Using IODyne

Illustrations and examples of:

IODyne toolbar
Information object agency
Search documents
Keyword and keyword-in-context tools
Creating search parameters
Hold files

Eric H. Johnson
ejohnson@uiuc.edu

Introduction

IODyne (pronounced "iodine") is an Internet client program which allows you to retrieve information from servers by dynamically combining information objects. Information objects are abstract representations of bibliographic data, typically titles (or title keywords), author names, subject and classification identifiers, and full-text search terms.

IODyne is being developed as part of a library research project, and so the emphasis on bibliographic data. The retrieval paradigm described here can be extended to other kinds of data as well.

IODyne embodies several key ideas:

• simultaneous connection to and querying of multiple servers, regardless of the query language or protocol required by each;
• abstraction of commonly understood kinds of searches (title, author name, subject, etc.) from the particular representation of the searched data on the server. Regardless of the particular bibliographic server you are connected to, you always do the same kinds of searches in the same way;
• construction of searches as persistent, arbitrarily complex, storable, and recallable objects, rather than the end result of a series of steps that cannot be recalled;
• caching of search results in the client, as well as set combination (e.g. set union, intersection, and difference) of search results in the client (various kinds of client-server load-balancing optimizations are possible, doing some kinds of set combinations in the client, and others in the server);
• normalized internal representation of abstract queries to facilitate set combination and other set-theoretic operations such as query subset recognition;
• a modeless, visible, multiple-windowed user interface in which, rather than learning commands, you learn how information objects behave and how to manipulate information objects. You do information retrieval in IODyne by moving information objects among different information spaces, and combining information objects in various ways in these spaces;
• visual, navigable, and simultaneously viewable information spaces, such as subject thesauri and classification schema, to facilitate search term suggestion.

Underlying all these ideas is the desire for the best use of each computing resource involved in getting information from the source to the user. The client-server paradigm is a sound one, but the prevalent use of World Wide Web browsing software and server-side CGI binaries does little to use the power of client machines to their best advantage, particularly in the realm of information retrieval. We are now in the situation where the desktop computers commonly used for Web browsing, though they rival the capacity and speed of the servers from which they retrieve information, are essentially being used as dumb terminals. Granted that much of this power goes toward supporting nice user environments and presenting multimedia content, there is much more that can be done by client machines to enhance the informational aspects of client-server interaction and take much of the query processing load off the server.

As for the user interface, HTML forms provide only the barest level of interactivity for the user of an information retrieval system. Server-intensive processing must occur between each stage of retrieval feedback to the user, not only to process the query but to generate the HTML required to display the results as well as the particular controls required for the next stage of query refinement. This is another artifact of the outdated host-terminal paradigm, where the host must compose the terminal’s screen, which impinges on client-server interaction in WWW-based information retrieval. Server administrators have enough to worry about without also having to write and maintain code for user screens.

The presentational power of desktop user environments, especially those which support drag-and-drop operations, allow for far higher levels of interactivity than is possible with HTML forms. For an information retrieval environment, a useful interface would allow you to literally "draw" queries on the screen, send any parts of theses queries to any number of connected servers at any time, would give you different kinds of feedback in multiple windows, etc. Though well within the capabilities of any currently available desktop PC, such an environment is all but impossible to implement in HTML, though with "frames" available in most Web browsers you might be able to do it in a limited and unsatisfactory modal sort of way.

The HTTP protocol is another obstruction to good information retrieval interaction, in that it imposes far too much overhead for the needs of query and relevance-feedback interchange. In such cases, connection-based client-server communication, with lower-overhead protocols tailored to the needs of the particular kind of data exchanged, are far more desirable.

None of this is to say that Web browsers have no place in an information retrieval environment. Web browsers are fine for displaying documents once they have been located, and for moving between documents in the browsing sort of way allowed by hypertext. This was the intent when Mosaic was first developed. In the interests of maintaining an open standard for digital library documents, the IODyne client will support both HTML and SGML displayers through CCI, as well as other kinds of viewers applicable to particular collections being accessed.