SIMPLE EVOLUTION OF COMPLEX WEB SYSTEMS

Lech Krzanik

University of Oulu, Department of Information Processing Science

Box 3000, FIN-90014 Oulu, Finland

Keywords: Web system evolution, pervasive context sensitive applications, user-participative development.

Abstract: Usability implies effective, efficient and satisfactory interaction with a system when using it. Similar

properties are not usually considered when an end user is faced with evolutionary developing or improving a

system – an activity increasingly more often, particularly when the context of using a system undergoes

frequent and significant changes. Project Nomadic Media investigated possibilities for user-participative,

effective, efficient and satisfactory evolution of nomadic systems. We considered geographically

distributed, context-sensitive, personalized systems, exploiting various web services and allowing for a

general function referred to as nomadic blogging. In this paper we demonstrate general policies of evolution

and the specific ways of end-user interaction with the evolving system. The overall strategy may be

generally described as simplicity – effective with clear and straightforward goals, efficient in terms of

resources spent at each small evolutionary delivery step and satisfactory in terms of end-user participation

and reaching the explicit planned user targets. Results of experiments with such systems are demonstrated.

1 INTRODUCTION

We investigate user-participative evolution of web

systems characterized as follows. Complex web

systems are considered including federated context-

sensitive pervasive local subsystems linked to global

web services that contribute to the local behaviour.

In this environment various types of local and

distributed businesses are supported as well as a

range of personal and social links and structures.

Both server and user sides of the system are

considered, and the users interact with the system by

using a broad range of devices, mobile and

stationary. The main goal for such a complex web

system is to coordinate the offered services in a way

which assures that the users may freely change their

physical locations geographically and structurally -

between various business points at a location,

between locations within an environment, between

environments, etc. This type of systems originated

from the investigation of media systems for nomadic

users, later extended to cover for various kinds of

businesses, whether dealing with media directly or

indirectly. The domain was investigated and

validated in project Nomadic Media. The mission of

the project was to “find solutions that allow

consumers to enjoy their content and use interactive

services at the times and in the places they prefer,

using the devices that best suit their circumstances.”

The domain is currently characterized by

considerable uncertainty of requirements,

technologies, markets and regulations. The

evolutionary strategy has therefore been selected for

iterative and incremental development of systems.

An important assumption is to include the end users,

possibly belonging to various stakeholder classes, in

the evolutionary feedback loop. Can we make such

interactions simple enough to be acceptable for end

users?

Usability implies effective, efficient and

satisfactory interaction with a system when using it.

Similar properties are not usually considered when

an end user is faced with evolutionary developing or

improving a system – an activity increasingly more

often, particularly when the context of using a

system undergoes frequent and significant changes.

In order to systematically investigate evolution in

such systems we introduce a general quality attribute

referred to as evolutionary simplicity. The attribute

has the interpretation of usability of evolution

support. It uses similar sub-attributes –

effectiveness, efficiency and satisfaction. Project

Nomadic Media investigated possibilities of user-

participative effective, efficient and satisfactory

340

Krzanik L. (2007).

SIMPLE EVOLUTION OF COMPLEX WEB SYSTEMS.

In Proceedings of the Third International Conference on Web Information Systems and Technologies - Web Interfaces and Applications, pages 340-346

DOI: 10.5220/0001288803400346

 SciTePress

Figure 2: A product line implementation of the variato

process.

evolution of nomadic web systems, allowing for a

general function referred to as nomadic blogging. In

this paper we demonstrate several policies

supporting evolution and allowing for various ways

of end-user interaction with the evolving system.

They should be effective with regard to clear and

straightforward goals, efficient in terms of resources

spent at each small evolutionary delivery step and

satisfactory in terms of reaching the explicit planned

user targets. We introduce three policies,

respectively: For effectiveness it is the

communicative specification (CS) policy. For

efficiency it is the minimalist architecture (MA)

policy. For satisfaction it is the feedback and

cooperation (FC) policy.

Sections 2 and 3 describe the approach to

evolution and the quality attributes. Section 4

presents the policies. Section 5 demonstrates

selected results. Sections 6 and 7 discuss other work

and formulate conclusions.

2 EVOLUTION

Evolutionary approaches to software development

are well suited for projects of various sizes and

under considerable uncertainty of requirements,

technologies, market and regulations (Larman 2003,

Madhavji et al. 2006). An evolutionary project

necessarily consists of two interleaving processes of

variation and selection. The former delivers variant

instances of the system under consideration, e.g., a

mobile service system, while the latter selects the

instance which is best fit and most responsive to

change. Software evolutionary projects are different

from natural evolutionary systems in the sense that

they have considerably accelerated pace, controlled

variation, and the selection occurs within a limited

population of specimen – hence neither the variation

is as random nor is the selection as natural.

Therefore even more attention must be attached to

efficient and well-targeted ways of generating

variants and assessing them by real stakeholders. Of

most importance is such stakeholder communication

which assures a successful selection process in a

community of many stakeholders – that is the

selection process which is the focus of this paper.

An overview of the evolutionary selection process is

shown in Figure 1.

Thanks to the embedded handling of variability,

software product lines (Bosch 2000, Pohl et al.

2005) offer a suitable way of handling the variation

process of evolutionary projects. An example

implementation from Nomadic Media is shown in

Figure 2.

With the focus on quality requirements, we

consider the attribute-driven evolution. In the

considered project the foreground quality attribute

was usability – in the conventional interpretation of

ISO/IEC (1998; Nielsen, 1993; Faulkner, 2000).

Improvements in system qualities such as usability

may result in new function opportunities – as in the

demonstrators discussed in next sections. The

overall function of the considered systems was

nomadic media management and specifically

nomadic blogging (Blood, 2004; Krzanik, 2005).

Evolution is tightly connected with the

architecture of the evolving system. By architecture

we mean a collection of architectural design

decisions, in particular the evolution decisions,

corresponding to the changing requirements and

constraints. According to Jansen and Bosch (2005)

architecture is the structured composition of a set of

business-driven architectural design decisions.

According to Gilb (2005) architecture is the set of

design artefacts, which are selected or exist to

impact a set of stakeholder requirements, by

Application

Evolutionary

specification

Application

Product Line

Architectural

Evaluation

Domain Model Context Model

Derivation

(

l. En

)

Solution Space

Quantitative

Property-

Based Notation

Scoping

(

Dom. En

)

Evolutionary

specification

Evolutionary

Specification

Software

Concept

Preliminary

Requirements

Develop a

Version

Deliver the

Version

Elicit

Customer

Feedback

Deliver Final

Version

Incorporate

Customer

Feedback

Update the

Architecture

Design the

Architecture

Figure 1: The iterative selection process of software

system evolution.

SIMPLE EVOLUTION OF COMPLEX WEB SYSTEMS

341

constraining, or influencing, related systems

engineering decisions.

3 QUALITIES OF EVOLUTION

Because of the likely fundamental role of end users

in web systems evolution, we have investigated the

hypothesis that the quality of evolution can be

measured by the usability of the evolution support

made available to the users. Usability defined by

ISO DIS 9241-11 (ISO/IEC, 1998) refers to products

rather than services; hence we reinterpret it as the

extent to which a service (evolution support) “can be

used by specified users to achieve specified goals

with effectiveness, efficiency and satisfaction in a

specified context of use”. By analogy to good

product usability - meaning that users can easily

perform the different tasks with the product - good

evolution means that users can easily perform

evolution tasks such as providing feedback and

selecting evolution options. By analogy we consider

the evolution simplicity attribute for user-

participative projects, including the three main sub-

attributes: effectiveness, efficiency and satisfaction.

Effectiveness measures the extent to which the

evolution support is able to determine whether the

selected or existing design artifacts provide

contribution toward clear and straightforward goal

specification - suitable for user interaction and

supporting the evolution process. Effectiveness

focuses on fundamental and critical system decisions

(its architecture) and promotes easy communication

between various stakeholders. Efficiency promotes

accomplishing evolution tasks with least resources.

Finally, the satisfaction attribute requires that the

participation in the evolution process is a pleasure to

the participating users. The measurement of the

attribute should be based on the observations of

stakeholder attitudes towards the evolution.

4 POLICIES FOR SIMPLE

EVOLUTION

We propose three policies aiming at contributing to

the three above attributes of effectiveness, efficiency

and satisfaction respectively, for evolving complex

web systems: communicative specification (CS),

minimalist architecture (MA) and feedback and

cooperation (FC). In the proposed configuration CS

is the basic policy, MA and FC depend on it.

4.1 Communicative Specification

The goal of the communicative specification (CS)

policy is to contribute to the effectiveness of the

evolution processes. The solution provides clear and

straightforward web system specifications, which

are suitable for system evolution and can serve as

easy communication media between various

stakeholders. To support effective communication

the specifications shall include elements of both (1)

requirements and (2) design, (3) be able to define

multiple stepwise requirement targets for an

evolving system such that at each delivery step the

process explicitly defines software architecture with

the respective (4) impact on critical requirements so

that evaluation of the effects of the steps is

immediate. Moreover (5) quality of the specification

can also be assessed. Thus the specification

aggregates elements of both problem and solution

spaces. The structure of the specification is shown in

Figure 1. The policy assumes a three-partite

structure of the requirement set including functions,

and attributes (non-functional qualities), extended

with features that associate the two.

Stakeholder

Architecture

Design Artefact

directly supports

Evaluation

Step 2:

Artefactt Level

Evaluation

Step 1:

Architecture Level

Evaluation

Step 3:

Contention

Design Decision

impacts

Performance Resource

Measurement

Scenario

Scale

Unit

Authority

Benchmark

Target

Feature

Function Quality

Requirement

Gist Description

Figure 3: The structure of a communicative specification.

On the solution space side the respective

software architecture is defined as a collection of

architectural design decisions to meet the specified

requirements (Jansen et al., 2005; Gilb, 2005). The

approach includes online architecture evaluation,

which allows for immediate reaction to stakeholder

feedback. The prospective evaluation methods for

WEBIST 2007 - International Conference on Web Information Systems and Technologies

342

Figure 4: A multi-tier architecture of the federate

context-sensitive pervasive local subsystem for nomadic

lications.

the policy are extensions of popular software

architecture assessment methods such as SAAM and

ATAM (Bass et al., 2003; Kazman et al., 2004) as

well as innovative product line evaluation methods

such as the BAPO method (van der Linden et al.,

2004). Evolution can be expressed in terms of both

the requirements and design decisions. The CS

policy provides for a suitable support of the other

proposed policies.

4.2 Minimalist Architecture

The goal of the minimalist architecture (MA) policy

is to contribute to the efficiency of the evolution

process. The solution is to optimize the size of the

architecture. It is expected to allow for architectural

decisions that are accomplished with the least

resources, including as little effort as possible. For

complex web systems software architecture may

include thousands of architectural decisions.

Optimal target architecture should match the current

delivery steps and minimize the number of involved

architectural decisions by including only active and

critical decisions. Moreover not all the remaining

decisions have to be dealt with explicitly and some

can possibly be deferred until later in the iterative

lifecycle. For example Malan and Bredemeyer

(2002) argue that as few decisions as possible should

be actually made.

4.3 Feedback and Cooperation

The goal of the feedback and cooperation (FC)

policy is to contribute to the satisfaction attribute of

the evolution process. The solution is to provide

suitable media to support (a) feedback collection, (b)

formulation and evaluation of candidate and final

architectural decisions, and (c) cooperative decision-

making, with regard to new or revised decisions.

The support is arranged so as to promote continued

and enhanced participation in the evolution activity.

The aim is to ensure that the stakeholders have

positive feelings towards the activity. The solution

we propose is based on visualization and sharing of

the architectural specifications and on providing

access to evaluations which adapt and enhance

selected known assessment methods.

5 DEMONSTRATORS

We consider a complex web system including

federated context-sensitive pervasive local

subsystems linked to global web services that

contribute to the local behaviour. In each of the local

environments various types of local and distributed

businesses are supported as well as a range of

personal and social links and structures. The users

interact with the system by using a broad range of

devices, mobile and stationary. The mission of the

project Nomadic Media has been to “find solutions

that allow consumers to enjoy their content and use

interactive services at the times and in the places

they prefer, using the devices that best suit their

circumstances.” The content may be linked to

individuals, groups, local or distributed businesses,

etc. The domain of such an exploratory project is

characterized by considerable uncertainty of

requirements, technologies, markets and regulations.

Hence the evolutionary strategy has been selected

for project deliveries. We considered a user-

participative evolution process. Figure 4 shows the

architecture of a federated context-sensitive

pervasive local subsystem. The evolution of both

server and user sides of the system were considered.

There were two variants of the subsystem that

provided support for user interaction for system

evolution: (I) One was based on automated regular

end-user interaction analysis and evolution planning

and evaluation facility intended for system architects

and developers only. That system did not allow all

regular end-users to directly access the evolution

support function and was used as a benchmark for

(II) the other variant which provided direct evolution

support for all important categories of stakeholders.

The implementation of both variants was based on a

cooperative wiki engine that followed the FC policy.

Not all devices used in the system had equal access

to all features (a)-(c) (cf. section 4.3) of the

evolution support applications. The results have

shown that variant (II) performed significantly better

Front-End

Infrastructure

Content servers

Access server (Portal server)

Gateways

Wireless network

Internet

Client devices

Internet

Client devices

Back-End Infrastructure

SIMPLE EVOLUTION OF COMPLEX WEB SYSTEMS

343

in terms of the introduced evolution criteria

(conventional product usability). Detailed analyses

of these and other results, also taking into account

the exploratory nature of the project, are currently in

progress. A relatively stable nomadic blogging

functionality (Blood, 2004; Krzanik, 2005) had been

defined and a product line platform, the PORO

system (Krzanik et al., 2005), implementing the

variability process for context-sensitive nomadic

blogging has been built. It has been used for

systematic evolution of nomadic media systems.

Examples of the functionality are visualized in

Figures 5-6.

Figure 5: Public screen control: An application derived

from the nomadic blogging functionality.

Figure 6: Virtual city: An application derived from the

nomadic blogging functionality.

The evolution experiments followed an attribute-

driven strategy with usability as the most critical

driver. Six different experiments and deliveries are

illustrated in Figure 7. The experiments demonstrate

the enabling nature of usability: the incremental

quality improvements not only lead to better

systems, but also were instrumental in discovering

new functional opportunities visualized in Figure 7.

It should however be stressed that the functionalities

of all systems shown in Figure 7 were essentially

very close to one another. In the evolution paths the

sub-attributes of efficiency, effectiveness and

satisfaction were main drivers. It is not yet known to

what degree the inventions of new functions are to

be attributed to the learning process associated with

using the evolution support.

Figure 7: Generating application variants driven by

usability and functionality changes.

We found that the domain vocabularies,

associated with requirements and design engineering

(and the revised orthogonal variability model

included in PORO), were critical in supporting the

stakeholders in the evolutionary process.

Vocabularies as well as other introduced support

services and their positive effect on the process

demonstrated the importance of explicit

consideration of stakeholder communication.

Various investigated applications were integrated

into innovative packages such as the Mobile Bastide

package (Krzanik, 2005) implemented as a dedicated

product line for nomadic media.

6 RELATED WORK

The main elements of the proposed approach include

the user-participative evolution, the tailored

approach to usability and the innovative approach to

software architectures. The revised approaches to

architecture assessment, to stakeholder cooperation

WEBIST 2007 - International Conference on Web Information Systems and Technologies

344

in project evolution, the exploratory wiki application

and the innovative nomadic blogging functionality

are all in a sense consequences. All these introduce a

number of changes to existing practices.

The user-participative evolution is essentially a

voluntary activity. Not only it differs from

conventional approaches (Madhavji et al., 2006) that

establish routine evolutionary processes but also

requires specific approaches which promote user

cooperation – in our case through improved usability

of the evolution support function. Our architectural

approach is based on the decision-oriented variant of

architecture definition (Jansen and Bosch, 2005;

Gilb, 2005), and explicitly integration of problem

and solution spaces - different from conventional

approaches to software architecture (IEEE 1471,

2000). It has had far-reaching consequences for

architecture assessment methods in our evolution

support – the possibility for evaluating architectures

with regard to both requirements and design

artefacts, in terms of end-user characteristics, e.g.,

system properties such as usability. The revised

methods could be used to equip the collaboration

tool such as wiki with more support for various

stakeholders as compared with more conventional

approaches, e.g., by Bachmann and Merson (2005).

The nomadic flavour created many new interesting

applications of the conventional blogging function

(Blood, 2004) but perhaps more interesting is the

innovative exploratory and evolutionary way in

which we propose to discover, analyse and validate

these new applications.

On the implementation side the web and

pervasive solutions tend to use conventional

technology (Hansmann et al., 2004). On the other

hand, with reference to the conventional model of

product line engineering (Pohl et al., 2005), we

introduce a revised orthogonal variability model that

better supports variation and selection processes of

evolutionary development for software systems.

Another proposed change was a closer cooperation

between the processes of domain and application

requirements engineering. We also offload some

non-functional attribute process areas from product

management and shift them to requirements

engineering. Consequently, the product roadmap

artefact that links the two process areas (Pohl et al.,

2005) changes respectively. On the other hand

product management takes responsibility of more

extensive evolutionary variation support. It is also

proposed that responsibilities regarding other new

process areas, such as evolutionary selection

support, are divided between product management,

integrated requirements engineering and the

variability model.

7 CONCLUSION

Evolutionary approaches can be applied to software

development projects of various sizes under

considerable uncertainty of requirements,

technologies, markets and regulations. In doing so

we should carefully investigate issues connected

with realizing the evolutionary strategy. Revisions to

conventional approaches may prove necessary. We

propose a user-friendly evolutionary methodology

for complex web systems. The methodology is still

in its validation stage. We illustrated the concept and

proposed solutions with selected results from a

recent project.

ACKNOWLEDGEMENTS

This work was partly supported by project E!2023

ITEA Nomadic Media.

REFERENCES

Bachmann, F., and Merson, P., 2005. Experience Using

the Web-Based Tool Wiki for Architecture

Documentation. Technical Note CMU/SEI-2005-TN-

041, Software Engineering Institute, September 2005.

Bass, L., Clements, P., and Kazman, R., 2003. Software

Architecture in Practice, 2

Ed. Addison-Wesley

Blood, R., 2004. How Blogging Software Reshapes the

Online Community, Comm. ACM, 47, Nr. 2.

Bosch, J., 2000. Design and Use of Software

Architectures. Addison-Wesley.

Faulkner, X., 2000. Usability Engineering, Macmillan

Press, London.

Gilb, T., 2005. Competitive Engineering: A Handbook For

Systems Engineering, Requirements Engineering, and

Software Engineering Using Planguage, Butterworth-

Heinemann.

Hansmann, M., et al., 2004. Pervasive Computing, 2

Ed.,

Springer.

IEEE 1471, 2000. ANSI/IEEE Std 1471-2000,

Recommended Practice for Architectural Description

of Software Intensive System. IEEE, October 2000.

ISO/IEC, 1998. DIS 9241-11. Part 11: Guidance on

usability. ISO/IEC.

Jansen, A., and Bosch, J., 2005. Software Architecture as a

Set of Architectural Design Decisions. 5th Working

IEEE/IFIP Conference on Software Architecture,

WICSA 2005. IEEE, pp. 109 – 120.

SIMPLE EVOLUTION OF COMPLEX WEB SYSTEMS

345

Kazman, R., Nord, R.L., and Klein, M., 2003. A Life-

Cycle View of Architecture Analysis and Design

Methods. Technical Note CMU/SEI-2003-TN-026,

Software Engineering Institute, September 2003.

Krzanik, L., 2005. Mobile Bastides: Flexible and

Evolvable Business-Oriented Contextual Media, Proc.

11th Intl. VSMM Conf., Ghent.

Krzanik, L., et al., 2005. Usability-Driven Evolution of

Context-sensitive Nomadic Media, Proc. MUM’05

Conf., Christchurch.

Larman, C., 2003. Agile and Iterative Development: A

Manager’s Guide. Addison-Wesley.

Madhavji, N. H., Fernandez-Ramil, J., and Perry, D.,

2006. Software Evolution and Feedback: Theory and

Practice. Wiley.

Malan, R., and Bredemeyer, D., 2002. Less is More with

Minimalist Architecture. IT Professional, Sept./Oct.

2002.

Nielsen, J., 1993. Usability Engineering, Academic Press,

Inc., San Diego.

Pohl, K., Böckle, G., and van der Linden, F.J., 2005.

Software Product Line Engineering: Foundations,

Principles and Techniques, Springer, Heidelberg.

van der Linden, F., Bosch, J., Kamsties, E., Känsälä, K.,

Krzanik, L., and Obbink, H., 2004. Software Product

Family Evaluation. In: Frank van der Linden (Ed.),

Product Family Engineering. Proc. of PFE-5, Siena,

Springer Verlag, Heidelberg.

van Gurp, J., and Bosch, J., 2002. Design Erosion:

Problems & Causes. Journal of Systems & Software,

61(2), pp. 105-119, Elsevier, March 2002.

WEBIST 2007 - International Conference on Web Information Systems and Technologies

346