Only a few institutions have made a large-scale commitment to Fedora, but the comminity is continually growing (and this list is probably outdated).
University of Virginia
The UVA collections are a great example of Fedora use. Note that some media content is only available to people on the UVA campus.
Some useful information:
- The Content Model Slides have some information about
UVA content models, and the Fedora Best Practices are good, too.
- The Content Model web pages have more details, but (as of May 2006) they are somewhat out of date.
- The UVA metadata group has some details on their metadata standards.
- UVA's indexing/delivery system is based on Fedora, Cocoon, XPAT, and stylesheets. Their Cocoon apps always produce pages with absolute URLs, so they don't have to worry about redirecting disseminators.
- UVA uses "uva-lib" for their object PID prefix, and "uva-lib-bdef" for behavior definition PIDs.
- They use IRIS for cataloging, and translate IRIS's output into uvaDescMeta format.
- They're using MySQL for storage of Fedora data.
- They store most (all?) digital content as referenced, rather than managed by Fedora for speed of ingest when they make changes. It is stored on a single (very large) filesystem.
- They consider the metadata to be more important than the digitized object. This means that when you search for something, you find the metadata object, but the media files are usually separate Fedora objects that are children of the metadata object.
- Their object hierarchy looks like:
- Aggregation object (roughly equivalent to our collection)
- Work object (metadata only)
- Media object
- Disseminators are directly derived from end-user requirements.
- A large amount of their repository's structure is based on complex (and hardcoded) workflows.
National Science Digital Library
The NSDL has the world's largest Fedora installation, with well over 2 million objects.
The detailed description of their implementation provided a lot of useful information:
- Tufts is running Fedora on multiple machines. They have a load balancer, but it is not clear whether the repositories contain
separate or overlapping data.
- The Naming Service looks interesting; how does it differ from a PURL server?
- They're using MySQL for storage of Fedora data, but Oracle for searching.
- Using a drop box to manage ingests from other parties. Items placed in the drop box are validated before any other work is done.
- Their behavior definitions are very similar to UVA's.
- Tufts uses "tufts" for their object PID prefix, and "bdef" for behavior definitions.
- A Tufts Fedora server is available for direct access.
- Some sample Tufts objects:
- Image Object (need to find one)
- Sample TEI Object (need to find one)
- Sample EAD Object
Documentation for their content models
The Encyclopedia of Chicago is a nice site. It has some great page-turning and image-viewing tools. The production site is not based on Fedora, only the cataloging version. The Fedora repository is exported via a chain of disseminators to create the production site.
- Digital Case
- Based on Fedora, with Aware for JPEG 2000 delivery
- Uses handles, both at the item level and the datastream level (itemID/datastreamID)
- Doesn't expose actual Fedora interface, it's all managed by external servlets.
The RepoMMan Project at the University of Hull is starting to use Fedora heavily.
Library of Congress – I Hear America Singing is a fairly complex project. It's unclear exactly how much of the content is in Fedora, and how much is "regular" HTML.
The VTLS VITAL product is based on Fedora.
The ARROW project is starting to develop a large Fedora implementation in conjunction with VTLS, but they don't seem to have anything completed yet.