MODELING WEB INFORMATION SYSTEMS FOR

CO-EVOLUTION

Buddhima De Silva and Athula Ginige

School of Computing and Mathematics, University of Western Sydney, Sydney, Australia

Keywords: Co-evolution, web information system, end-user development.

Abstract: When an information system is introduced to an

organisation it changes the original business environment

thus changes the original requirements. This can lead to changes to processes that are supported by the

information system. Also when users get familiar with the system they ask for more functionality. This

gives rise to a cycle of changes known as co-evolution. One way to facilitate co-evolution is to empower

end-users to make changes to the web application to accommodate the required changes while using that

web application. This can be achieved through meta-design paradigm. We model web applications using

high level abstract concepts such as user, hypertext, process, data and presentation. We use set of smart

tools to generate the application based on this high-level specification. We developed a hierarchical meta-

model where an instance represent a web application. High level aspects are used to populate the attribute

values of a meta-model instance. End-user can create or change a web application by specifying or changing

the high level concepts in the meta-model. This paper discusses these high level aspects of web information

systems. We also conducted a study to find out how end-users conceptualise a web application using these

aspects. We found that end-users think naturally in terms of some of the aspects but not all. Therefore, in

meta-model approach we provide default values for the model attributes which users can overwrite. This

approach based on meta-design paradigm will help to realise the end-user development to support co-

evolution.

1 INTRODUCTION

In traditional approach to develop web information

systems, when an organization finds the need for

new information system, the requirements are

analysed. Then if they go for a custom made

information system the design specification is

produced and the system gets developed tested and

deployed. Otherwise organization may buy a product

to match their requirements and adapt their

processes accordingly. The system then needs to be

maintained until decommissioned. However, we

have found that many such systems that were

deployed with the university and client organizations

after a period of time no longer meet the user

requirements. The failure is due to 3 reasons:

1) When an information system is introduced to an

ganisation it changes the original business

environment thus changing the original

requirements.

2) The processes that information system that

pporting may change.

3) When users get familiar with the system they ask

for m

ore functionality.

This gives rise to a cycle of changes known as

-evolution (Costabile et al., 2006). Meta-design is

proposed as a solution to the co-evolution (Costabile

et al., 2006, Costabile et al., 2005). Meta-design

paradigm characterises objectives, techniques, and

processes for creating new media and environments

allowing ‘owners of problems’ (that is, end-users) to

act as designers (Fischer and Giaccardi, 2004,

Fischer et al., 2004). A fundamental objective of

meta-design paradigm is to create socio-technical

environments that empower users to engage actively

in the continuous development of systems rather

than being restricted to the use of existing systems.

If we are to empower end users to actively

participate in development tasks then, they should be

provided with a suitable web application

development environment. Rode’s study on mental

model of end-user developers reveals that end-users

do have no or very little concerns about the critical

issues in web applications such as authentication,

De Silva B. and Ginige A. (2007).

MODELING WEB INFORMATION SYSTEMS FOR CO-EVOLUTION.

In Proceedings of the Second International Conference on Software and Data Technologies - Volume ISDM/WsEHST/DC, pages 54-61

DOI: 10.5220/0001341600540061

 SciTePress

session management, etc.(Rode and Rosson, 2003).

Therefore we identify the need to bridge the gap

between end-user mental model and developer’s

mental model for the success of meta-design

paradigm. This gives rise to following 2

requirements:

1) End-users need to model a web application using

high level aspects adequate to specify a web

application.

2) Then we have to provide a set of tools and a

framework that they can use to develop and change

the application by specifying and changing these

aspects.

Thus rather than developing “the application” for

end-users to use we need to develop a meta-model to

represent various aspects of applications and a set of

tools which the end-users can use to create the

applications that they want by populating the

instance values in the meta-model. Then end-users

can populate or change the values relating to various

aspects according to their requirements to instantiate

or change the web application. We have created a

meta-model for web based information systems to

support the meta-design paradigm (Ginige and De

Silva, 2007).

We analysed a set of requirement specifications

end-users written for different information systems

within the University to identify how they specify

web applications. In that study we found that end-

users tend to specify aspects of information systems

at a conceptual level. Therefore, we reviewed the

literature on conceptual modelling of web

applications to find different aspects required to

model web applications at a conceptual model. We

found that many conceptual models are proposed to

model a specific type of web application such as

data-intensive web applications, process intensive

web applications. From these, we derived a set of

aspects required to define web conceptual model for

any type of web application. Then we refined the set

of aspects by modelling different types of web

applications. These systems varied from simple web

sites to e-commerce web sites. We have excluded

the special types of web applications such as e-mail

or chat applications which are optimised for specific

functionality. We found that there are 2 broader

categories of web applications; information centric,

and process centric.

• Information centric:

simple web sites with

unstructured information. The focus is on

effective presentation of information.

• Process Centric:

web sites that support business

processes by enabling users to perform actions

such as filling a form, approving a form etc. These

can further be divided into 2 types;

 Data intensive: The focus is on efficient

presentation on structured data such as

product catalogue.

 Workflow intensive: The focus is on efficient

automation of business process consisting of

sequence of steps such as order processing

system.

In this paper, we discuss the ‘holistic’ web

conceptual model to define different types of web

information systems. This will answer the research

question; how can we specify Web Information

Systems? We use WebML, Web modelling language

with extensions to model these aspects. Then for

end-users we provide techniques such as visual tool,

program by example to model these aspects. In

section 2, we review the existing web conceptual

models to identify the required aspects to model web

applications. Section 3 presents the different aspects

of web information systems. In section 4, we analyse

end-user specifications to find out the concepts they

use to define these aspects. Section 5 concludes the

paper.

2 RELATED WORK

Most of the web modeling languages developed at

early days such as HDM-Lite (Fratenali and Paolini,

1998), UWE (Koch and Kraus, 2002), WebRE

(Escalona and Koch, 2006), W2000 (Baresi et al.,

2001), OOHDM (Baresi et al., 2001) and initially

WebML (Ceri et al., 2000) emphasized aspects

required to model data intensive web applications

such as data and hypertext. All these approaches

provided data models to define the entities used and

relationships between them, navigation &

composition models to organize and to access pages

and presentation model to define look and feel.

WebML consists of models such as structural model

(i.e. data model), hypertext model (composition

navigation models), presentation models,

personalised models and operational models. The

operational model is providing a way to model

functions in a data intensive web application.

Modelling in WebML is based on visual notations

which are simple and complete.

Later some of these languages such as WebML

(Ceri et al., 2000), UWE (Koch and Kraus, 2002)

and OOHDM (Baresi et al., 2001) were extended to

model process intensive web applications. WebML

was integrated with Business Process Modellign

MODELING WEB INFORMATION SYSTEMS FOR CO-EVOLUTION

Language (BPML) to model workflows in web

applications. In OOHDM (Schwabe et al., 1996), the

conceptual model is derived based on Object

Oriented Modelling (OMT) principles. It is defined

using classes and relations. The navigation model is

also modelled as two classes, the navigation class

schema and navigation context schema. State charts

are used to define the browsing semantics. The

interfaces are defined using configuration diagrams.

HDM-Lite (Fratenali and Paolini, 1998) uses

structure schema, navigation schema and

presentation schema to model web applications.

Two types of use cases exist in UWE (Koch and

Kraus, 2002) approach called navigation and

functional. Use cases are explained using

behavioural diagrams or textual format.

Recently, Koch et.al have modelled the

conceptual model of web application based on the

web application behaviour and structure (Escalona

and Koch, 2006, Koch et al., 2006). Their

Requirement meta-model is focused on navigation

and presentation of information. Therefore, business

process is modelled as a kind of navigation. This

type of generalization leads to complex models of

web applications.

Jakob, et.al. (Jakob et al., 2006) and Schimid

et.al (Schmid and Rossi, 2004) have identified the

problem with most of these web application models

as neglecting the business processes at the early

stages of modeling. Jakob, et.al. (Jakob et al., 2006)

use “Operational model” to model the business logic

aspect of data intensive applications. However, their

operational model is in the logical level, not in the

conceptual level. They are more focused on

application generation not on requirement

elicitation. In Kobti et.al.’s Conceptual framework

(Kobti and Sundaravadanam, 2006), they use

WebML to model the hypertext aspects of web

applications and Web Service Choreography

Description Language (WS-CDL) to model the

processes. In this approach they model the business

process independent of hypertext modeling

language. But the transaction nature of WS-CDL

won’t be sufficient to define complex business

processes. Also it doesn’t provide a mechanism to

model the access control.

Olveira et.al (De Oliveira et al., 2001) have

modeled the process-centric web applications using

state models. They have modelled Access control as

conditions associated with states and navigation as

state transition. They have associated views with

states. Even though this is an attempt to model the

web application at logical level we see the state chart

can be used to model the state dependant behaviour

of use cases.

However, most of these modelling approaches have

embedded the aspects such as process model and

user model within other aspects rather than defining

these separately during the analysis phase of web

applications. Also we have identify the possibility of

abstracting composition model at a higher level than

the granular elements such as label, text, list box for

process intensive web applications.

3 ASPECTS OF WEB

CONCEPTUAL MODEL

We have analysed modelling of different web

applications from information-centric and process-

centric categories. Based on our finding we have

revised WebML conceptual modelling aspects to

enhance the naturalness and completeness. We

believe these modifications will help end-user

developers to efficiently develop or modify their

web applications. We have identified 5 high level

aspects required to completely define any type of

web application called Hypertext model, User

model, Process model, Data model and Presentation

model.

• Hypertext model: Hypertext model consists of

two sub models: composition and navigation.

Navigation model defines how to get in to a

specific page and how to go through a group of

pages in the web application. Thus navigation

defines the sequence of views presented to the

user when user is interacting with the

application. For example possible sequence of

views present in the e-commerce application is

customer details view, order detail view,

payment detail view. Composition model

defines the collection of elements in the view

which help users to interact with web

application. For example the composition model

of a “log in” page consists of log in instruction

provided at the top of the page, user name and

password elements submit and cancel buttons.

• User Model: The underline structure in user

model is that users belong to groups. User

model consists of two sub models called access

control and personalisation models. Access

control model defines what functions user group

can access in an information system. For

example prospective buyer can access the

shopping cart function on an e-commerce

application. But only users who are waiting for

delivery of an order can access their order data.

Personalisation model defines the attributes for

the user that can be used to personalise the web

ICSOFT 2007 - International Conference on Software and Data Technologies

application at group level and user level. When

a user logged on in the same e-commerce

application the list of products a user see on a

personalized home page can be based on user

profile.

• Data Model: Data model defines relevant

objects and relationships among objects used in

the web application. Data defines the domain

objects stored and retrieved through the web

application. For example in the above e-

commerce application we have to store product

information such as name, type, price,

availability, available units, etc. We may also

want to store order information such as

customer name, product quantity, payment

amount, etc. We may want to keep relationships

between order and product to maintain stock

management.

• Process Model: Process model defines high

level abstraction of tasks and workflows to be

performed in a web application. When

modelling web applications sometimes it is

necessary to model some routing tasks that

should happen when a user interact with the

application. Typically this is expressed in a use

case such as use case description of “request for

more information” in an e-commerce

application. These functionalities can be

executed independent of the status of any other

processes. Details of these functionalities are

captured in a Task Model. On the other hand

there may be workflows in web applications to

sequence series of tasks to carry out a business

process. Some examples are order processing

and stock management. In a workflow, when a

task needs to be performed may depend on the

status of some other tasks. The order in which

tasks need to be performed is captured in the

Workflow model.

• Presentation: Presentation model captures the

layout and graphic appearance of all generic

elements appearing on the views of the web

application. This is modelled at two levels in a

web information system: site level, and page

level. At site level we define the look and feel

of the web application. For example in the e-

commerce site we may define the size and style

of the fonts used in labels and style, size and

color for the instructions etc. At page level we

define the same attributes only applicable to a

specific page.

We find that some of these aspects can be specified

independently. For example data aspect can be

specified using objects and relationships among the

objects. However, other aspects such as composition

model and process model depend on data model to

find out the objects or object attributes which

associate with user interfaces or tasks. The process

model also depends on composition model for

actions which trigger states and data model for data

updates. Presentation aspect depends on

personalisation sub model of the user model.

Navigation model depends on access control sub

model of the user model to find out access rights of

the user when authorisation is required. These

dependencies are shown in figure 1.

Figure 1: Dependencies between aspects .

3.1 Hypertext Model

Hypertext model consist of composition and

navigation models. Navigation can be defined

explicitly using menus or implicitly with in the

composition model in a web information system. We

have to define navigation for two different types of

tasks: state dependant, and state independent. In an

e-commerce application the tasks such as “view

product catalogue”, “order” are available for all

users. Also views such as “home page” or “contact

us” pages are available for all users. Then there can

be tasks such as “view order summary” which is

available for authorised users. These tasks which are

always available for users are categorised as state

independent tasks. These tasks can be executed

independent of the state of any processes. Once a

user log in, user will be provided with a menu to

access state independent tasks. On the other hand,

the “view order status is an example for state

dependant functions. These tasks are available to

users only if an order waiting for processing for that

particular user at that time. Navigation model can

identify views to sequence from the access control

model.

Composition model defines the user interfaces of a

web application. There are three different types of

Composition models called unstructured, form, and

MODELING WEB INFORMATION SYSTEMS FOR CO-EVOLUTION

table. Unstructured page can contain content such as

images, links and, text. Form or table model consists

of UI guide, UIElementGroup, and or UIElements,

and UI Actions. UI elements can be WebML data

units, which store or present data. UI elements can

be in input mode or output mode. In a form we have

UI elements in Input mode, i.e. called entry units in

WebML. In a table we can have UI elements in

output mode. It can be WebML index, single data,

multi data, scroll data or search data unit. UIGuide

provide the guidelines to use the particular interface.

For example, in a form interface, the guidelines can

help the users to understand the purpose of the form.

In a table interface, the guidelines can help the users

to interpret the report properly. UI is associated with

a primary business object. For a UI element in input

mode, we can also have associated help tip. Help

tips help users to enter the values of the form UI

element correctly. We extend WebML entry unit by

adding help information for each field of entry unit.

Sometimes, it is required to logically group

UIElements. For example a product order may

include more than one product. The data of the each

product order can consist of quantity of product,

price of product. Thus, product details can be in a

UIElement group of the order UI. UI Actions such

as add product, amend product actions can be

associated with that UIElement group. Order UI

Model can have UI Action to process the order.

Figure 2: Example Hypertext model from Amazon.

Figure 2 shows the partial hypertext model for

amazon site. The composition model consists of

table of products. We model table with WebML

multi data unit as shown in figure 2.

3.2 Presentation Model

Presentation model specifies the look and feel of a

web application. It dictates the layout and graphic

appearance of all generic elements appeared on a

page. We specify the presentation model at two

levels: site level and page level. The template and

style settings of the presentation model at site level

apply to all the web pages of the application. If we

want to model a custom page we have to define a

template and style settings at page level.

If we consider personalisation then presentation

model has to get the values of the presentation

model for the user from the user model. Figure 3

shows the site level template for amazon. Then for

some pages it may have page level templates.

Figure 3: Site level Template for Amazon.

3.3 User Model

User model is based on user, group concept. User

can belong to many groups. Group can have one or

more users. Group can access one or more tasks or

pages using access control model. “Public” is a

special group which, has access to all tasks and

pages which doesn’t require authentication.

“Admin” is another special group which has access

to all the tasks and pages.

We can store values for presentation model at

user level or group level. These personalised values

are used by presentation model to generate views for

given user. We have extended WebML conceptual

model by adding a user model. We identify the need

to define access control explicitly at conceptual level

by end-user developers since this is required to

generate the applications. We also identify

possibility for expressing personalisation model with

in the user model. Since user has attributes and

relations similar to object we use class diagram to

model users. User model is a special data model

ICSOFT 2007 - International Conference on Software and Data Technologies

because it is common in any web information

system. Figure 4 shows the class diagram for user

model to define the access control and

personalisation sub model.

Figure 4: user model.

3.4 Process Model

Pro odel consist two different types of

when a customer

sub

cess m

processes: task related and workflow related. We

define business logic included in processes and this

determines the behaviour of web application. In

other words in a web information system, the

business rules govern what happens next when a

task is performed. Presence of process aspect in an

Information-centric web application is minimum

compared to a process-centric web application.

Jacob et. Al. (Jakob et al., 2006) defines the

minimum operations a data intensive web

application should support as to Add new content

objects or new relations between content objects,

alter existing content objects or existing relations

between content objects, delete existing content

objects or existing relations between content objects,

filter content objects according to conditions and

sort content objects by specified criteria. There can

also be tasks such as send notification associate with

data-intensive web applications.

In the e-commerce example

mits an order, order processing officer should be

able to see the order information. Conditional rules

can also be associated to the flow. We define the

flow of information in the workflow model.

Workflow model defines the state, entry condition,

exit condition, transition, transition activity, etc.

Transition also associated with a task and UI action.

A simplified process model for order processing

system is shown in figure 5. User can place an order.

Then order processing officer process order.

Figure 5: Simplified workflow model for order

processing.

3.5 Data Model

Data Model defines the relevant objects and

relationships among objects used in the web

information system. Data model is the key concept

in data-intensive web applications. Objects have

attributes. We define high level attribute types such

as photo, e-mail ,etc as defined in Smart Business

Object Modelling Language (SBOML) (Liang and

Ginige, 2006). This high level abstract attribute

types will help to improve the naturalness in

modelling data aspect. The partial data model for

amazon is shown in figure 6. There are 3 different

kinds of products such as book, software, electronic.

Product can has discount during a given time frame.

Product can have 0 or more reviews.

Figure 6: Partial Data model.

4 STUDY ON END-USER

SPECIFICATIONS

We analysed 32 proposals submitted by users in

administration positions of an academic organisation

to develop or modify existing information systems.

From these 32 proposals 17 were for new

Information Systems or enhancements to existing

Information Systems. Rest were for new hardware or

hardware upgrades. We further analysed the

software proposals to identify the concepts they have

used. 12 out of 17 users have specified objects and

MODELING WEB INFORMATION SYSTEMS FOR CO-EVOLUTION

some attributes for the objects. In most cases users

had a manual system. Therefore they have an

understanding of data that needs to be managed. No

one has specified navigation personalisation or

presentation models. Only one has specified the

composition model. 50% have specified processes.

80% has specified user model with access control

and some attributes for user groups. Usage of

concepts as a percentage of 17 is shown in table 1.

Table 1: Covered Aspects in end-user specifications.

Aspect Usage

Data 70%

Hypertext 4%

Presentation 0%

Process 50%

User 80%

This is only a sample of end-users. However,

this sample of end-users also confirms Rode’s

conclusion on end-user mental model. They want

systems to store data and then manipulate data. They

are not concern about other aspects such as

presentation and hypertext. All of that will come as

usability issues when they start to use the system.

However, when they become the owners of the

system they need to create or modify the system.

Then they also need to consider the other aspects

such as hypertext model. When designing the tools

for end-users to develop web information systems

we have to incorporate default values for these

aspects in presentation model and navigation model.

This will help to match the tools to end-user mental

model.

5 CONCLUSION

This paper presents high-level aspects of web

information systems that can be used by end-users to

specify the web applications that they want. These

high level aspects can be used in end-user

development of web information systems to support

co-evolution. In our study of end-user specifications

we found that some of the aspects such as data, and

processes are naturally specified by end-users while

some of the aspects such as navigation and

composition models are rarely specified. Even for

the aspects such as data they hardly define the

relationships between objects. These missing aspect

specifications are necessary to define the web

applications completely. Therefore in the process of

matching end-user mental model to developers we

have to use some strategies and techniques to

capture minimum aspects necessary to define a web

application.

REFERENCES

Baresi, L., Garzotto, F. & Paolini, P. (2001) Extending

UML for Modelling Web Applications. Annual

Hawaii Int.Conf. on System Sciences. Miami , USA.

Ceri, S., Fratenali, P. & Bongio, A. (2000) Web Modelling

Language (WebML): a modelling language for

designing Web sites. WWW Conference. Amsterdam.

Costabile, M., F., Fogli, D. & Marcante, A. (2006)

Supporting Interaction and Co-evolution of Users and

Systems. Advanced Visual Interfaces -AVI.

Costabile, M., F., Fogli, D., Mussio, P. & Piccinno, A.

(2005) A meta-design approach to End-User

Development. VL/HCC05.

De Oliveira, M. C. F., Turine, M. A. S. & Masiero, P. C.

(2001) A Statechart-Based Model for Hypermedia

Applications. ACM Transactions on Information

Systems (TOIS), 19, 28-52.

Escalona, M. J. & Koch, N. (2006) Metamodeling

Requirements of Web Systems. International

Conference on Web Information System and

Technologies (WEBIST 2006). Setúbal, Portugal.

Fischer, G. & Giaccardi, E. (2004) A framework for the

future of end user development. IN Lieberman, H.,

Paterno, F. & Wulf, V. (Eds.) End User Development:

Empowering People to flexibly Employ Advanced

Information and Communication Technology. Kluwer

Academic Publishers.

Fischer, G., Ye, E. G. Y., Sutcliffe, A. G. & Mehandjiev,

N. (2004) Meta Design: A Manifesto for End -User

Development. Communications of the ACM, 47, 33-

37.

Fratenali, P. & Paolini, P. (1998) A conceptual model and

a tool environment for developing more scalable and

dynamic Web applications. EDBT 98. Valencia, Spain.

framework to support Meta-Design Paradigm. 12th

International Conference on Human-Computer

Interaction (HCII 2007). Bejing, P.R. China,

Springer.-accepted.

Jakob, M., Schwarz, H., Kaiser, F. & MITSCHANG, B.

(2006) Modeling and Generating Application Logic

for Data-Intensive Web Applications. Internationa

Conference on Web Engineering (ICWE06). ACM

Press.

Kobti, Z. & Sundaravadanam, M. (2006) An enhanced

conceptual framework to better handle business rules

in process oriented applications. International

Conference on Web Engineering (ICWE06). Palo Alto,

California, USA, ACM Press.

Koch, N. & Kraus, A. (2002) Power of UML-based Web

Engineering. Second International Workshop on Web-

oriented Software Technology (IWWOST02). Málaga,

Spain.

ICSOFT 2007 - International Conference on Software and Data Technologies

Koch, N., Zhang, G. & Escalona, M. J. (2006) Model

Transformation from Requirements to Web System

Design. International Conference on Web Engineering

(ICWE06). Palo Alto, California, USA, ACM Press.

Liang, X. & Ginige, A. (2006) Smart Business Objects: A

new Approach to Model Business Objects for Web

Applications. 1st International Conference on

Software and Data Technologies. Setubal, Portugal.

Rode, J. & Rosson, M. B. (2003) Programming at

Runtime: Requirements and Paradigms for

Nonprogrammer Web Application Development.

Human Centric Computing Languages and

Environments-2003. Auckland, New Zealand, IEEE

Symposium.

Schmid, H. A. & Rossi, G. (2004) Modeling and designing

processes in e-commenrce applications. Internet

Computing IEEE, 8, 19-27.

Schwabe, D., Rossi, G. & Barbosa, S. D. J. (1996)

Systematic hypermedia application design with

OOHDM. seventh ACM conference on Hypertext.

Bethesda, Maryland, United States, ACM Press.

MODELING WEB INFORMATION SYSTEMS FOR CO-EVOLUTION