IICE does not display search data until all searches are complete, which unfortunately puts performance at the mercy of the slowest back-end system. You may remove a given system from a search, but there is no option for displaying partial search results in the name of speed. In a simple demo across four back-end systems, we waited several minutes for a search to complete, because one repository was hosted on a severely underpowered platform.

Contributing to this slowness is the fact that IICE does not maintain an index of items found on the back ends, nor does it cache searches or results. IBM attributes this design to IICE's focus on completeness of search results, rather than speed. If fast searches are important, IBM recommends its

OmniFind edition, but I don't buy this either/or paradigm. Users would be well served by options that allows them to tune activities between performance and completeness.

After the results have been displayed on the console, users can modify the data. For example, highlighting an item in IICE can be saved to the CMS via IICE. The product also exposes administrative options to these repositories --most data manipulation capabilities that the system provides are passed through to the user in an integrated UI.

The underlying technology that makes this all possible is a Java package IBM obtained in its 2004 acquisition of Venetica. The software can run on its embedded J2EE server or on an existing IBM WebSphere or BEA WebLogic server. It integrates with the various repositories via connectors or Web services. Developers can write applications that interact with these connectors and even write new connectors for unsupported systems --an SDK is supplied for this purpose.