1 Introduction

The Internet and the World Wide Web provides vast and ever growing amounts of data with possibly great value. However, it is also a very messy place, and there clearly is demand for tools that can help clear away the rubble and transform the data into useful information for a particular user. One clear indication is that online services as search engines, directories and more or less specialised portals rate among the most visited sites on the Web [Waxman 00]. Albeit highly visible and advertised, these online services are only one way to get the work done: finding useful information and adapt it to the needs of individual users. This work will examine an alternative approach.

1.1 Aim and scope of this thesis

The aim of this thesis is to evaluate the use of client-side proxy servers for finding and adapting information according to user preferences, and to present Blueberry, a partial proxy prototype that highlights some ideas for development of the approach. Traditionally, proxy servers are specialised large-scale Web servers aimed at improving use of network bandwidth through document caching and improving network security by, for example, restricting access to certain content. In contrast, a client-side proxy is a small-scale server running on the user's local computer. It could also be used for improving performance and security, but what is of interest in this context is a more versatile type of proxy server, a proxy with the ability to support information retrieval and adaptation.

To estimate the value of such proxies as opposed to the value of using other methods, some more specific questions need to be answered: Could applications using the client-side proxy approach be better at providing information that fits the needs of an individual user? If so, under which circumstances and for whom? Are there situations when some other technique is preferable, such as browser plug-ins, online services or stand-alone applications? Are the potential benefits of this approach realised in existing systems? If not, in what ways could they be improved? These are the questions that will be examined throughout this work and some of the answers will be visualised in the proxy prototype implementation that is also a part of this thesis.

The scope is limited to the merits of using client-side proxies as an architecture in which functionality for information retrieval and adaptation can be implemented. The functions themselves, such as filtering, collaborative rating, privacy enhancement, etc, will certainly not be overlooked, since there is often a close relationship between the architecture of a system and its functionality. However, they will not receive the full scientific treatment because each of these fields could easily qualify as a separate thesis topic.

1.2 Purpose of examining client-side proxies

The common denominator of many of the popular online services mentioned earlier is that they act as a middleman between the user and massive amounts of information. Just like real-world travel agents, newspapers and libraries, the middlemen on the Web use their domain-specific knowledge and analytical skills to make it easier for the individual to find what he is looking for. This function is clearly in demand despite the prediction that the Web would replace the human middlemen with "Cool Software" that would analyse the user and efficiently provide relevant information [Ganesan 99].

Client-side proxies clearly fall under the "Cool Software" category. So why bother examine the merits of this approach, if the advantage lies with those who provide some more subtle and non-computable service?

One reason is that some of the services provided today are simply not good enough. A popular service (at least among parents) is filtering or blocking content that is "harmful to minors". Such censoring filters have blocked access to Web-sites such as middlesex.gov (due to the domain name), The Privacy Forum (due to a discussion about cryptography that was rated as criminal skills), and other subversive material such as the United States Constitution, the Bible and the plays of William Shakespeare [Neumann and Weinstein 99]. Furthermore, a search engine advertised to be "family-friendly" filtered away about 90% of the relevant hits when queried for material about the American Red Cross, San Diego Zoo and Christianity.

These results certainly indicate that there is a need for a better solution, a solution more adapted to the needs and wishes of the individual user and not coloured by the biased opinions of the middleman. As one among several alternative solutions, client-side proxies could provide a framework for making content retrieval and adaptation more flexible and better adjusted to individual needs. As there probably are situations when this is indeed a better solution than what is available today, and since the merits of this architecture are relatively unexplored, it is an approach worthy of attention.

1.3 Contributions from this thesis

The evaluation of the client-side proxy approach to information retrieval and adaptation is in itself a contribution, since this is a relatively unexplored field. The results of the evaluation should be able to provide useful guidelines for those interested in using this approach.

Blueberry, the prototype implementation that is part of this thesis, could also be viewed as a contribution to the field. As an example application, the goal is that it will provide ideas as to how a proxy can be designed to take advantage of the potential strengths of the approach. As noted, there are already a number of existing client-side proxy applications available, but this one will highlight some issues that has not been fully utilised in the available systems.

1.4 Thesis outline

Section 2 provides the background for this work, giving a brief description of other material about the use of client-side proxies for content processing and identifying some key issues regarding use of such proxies. The methods used in the evaluative sections of this thesis are outlined in section 3. Results of the evaluation are described in sections 4 and 5. The former focuses on the evaluation of client-side proxies compared to other types of applications performing similar tasks. In the latter, the architectural differences between existing client-side proxies are examined. The proxy module implementation of this thesis, Blueberry, is described in detail in section 6. Concluding the thesis, a discussion of the results and ideas for further research can be found in section 7.