Long term preservation can be considered a relay of content over time, a series of hand-offs occurring repeatedly at many levels: between types of media and storage systems, object frameworks and software systems, different institutions and policy regimes, and different communities. As time passes, all links to the original creators and context will be severed, exacerbating the challenges. The National Geospatial Digital Archive (NGDA:  ) has identified three architectural design principles beyond OAIS ("Reference Model for an Open Archival System (OAIS)": ) which are necessary to preserve content over time: the "relay" principle, the "fallback" principle, and the "resurrection" principle.
The "relay" principle states that a preservation system should support its own migration.
In the event that the system itself is no longer functional, then a preservation system should support some form of hand-off of its content (this is the "fallback" principle).
Since the costs of continued migration of content is very high, a method of mitigating that cost is to allow content to become obsolete, but to support sufficient metadata and contextual information to be able to resurrect full access and use at some future time (the "resurrection" principle). Preserving any type of digital information requires preserving the information's context so that it can be interpreted correctly.
Clifford Lynch pointed out at the iPres2008 Conference  that we should preserve open source source code to document the standards we develop and to enable reconstruction of the content that we save.
Steps and Agreements
Whatever we do must be feasible, reasonable, scalable, and specific to our needs at the University of Alabama. The following are the current steps being taken:
In support of the Resurrection and Relay Principles:
- Given that at the time of discovery, the file system itself may be obsolete, one potentially important task is storing inode information and information about the parameters of the file system (type of partition support needed, for example) at the top level in a plain text or XML manifest. Some information that should be included are then number and type of characters (Unicode? Latin 1?) allowed in file names, and other expectations for support. Jason Battles, Head of Web Services, will determine what information should be collected and stored for the University of Alabama Libraries digital archives.
- Janet Lee-Smeltzer, Head of Cataloging and Metadata Services, will engage the Metadata Unit in:
- Creating a descriptive metadata file about the archive itself, to inform archivists of the future about what it contains, and its provenance
- Assessing which of our selected technical metadata fields (from MIX, for images:  and AudioMD for audio: ) cannot be generated via open source tools such as JHOVE () and FITS() , and thus must be captured in some other manner
- Determining what technical metadata (such as version and type of format, to inform migration needs) needs to be readily available for file management (stored in a database for easy access, periodically exported as plain text for storage with the files)
- Analyzing and rating PREMIS (Preservation Metadata: Implementation Strategies: ) recommendations as to highest priority, middle priority, and lowest priority, to assist us in determining feasible methods for implementing the portion of these recommendations which are critical to our success
- Jody DeRidder, Head of Digital Services, has developed a File:ArchiveMap.txt "map" of the file structure, and created some level of text documentation or metadata about each level.
- Digital Services personnel has selected appropriate open source tools for rendering current content, and will work with Jason Battles to determine the necessary open source operating system for support of these software tools.
- Jason Battles will determine how we should periodically test the selected software and operating system for verification that they will indeed reconstruct access to our content.
- Tom Wilson proposed that we test the selected software and operating system yearly; we have agreed to this as a policy, to ensure that the stored software is the most appropriate. We also agreed to store it on a top level in a single location, with instructions as to how it should be used, and to what files it should be applied.
In support of the Fallback Principle:
We continue to pursue use of LOCKSS (Lots of Copies Keeps Stuff Safe: ) via the Alabama Digital Preservation Network (ADPNet: ) as our primary method of "fallback" as this provides a method of handoff of content in the event our system fails completely.