To appear in Proceedings of SDNE '95 (June 5-6 Whistler, Canada)

Optimizing World-Wide Web for Weakly Connected Mobile

Workstations: An Indirect Approach

Mika Liljeberg, Timo Alanko, Markku Kojo, Heimo Laamanen, and Kimmo Raatikainen

Department of Computer Science, University of Helsinki

P.O. Box 26 (Teollisuuskatu 23), FIN-00014 University of Helsinki, Finland

Abstract

Modern portable computers and wireless connections over cellular telephone networks have created a new platform for distributed information processing. We have designed a communication architecture that makes it possible to exploit the existing TCP/IP communication protocols but that also takes into account specific features of cellular links. Our communication architecture is based on the concept of indirect interaction. The mediating interceptor is the bridge between the worlds of wireless and wireline communication. It also provides enhanced functionality that improves fault-tolerance and performance. In this paper we demonstrate how the architecture is used to improve the performance of the WWW information browsing. Similar solution methods can be applied to other existing applications and protocols.

1 Introduction

Recent developments in mobile communication and personal computer technology have created a new and challenging platform for mobile information processing. Today the application of cellular telephone networks, such as the Global System for Mobile Communications (GSM) [12, 14], for data transfer between portable personal computers and traditional fixed data communication networks already makes mobile computing possible. However, cellular telephone links have a low and variable throughput. They are also prone to sudden disconnections and are often expensive to use. These characteristics create problems for many existing applications.

New solutions are needed to fully exploit the developing wireless data communication technology. In [9] we presented the Mowgli communication architecture designed to solve many of these issues. In this paper we give a concrete example how the proposed architecture is used to improve the performance of an existing application.

Information retrieval is predicted to be one of the most important applications for future mobile workstations. The World-Wide Web (WWW) [4] is a globally used distributed collaborative information system that has a rapidly increasing user community. Hence, there is real interest for bringing WWW to the mobile workstation. However, the HyperText Transfer Protocol (HTTP) [5] used by the World-Wide Web exhibits

a number of performance problems which need to be solved to make HTTP a practical tool on a weakly connected mobile workstation.

In this paper we examine these problems in detail. We show how to address these problems within our architecture. Our goals are: 1) to improve the performance of the WWW browsing over a wireless cellular link without modifying the existing software and 2) to demonstrate the viability of the Mowgli architecture through this example application.

2 Mobile nodes and World-Wide Web

2.1 Hypertext documents

A hypertext document in WWW is a collection of objects containing various information elements such as text, graphical images and links to other (normally hypertext) objects. The representation of a hypertext object is specified using a special language, the HyperText Markup Language (HTML). It defines several kinds of basic markup as well as support for the hypertext links, inline images, and interactive forms. HTML is an application of the ISO Standard Generalized Markup Language (SGML) and its format has evolved spontaneously among the Internet community. Currently it is being prepared for the formal adoption as an Internet standard [3].

Hypertext links are represented in the form of anchors. An anchor is the start and/or the destination of a hypertext link. An anchor can either point to a whole document or to another anchor in a hypertext object marking a specific point. In this paper we consider, unless stated otherwise, a hypertext link to point to a whole document. The document pointed to need not be a hypertext document. However, the way other document formats (e.g. plain text or postscript) are treated depends entirely on the WWW client software.

2.2 HyperText Transfer Protocol

The HTTP protocol is a stateless protocol designed to transfer data objects between a WWW client and a WWW server. It is currently being standardized within the Internet community [5]. Each HTTP transaction consists of a single HTTP request and of the associated response. The request begins with a request line specifying a document object and an operation.