By definition, federated search involves the simultaneous search of multiple databases or collections. In intranets, it runs a single query across the staff directory and the subsites of multiple business units. On the Internet, it takes queries where crawlers can't by tapping into dynamic databases of the deep Web. In libraries (Figure 4-15), it lets users search multiple catalogs, collections, databases, and websites all at once.
Federated search may be necessary when managing dynamic content from multiple sources with different data models, but it does present challenges. First and foremost, performance is notoriously slow. As the number and size of collections grow, speed suffers. This is a serious flaw. Plus, the query language may be limited to the lowest common denominator in the face of disparate vocabularies and data models. Sophisticated approaches like faceted navigation may be precluded by federated search.
So, before committing to this solution, it's worth reframing the problem. When content is scattered into silos, users don't know where to search. Fragmentation is the root cause. Federated search addresses this problem, but it's not the only way. Instead, we can defragment the content by pulling it all into an integrated index. This may enable fast, powerful, unified queries, but is it still federated search? At this point, our definition needs clarification. Generally, when we talk about "multiple databases or collections," there's a symmetry between the front and back ends. But this needn't be true. Mirroring the technical architecture within the user experience is a design decision. Is it valuable for users to know these resources are from different sources? Is source a more important distinction than topic, format, or date? Will users need to define subsets of collections to search or execute more powerful queries using the native interfaces of individual databases? If the answer to these questions is, "No," federated search may become an antipattern that adds drag to the back end and visual clutter to the front.
That said, federated search can serve as a stepping stone by helping people visualize and play with possible and desirable outcomes. At the Library of Congress, the prospect of a single search across multiple collections was met with some skepticism. Federated search was a first step to get the ball rolling. However, the very next step may involve moving beyond federated search by unifying indexes and embracing a faceted navigation user interface that emphasizes topic and format more than source. Similarly, the alpha version of Boxee (shown in Figure 4-17) is an intriguing first step. Boxee is a social media center that supports multiple sources and formats of videos, music, and pictures from desktop computers, broadcast networks, and the Internet. It's a web-based service designed for big-screen television, and it's a radically federated solution that makes us think differently about how we might find and share media in the not-so-distant future. Boxee is also pretty difficult to use, in part because it spotlights sources rather than focusing on user-friendly ways to search and browse. It's a stepping stone, not a bridge.
In summary, federated search is important because users don't know where to search. Whenever content is fragmented into multiple locations, this pattern merits discussion. However, it's important to focus on the goal, not fixate on the pattern. If it's possible to unify the indexes, that may be the right solution, even if it's not technically federated search.
If source isn't relevant to users, there's no need to highlight it in the results. It may suffice to list source on the detail pages or include it as a metadata field within the faceted navigation display. Indeed, federated search must be carefully integrated with other patterns. Autocomplete and best first will need to draw suggestions from multiple sources. Advanced search may enable database selection, while the complementary modes of ask and browse may be harder to align across multiple sites. But no pattern is more closely tied to federated search than faceted navigation, since it offers users a great way to see and select sources within an integrated model of clarification and refinement.
