Case Study: Wide-Angle Searching Is a Web Service

Computerworld - Ness Technologies Inc.
Hackensack, N.J., and Washington
Who they are: IT services provider formed in 1999 from the merging of several Israeli companies.
Goal: To use its expertise with content management technologies from Documentum Inc. in Pleasanton, Calif., and Interwoven Inc. in Sunnyvale, Calif., to develop a multitier document-listing service that supports federated search - that is, the ability to search multiple sources simultaneously.
Strategy: Using a beta version of WebLogic Workshop from San Jose-based BEA Systems Inc.
Challenges: Wizards that hack away at the dense underbrush of supporting code are only part of the story. Web services that work robustly can't rely on the synchronous model of HTTP communication, which expects an immediate response to every request.
Issues: The tool kit's support for asynchronous and long-running calls, Ness says, does as much for a business developer's productivity as the wizards. Alan Zenreich, the company's director of Web services, amplifies the point: "The coarse-grained and loosely coupled design is best for scalability and availability," he says.
Given the intrinsic unreliability of the Internet as a message delivery mechanism and the stateless nature of HTTP, there's a need to create a "conversational context" for Web services, and WebLogic Workshop addresses that need, Zenreich says. It offers support for fire-and-forget messaging, queuing, registration of callback methods and managed polling for endpoints whose callback addresses are blocked by firewalls.
Payoffs: Simplifying Java 2 Enterprise Edition development is a key benefit of the BEA tool kit, according to Ness Chief Technology Officer Joe Fung. Freed from "plumbing" concerns, Zenreich has prototyped a service intended to support federated search. In this scenario, a Web-services-enabled client issues an asynchronous query to a search service that in turn issues asynchronous queries to several different Web-service-enabled document repositories. Maintaining context in each tier means that a client can refine a search without requiring the searcher to requery its sources, and the searcher can assemble results incrementally without waiting for all sources to return complete results.