Readings in Information Retrieval
Karen Sparck Jones and Peter Willett (eds.)
Morgan Kaufmann Publishers, 1997, ISBN 1-55860-454-5.

Wide Area Network Design: Concepts and Tools for Optimization
Robert S. Cahn
Morgan Kaufmann Pulishers, 1998, ISBN 1-55860-458-8.

Switching in IP Networks: IP Switching, Tag Switching, and Related Technologies
Bruce Davie, Paul Doolan, and Yakov Rekhter
Morgan Kauffman Publishers, 1998, ISBN 1-55860-505-3.

Readings in Information Retrieval
Karen Sparck Jones and Peter Willett (eds.)
Morgan Kaufmann Publishers, 1997, ISBN 1-55860-454-5.

Reviewed by Rajesh Krishnan.

Readings in Information Retrieval is an excellent collection of papers spanning over four decades of research in the field of Information
Retrieval. The editors Karen Sparck Jones and Peter Willett are pioneers in the field. Sparck Jones has contributed to and witnessed the evolution of this field since its nascence in the '50s and Willett has been active in this field since the '70s. Having actually fought in the trenches, they provide an excellent historical perspective of the field, which they are uniquely qualified to offer.

Information Retrieval is presented in this book in the classical context of textual documents. The papers in this collection are divided into 7 sections -- history, key concepts, evaluation, models, techniques, systems and extensions. Each section begins with a well-written introduction from the editors. The collection begins with an overall introduction by the editors and concludes with a historical note by D.R. Swanson. Additional references are provided for each section, which includes papers in the collection. For a list of included papers, see the publishers' web page: http://www.mkp.com/books_catalog/1-55860-454-5.asp.

In the preface, the editors provide their rationale for selecting papers for inclusion. With a view to motivate new practitioners in the field, they
have chosen original works over surveys, classic citations over papers with a pedagogic orientation, and time-tested research over current research.  They have also sought a balance between depth and coverage. For the most part, they stay loyal to these goals.

I was surprised to find that the editors present hypertext in a negative light as unimportant and having the potential for cognitive overhead,
rather than highlighting the research challenges and opportunities therein.  Vannevar Bush's classic 1945 article, "As we may think," is cited but not included. There is no mention of Ted Nelson's work on hypertext except as a citation in one of the included papers. Probably, my surprise is due to the success of the Web.

Their choice to eschew recent works is rather unfortunate, considering that this is an applied field making a tremendous impact in the Information Age of the nineties. The extensions section does not do justice to the highly interdisciplinary nature of current research in this area or to the content diversity. Due to the bias towards "classic citations," the reader is not introduced to most of the challenges in multimedia information retrieval.  For example, the reader is not introduced to the fact that although metadata-based retrieval effectively leverages decades of earlier research in textual information retrieval within the limitations of today's technology, it is not the only way.

Most of the papers in the first six sections are original classics.  However, it would be difficult to classify the papers in the extensions section, on hypertext, spoken text retrieval or video as original classics.  The extension section has a strong textual linguistic focus. The paper on news story categorization is an application of ideas presented earlier and could have been replaced by more contemporary work on news video.

The concepts of initial data entry, resource discovery, collaborative indexing, hypermedia and agent technology do not get adequate coverage. I
could not find other recent concepts like inclusion of support for retrieval at content creation time, and for content expiration, caching, and versioning. Recent work on retrieving relevant portions of large documents is missing.

I was disappointed that the editors do not describe the state of the art in the section introductions, nor do they provide future research directions.  An electronic companion to the book on the Web or on CD-ROM that includes pointers to information and software resources will be a
great asset.

Like most collections of this nature, the papers in this book have been photographically reproduced from their original sources. As a result, typographic conventions vary a lot and some papers are in landscape mode. I was almost tempted to cut the book along its spine and staple each paper separately. There is also an earlier collection which has the same title.  (Readings in Information Retrieval, Harold S. Sharp, ed., Scarecrow Press, New York, 1964.)

This present collection can be used as the nucleus of a single semester course on information retrieval in a seminar format. The collection does cover a wide range and provides a good overall appreciation for the field, but at the same time, it can not play the role of a textbook which would be more appropriate for most of the historic material covered. The teacher using this collection can provide a contemporary bias by supplementing more recent work, a wider coverage by adding a survey component and hands-on experimentation using public-domain software tools. I recommend this collection to instructors and libraries.

Wide Area Network Design: Concepts and Tools for Optimization
Robert S. Cahn
Morgan Kaufmann Pulishers, 1998, ISBN 1-55860-458-8.

Reviewed by Ayan Banerjee.

This book explains the design methodology of wide area networks. It can be divided into three main parts. In the first section, the reader is introduced to the general problem of network design and to the various parameters that are to be accounted for while designing networks. In the
second section, a number of design tools are presented based on the assumption that the network can be designed without any constraints. In the third section, approaches and algorithms are presented for network design with constraints.

The first two chapters introduce the reader to a generic network design problem and show that heuristic algorithms need not often lead to good
designs. Chapter Three initiates the reader in understanding the cost and delay tradeoff in the designs. It shows that a weighted combination of the Minimum Spanning Tree or Tour design algorithm with the Star design algorithm provides better results. Chapter Four focuses on providing models for traffic and tariff tables over a variety of traffic types (e.g., electronic mail, web page access, and connection-less user sessions) and cost functions (e.g., distance-based linear costing and fixed costs) respectively.

The next three chapters cover the local access design problem. Chapter Five talks about one-speed, one-center design using capacitated access trees. Chapter Six extends the algorithm to account for multi-speed local access. Chapter Seven is a generalization of the previous two chapters focussing on the multicenter local access design. It also discusses the modifications necessary to the base algorithms to produce designs for more complicated but practical constraints (e.g., access trees with a maximum number of nodes, central site less than some degree, and access trees with a maximum number of hops). The software package gives a hands-on experience with the design process.

Chapter Eight discusses backbone network design. An algorithm called MENTOR is presented. The interaction between the routing algorithm and capacity evaluation is highlighted by showing that if routing is not taken into consideration in the design, modification of the network could potentially lead to worse performance. Chapter Nine is a summary of all the previously-discussed algorithms that have been included in the MENTOR algorithm.

Chapter Ten deals with network design with constraints. A host of constraints are given (e.g., hop constraint, node-pair constraint, equipment cost constraint, degree constraint, performance constraint, and reliability constraint), and the modifications to the basic algorithms necessary to meet those constraints are also stated. Finally, Chapter Eleven deals with redesigning networks with increasing traffic flow and/or merging of companies.

The book is easy to read and each chapter builds upon previous ones. There is, however, not much reference to other work in the area. Designing data networks is the primary emphasis, and there are few examples in the book that address voice networks. The network design software package provided with the book is likely to be useful to students, network architects, and network designers. Exercises at the end of the chapters are insightful. The reader gets the feeling that the optimal result may not always be obtainable, but it is possible to search for fairly good designs in the correct region of the solution space.

The book gives detailed coverage of the algorithms that it discusses and highlights the variations on them to solve other design problems with
different constraints. It thus illustrates the methodology and prepares the reader to tackle other problems with more complicated constraints. However, the book does not help in developing a combined strategy for routing, security, and network management for wide area networks. An integrated algorithmic approach illustrated in a running example would be beneficial for the reader.

Switching in IP Networks: IP Switching, Tag Switching, and Related Technologies
Bruce Davie, Paul Doolan, and Yakov Rekhter
Morgan Kauffman Publishers, 1998, ISBN 1-55860-505-3.

Reviewed by Nicholas Shectman.

I picked up this book expecting to find a scholarly discussion of the advantages and disadvantages of tag switching and perhaps a philosophical consideration of network layers. What I got was a textbook. As I read on, however, I found that it was a pretty good textbook, and if I were teaching a beginning graduate - or perhaps advanced undergraduate - course in networking I would certainly put it on the syllabus.

In 232 large-print, easy to understand pages, this book lays out how various label-switched networking schemes work, although it's light on why.

It manages to do this while managing to address a wide range of familiarity with the subject and without unduly confusing the uninitiated or insulting the experienced. Only occasionally does it have to resort to asking the reader to skip ahead if he already knows, for instance, what an ATM cell is. Instead the authors hold the interest of more advanced readers through good, readable style. The text also gradually cranks up the level of discussion, and the assumptions of background knowledge, as the book progresses, to follow the progress of the students reading it (although for an undergraduate course the progress assumed is probably too high). Thus the text achieves its textbookish goal of leveling the field of understanding of the students who read it, so that the lecturer can get on with the teaching.

And teach you will have to do. This book leaves off or glosses over the interesting stuff, such as management, scalability, and pathology concerns, and includes only some of the security aspects. The technical progression of the book also results in earlier subjects being treated with less depth, which you may want to remedy. Fortunately, the text mostly avoids taking any political stances in its treatment, and so leaves open a wide range of further topics - QoS, for example - for discussion.

The authors avoid too much jargon but do tend to use acronyms without explanation - a tendency which is easily enough addressed by starting the course with a more general networking text. Any inconsistencies in style or content between the multiple authors are more than made up for by the fact that all three editors managed to write to the same audience, making this all in all a very effective textbook.