the construct for modeling a real-world entityhas been defined, it may be exported for use by other eGIS,which will import this construct. Each eGIS also providesimplementations for the roles that will be assigned to theatoma to be used within its context, either locally definedor imported. For example, the Ministry of Transport maydefine the roles taxi and bus that will be assigned to atomaof type vehicle (locally defined), while the Ministry ofFinance may define the role taxpayer, which will beassigned to atoma of type person, a type imported from theMinistry of Social Security. Correspondence betweenroles and atoma types is not necessarily one-to-one: theMinistry of Finance may define the taxable good role,which can be assigned to atoma of types building, car,value-added service and so on. The communication be-tween a role and the atoma construct to which it has beenassigned is based on the collaboration and reuse con-tracts provided by the atoma construct; thus, the taxablegood role can be assigned to any atoma construct thatimplements the collaboration contract price, which will beinvoked to obtain the net price (before taxation) of therelevant real-world entity. During the period that an entityis assigned the taxable good role, its behavior is en-hanced with the operations defined in the role (e.g., taxamount), while some of its operations may be overriddenby respective ones implemented by the role (e.g., the pricebehavior will be redefined to return the net price plus thetax due). Note that both the role and the underlying entitymay independently evolve without any disruption in theircommunication, as long as the collaboration contracts arerespected. Moreover, the interaction between a role andthe eGIS it is used within will function properly, as longas the existing elements of the collaboration contractimplemented by the role are not altered (new elements maybe added with no side-effects).
Similarly to atoma constructs, roles can also be ex-ported by the PA eGIS that has defined them and importedby other PA eGIS for use, if the functionality they provideis useful in the context of the importing eGIS. For instance,the taxable good role, defined by the Ministry of Finance,may be imported by the Ministry of Commerce and as-signed to atoma constructs, either locally defined (e.g.,value-added service) or imported (e.g., car).
Maintaining Role-Based eGIS
When a new requirement is added or an existing one ischanged, a new design aspect (a role) is created to addressit. The new design aspect can then be composed with the
1066
Tailorable E-Government Information Systems
existing design. Afterwards, the design aspect can berefined to a code aspect (role implementation), which issimilarly composed with the existing code. The changesare localized without risking the rest of the system’sstability, so there is no tangling, and both traceability andmonitoring are preserved. In addition to this, dealing withthe alignment problem, the decomposition into roles alle-viates the monolithic nature of the design, and allows forconcurrent development, while the composition underly-ing the ATOMA model provides a powerful mechanism forintegration, evolution, customization, adaptation,tailorability and improved reuse.
It is in fact this composition mechanism that providesa remedy to the problem of constructing rigid systems. Thenecessary inflexible mapping of requirements to designunits is no longer required, and design decisions, which fixrequirements into design, and subsequently code, are nolonger necessary. By separating the various cross-cuttingaspects of a system, modeling them independently andcomposing them back to deliver the required function ofthe eGIS, we advocate that deferred design decisions canbe realized, and thus enable the construction of trulytailorable eGIS.
Using Atoma and Roles: A Case Study
Figure 1 presents an example of an eGIS with three con-stituent PA eGIS. The case study has been drawn from theGreek public administration, by studying the related legis-lation and practices. Atoma are represented as roundedrectangles, whereas parallelograms are used for roles.Solid lines indicate that the construct has been definedwithin the eGIS it is depicted in, while dashed lines indicateimported constructs. Role assignment to atoma is shownthrough arrows. Only one atoma construct is defined inthis example, namely Person. This construct is defined
within the Ministry of Social Security, which additionallydefines the beneficiary role that it assigns to the personconstruct. In this manner, a Person exhibits the behaviorof beneficiary only when it is treated within the contextof the Ministry of Social Security, whereas outside thescope of this eGIS only the basic Person behavior isavailable. The two other ministries (Health and Finance)import the Person construct and assign to it locallydefined roles (Patient and Taxpayer, respectively). TheMinistry of Finance additionally imports the Patient rolefrom the Ministry of health, and assigns it to the Personconstruct, since the hospitalization expenses (includedin the Patient role’s contract) are needed within thecontext of the Ministry of Finance’s eGIS, to enable theirdeduction from the person’s taxable income.
Since roles are independent of one another, it ispossible that two or more roles implement different be-haviors under the same collaboration contract. For in-stance in Figure 1, it is possible that both roles, Patientand TaxPayer, include a Checkup contract with differentsemantics (triggering clinical tests and initiates a de-tailed audit of tax records, respectively). Since within theMinistry of Finance’s eGIS both roles have been as-signed to the Person construct, an ambiguity issue israised regarding which behavior should be selected.This ambiguity is resolved by defining priorities for roleassignments, effectively dictating the order in which rolecontracts will be scanned for elements matching anyindividual contract use. It is generally expected thatlocally defined roles should be assigned higher prioritiesthan imported roles, however no such restriction is im-posed by the model. The atoma framework encompasseseven more powerful methods for atoma and role combi-nations, such as molecules, which can be used in morecomplex cases (Theotokis, 2001).
Figure 1. Definitions, exports, and imports of atoma and roles
������������������� �����������������
�����������������
��������
������
�����������
������
�����������
��������
����������� ��!��" �������
�����������
������
�����������
1067
Tailorable E-Government Information Systems
�Comparing the Atoma Framework andOther Approaches
As noted in the background section, the dominant ap-proaches for modeling multiple aspects of real-worldentities within eGIS belonging to different PAs are; (a) theuse of totally independent systems, (b) the introductionof a centralized repository where all schemata are consoli-dated, and (c) the use of a federated schema approach.These practices, however, have a number of disadvan-tages seriously impeding the design process, modeling ofcontext-specific behavior, tailorability and maintainabil-ity.
The practice of using independent systems intro-duces multiple autonomous representations, which are apotential source of inconsistencies, both at the level ofschema representation and at the level of stored datavalues (Lenz, 1996; Wiesman, et al., 2000). Indeed, if achange affecting the schema of the data used for a specificpurpose is decided by the administratively responsiblePA, this change should be communicated to all other PAsthat handle such data, and their eGIS should be accord-ingly modified; the distributed nature and the scale ofthese maintenance activities increases the associatedcosts and the probability of errors. At data instance level,changes to data values stored in one eGIS do not affectthe corresponding values in other eGIS, leading to dis-crepancies in the representation of the same real-worldentity. Finally, multiple copies of the same information aremaintained, increasing the overall storage requirements.
The centralized repository approach eliminates incon-sistency problems (only one schema is defined and asingle copy of each datum is stored), it hinders howeverthe modeling of context-specific behavior for the entities.If entity behavior was modeled within the global reposi-tory, a single behavior would be provided for all eGISusing the entity, regardless of the context. If entity behav-iors were provided separately by each PA eGIS, changesto the “global schema” would necessitate maintenanceactivities for all affected PA eGIS. Note also that globalschema amendments and the PA eGIS maintenance shouldbe performed “almost synchronously”, since the PA eGISdata model should exactly match the repository datamodel upon any access.
Finally, the federated database approach cannot ad-dress the issue of context-specific behavior: if each PAeGIS fully models the behavior entities should have in itscontext, changes in some schema portion would again callfor maintenance activities to all PA eGIS importing themodified schema; if the behavior were provided by schemapublishers, context-specificity would be hindered and anadditional concern would be raised, regarding where tocode behavior that uses multiple schema portions, ex-ported by different PA eGIS.
The role-based model presented above successfullytackles all these issues, by allowing organizations tomodel both the data items and the behavior pertinent tothe entity within the role concept, which is attached to thebasic atoma construct corresponding to the real-worldentity. Roles can be shared between eGIS, at both schemaand instance-data level, removing thus any inconsisten-cies due to multiple copies. Context-specific behavior isdirectly modeled through new context-specific roles and/or by fine-tuning the assignment of roles to atoma. Finally,maintenance is facilitated, since changes in existing re-quirements or emergence of new ones can be addressedthrough the modeling of new roles. Note that the creationand assignment of new roles (as well as the de-assignmentof existing ones) can be performed at any instant, remov-ing the need for fixing design decisions at early stages,and increasing the overall system flexibility.
FUTURE TRENDS
eGIS have emerged in the past few years, aiming to exploitinformation and communication technologies for provid-ing public services with improved quality and reducedcosts (Forman, 2002). Insofar, most of the work has beendevoted to developing isolated informational and trans-actional services, but service interoperation and integra-tion, which provides added value for service consumers[e.g., for handling life-events (Wimmer & Tambouris,2002)], has not received comparable attention. It is ex-pected that the research and implementation agenda foreGIS will include these aspects, for which the work pre-sented herein provides the necessary technical infra-structure. Of equal importance is the provision of a meth-odological framework that would support the phases ofeGIS requirements analysis and design, enabling involvedstakeholders to unambiguously distinguish between ba-sic and context-specific functionalities of real-world en-tities, providing appropriate input to the implementationphase. Methodologies extending beyond traditional ser-vice-provision systems, to include the dimensions ofmanagerial effectiveness improvement and democracypromotion (Gil-Garcia, 2004; Cohen, 2002), providing aholistic framework for e-government will also be of es-sence.
CONCLUSION
In this article we have presented a framework for con-structing tailorable eGIS. The proposed framework em-ploys two basic constructs for modeling real-world enti-ties, namely atoma for encapsulating basic representa-
1068
Tailorable E-Government Information Systems
tional needs and fundamental functionality, and roles forrepresenting context-specific behavior that is attached toatoma on demand, according to the needs of the eGISwithin which the entity is examined. By employing thisdesign and implementation frameworks organizationsenhance the overall system maintainability, since atomaand roles may evolve independently, provided that thereuse and collaboration contracts are respected along theevolution. New behaviors, necessitated by changes ofexisting requirements or emergence of new ones can alsobe seamlessly integrated into the system, by modelingnew roles and attaching them to the pertinent atoma.Future work will address the formulation of a methodol-ogy for eGIS design according to the atoma framework, aswell as issues related to migration and mobility of atomsand roles, as well as versioning and persistency.
REFERENCES
Chorafas D. N., & Steinmann, H. (1993). Solutions fornetworked databases: How to move from heterogeneousstructures to federated concepts. Academic Press.
Codenie, W., De Hondt, K., Steyaert, P., & Vercammen, A.(1997, October). From custom applications to domain-specific frameworks. Communications of the ACM, 40(10).
Cohen, S. (2002). The future of e-government: A projec-tion of potential trends and issues. Price WaterhouseCoopers. Retrieved from http://www.columbia.edu/~sc32/future%20of%20egov.pdf
Demeyer, S., Meijler, T. D., Nierstrasz, O., & Steyaert, P.(1997, October). Design guidelines for “tailorable” frame-works. Communications of the ACM, 40(10).
Forman, M. (2002). E-government strategy: Implement-ing the president’s management agenda for e-govern-ment. Retrieved from http://www.whitehouse.gov/omb/inforeg/egovstrategy.pdf
Gil-Garcia, R. (2004). Electronic government. ISWorldEncyclopedia. Retrieved from http://ispedia.terry.uga.edu/?title=Electronic_Government
Jansen, B. (2004). Capturing legislation for egovernance.White paper. Retrieved http://www.egovernance-consortium.org/WhitePapers/Janssen2004a.pdf
Lenz, R. (1996). Adaptive distributed data managementwith weak consistent replicated data. Proceedings of the1996 ACM symposium on Applied Computing Table ofContents (pp. 178-185). Philadelphia.
ing. Retrieved from http://www.prisma-eu.net/deliverables/D4-1.PDF
Scholl H. J., Klischewski, R., & Moon M. J. (2005).Minitrack: “E-government infrastructure andinteroperability”. Proceedings of the 38th Hawaii Inter-national Conference on System Sciences.
Stamoulis, D., Theotokis, D., Martakos, D., & Gyftodimos,G., (2003). Ateleological development of “design-deci-sions-independent” information systems. In N. Patel (Ed.),Adaptive evolutionary information systems (pp. 81-104).Hershey, PA: Idea Group Publishing.
Theotokis, D. (2001). Object-oriented development ofdynamically modifiable information systems using com-ponents and roles. PhD Thesis, Deptartment of Informaticsand Telecommunications, University of Athens (in Greek).
Theotokis, D., et al. (1997). ATOMA: A component ob-ject-oriented framework for computer based learning.Proceedings of the 3rd International Conference on Com-puter Based Learning in Science (CBLIS ’97) (pp. G5-G15). De Montford University, Leicester.
Wiesman, M., et al. (2000). Database replication tech-niques: A three parameter classification. 19th IEEE Sym-posium on Reliable Distributed Systems. Nürnberg, Ger-many.
Wimmer, M., & Tambouris, E. (2002). Online one-stopgovernment: A working framework and requirements.Proceedings of the IFIP World Computer Congress.Montreal.
KEY TERMS
Atoma: The most abstract level of representation ofa real-world entity, encompassing only the most basicdata and functionality for representing and manipulat-ing the entity.
Centralized Representation Repository: A singledatabase in which all real-world entity representationmodels are stored, and any information system willing toaccess a model retrieves if from there. This approachkeeps the models consistent but shifts the maintenanceissues to the client information systems.
Export/Import: Information systems may define atomaand roles and then export them to be used by other informa-tion systems. An information system may import any ofthe exported atoma and roles and use it in its own context.
Federated Representational Model: An approach forbuilding representational models according to which each
1069
Tailorable E-Government Information Systems
�information system defines some aspects of an entitymodel, contributing thus to a “global” entity model.
Representational Diversity: the practice according towhich real-world entities are represented through totallyindependent and unsynchronised models in differentinformation systems, leading to inconsistencies and main-tenance problems.
Role: A situation- or context-specific aspect of a real-world entity. A role may be dynamically associated and
removed from an atoma construct when either the real-world entity evolves accordingly (e.g., a person becomesan employee or looses this property) or when the behaviormodeled by the role becomes/ceases to be pertinent to thecurrent context.
Tailorability: The dynamic accommodation of con-text-dependent behavioral variations in an existing soft-ware system.
