The Zombie Stack Exchanges That Just Won't Die
Our library has been trying to envision what a single web portal into all of our unique digital and special collections offerings would look like. We've got two separate special collections and archives departments with two websites, a digital collections website, a digital exhibits page, and our ETDs, all in separate silos.
For us,
Exhibits are curated digital items representing a larger physical collection, lots of text, sometimes in tandem with a physical exhibit (think Omeka)
Digital collections are entire collections as-is, usually lots of items, little to no accompanying description (think CONTENTdm)
...not to mention that we'd like to in some way represent the collections that we haven't digitized.
If you know of institutions that are presenting a graceful single web interface into all of their unique collections, please send them along.
Erin White
Well, I know of "discovery layer" implementers who solve the problem through new-generation OPACs. The University of Wisconsin's new Blacklight-based catalog interface harvests metadata from the IR and digital collections via OAI-PMH for search and display, for example.
My two favorites in this are are Dukes digital collections site and the National Library of Australia's Trove. I love how trove draws attention to collection areas and to the daily activity of civizen librarians and archivists who are working on enhancing the quality of the collection. Duke's site has a great look to it as well. The Library of Congress recently rolled out a single search across all the collections as well and it has some nice features to it. In particular, I like how it responds to Null search requests, and how it defaults to faceting to digitally accessible content but with one toggle becomes a search of all the rest of the records as well.
Yale has one, though it's certainly not comprehensive of content at the University: http://discover.odai.yale.edu/ydc/.
In North Carolina's research triangle, UNC-Chapel Hill, Duke, North Carolina State, and North Carolina Central have a shared catalog that doesn't just has bibliographic records, but also indexes full text of finding aids from those institutions' archival branches (though not museums), and is also starting to include digital content (though not all yet). You can find this catalog at http://search.trln.org/.
I love the way how the world digital library exposes their assets on a map. You can brows by various facets such as topic, language, item type or institution; all results can be narrowed with a time slider - neat.
The Library of Congress has a "search box" that provides a gateway into content found in a variety of subsystems (the online catalog, Prints and Photographs catalog, Thomas, the Veterans History Project, American Memory, etc). The web application is built using Django and Solr, using metadata obtained from crawling the various websites with Apache Nutch and extracting known HTML metadata for indexing with Solr. There are also specialized loaders to bypass crawling, for datasets like the Voyager online catalog. For an example check out the search results for "whitman".