We found interesting, useful search features in the latest version of Microsoft's Internet Information Server (IIS). The best feature required no testing whatsoever to discover; Microsoft bundles Index Server with its IIS product, which it in turn includes as part of NT Server. Do you have NT Server 4.0? Is IIS your Web server? You ha ve a search engine.

In the lab, Index Server accurately indexed Microsoft Office97 files and Adobe PDF documents in real-time, allowed full-text and HTML field searching by word or phrase, and integrated itself closely with NT Server. Index Server supports seven languages and several document formats, but not as many as Digital's SIX-97. However, Index Server lacks a natural language interface, and the supplied query interface doesn't support proximity searches or stemming. To its credit, Microsoft's indexer is "stem-aware," and the vendor has documented an IIS programming interface, which third-party software vendors can use to build query tools that incorporate stemming.

Offering an intriguing quality among the products we reviewed, Index Server 2.0 accepts SQL queries in addition to the traditional keyword and phrase arguments. IntelliServ indexes ODBC data sources but only allows keyword searching of the index. Index Server goes a step further to allow finding results via SQL statements. Query pr ograms deliver SQL statements to the ActiveX Data Object (ADO), which passes the request to Index Server for processing.

Microsoft says it added the SQL capability so programmers could develop applications (in Visual Basic, J++, C++, or VBScript) that could search a Web site's index. A small VBScript program we wrote successfully accessed the Index Server catalog, but our attempts to connect a variety of relational database query tools (such as Microsoft's Access97) failed. Nonetheless, the concept of SQL access to Web indexes is a good one, and we hope Microsoft develops the feature further in future versions.

Netscape Communications Corp. Netscape Compass Server 3.0
We liked Compass Server for its ability to support diverse document types, though it does not support as many as SIX-97. We also liked its easy administration--only Ultraseek Server and SIX-97 were easier to administer. However, Compass Server lacks stemming and a natural-language interface, and it produced more false matches than Ultraseek Server, IntelliServ and Microsoft Index Server did. Interestingly, Compass Server is based on the Verity search engine--the Netscape programmers built Compass Server from the same core of search functions Verity used to create IntelliServ.

If your intranet is running Netscape SuiteSpot, you're in luck: You have a useful search engine. Just as Microsoft includes Index Server with IIS, Netscape bundles Compass (formerly Catalog Server) in its SuiteSpot product. Because Netscape supports more platforms than Microsoft, you have a wider range of search engine hardware and operating system choices.

Without stumbling over the binary formats of the files, Compass Server indexed the Microsoft Office files on our intranet, as well as the Adobe PDF and WordPerfect files. In addition to HTML and plain text files, Netscape claims Co mpass Server can catalog the contents of Rich Text Format (RTF), PowerPoint, Interleaf, Lotus Ami Pro and FrameMaker files.

When we u sed the Compass Server search engine to hunt for data on our intranet, it let us specify words and phrases, with Boolean operators, and it performed phonetic lookups and synonym matching. This last feature helped us find university when we searched for college (but didn't give us film for movie). We located documents based on metadata entries we'd placed inside the HTML, and Compass Server let us segment our documents into topics (Netscape terms each topic a taxonomy). The resulting categorization helped make responses immensely more relevant, at the expense of hiding potential nuggets of information buried in other topics.

Compass Server's personalization feature pushes search results to you periodically in the form of a newsletter. In the lab, we painlessly created what Netscape calls a personal interest profile subscription. Thereafter, whenever we updated Web pages or documents, Compass Server's indexer automatically used our URL classification rules and taxonomy designations to prepare a My Compass n ewsletter, which it e-mailed--via our Simple Mail Transfer Protocol (SMTP) mail server--to the network address we specified. The newsletter contained Web pages meeting our search criteria.

We liked being able to schedule the newsletter profiler separately from the scheduled running of Compass Server's indexer. However, IntelliServ's implementation of search engine push technology was more useful because it gave us QBE refinement of search results.