Digital Preservation

Best Practices Guides: A Typology

After establishing the general issues of digital preservation, we can proceed to developing a typology for comparing and categorizing current best practices documents based on those issues. It is possible to identify eight focus areas addressed in the literature:

As part the of the literature survey, a number of guidance and best practices documents were identified as noted in the bibliography and using the above typology, their recommendations categorized. Although this survey not exhaustive, the goal is to identify key documents representing either the advice of organizations leading the digital preservation enterprise or commonly recommended in discussions of digital preservation. It is also important to note that most of the best practice guides are recent and the recommendations may not have empirical support yet. Indeed, the Arts and Humanities Data Service notes a lack of information about how standards and methodology may be applied effectively (AHDS, 2001).

Almost all guides identify the primary problems associated with digital preservation as media deterioration/obsolescence and technology obsolescence. In particular, the NINCH guide suggests preservation faces two types of long-term accessibility challenges: machine accessibility (bit integrity) and human accessibility (semantic integrity) (NINCH, 2002). Similarly, the Visual Arts Data Service guide distinguishes between physical reliability and continued usability (Grout, Purdy and Rymer, 2002). There is also general consensus on digital archiving or preservation as a continuous activity in the form of a series of managed activities (RLG/OCLC, 2002) or as a lifecycle management approach (Beagrie and Jones, 2002). On the other hand, the source of digital objects is an area of divergence. A number of guides focus on digitized surrogates of physical objects while others encompass both digitized and born-digital objects with the preservation recommendations biased accordingly.

In the area of organizational recommendations, the pressing need for institutional policies on digital preservation is king. As Howard Besser argues, "our community needs to develop a concrete set of guidelines that can be used by people and organizations wishing to make information persist" (Besser, 2000). NINCH goes even further to suggest that "[i]nsufficient institutional commitment to long-term preservation can create digital resources with limited sustainability" (NINCH, 2002). Beyond the creation of policies, digital preservation needs to be incorporated into the organization as a whole, as in the case of the National Library of Australia where digital preservation is "part of existing core operations and systems … [as opposed to developing] a special or separate undertaking requiring its own infrastructure" (Gatenby, 2000). Finally, most guides reference OAIS as the model for developing a digital preservation architecture as noted in the literature survey.

There is relative uniformity identifying the two dominant strategies in digital preservation as being emulation and migration. However, some guides only identify these two strategies while others present them as two in a wider spectrum of possibilities. In most cases though, there appears to be an implicit acceptance of migration as the primary strategy, ranging from presenting migration as the only strategy (Hodge, 2000) to identifying a list of strategies but explicitly recommending migration (NINCH, 2002). The Cedars guidance documents are unique in presenting a more nuanced migration with an emphasis on the retention of the original digital object supplemented by migration upon request (Cedars, 2002). In contrast, the NINCH guide suggests migrating with every version of a format (NINCH, 2002), an undertaking requiring the application of greater resources over the long-term. One reason given for why migration is favored is that emulation-based approaches are experimental (Besser, 2000); while emulation may be the best hope for complex digital objects in the future, there are few institutions with the technological expertise to create emulators in the short term. In general, there may not be a best strategy but rather efforts should be focused on refining existing strategies (Kenney, 2000) with an emphasis on providing a suite of digital preservation tools.

While there is not a clear answer as to which strategy to choose, there is agreement that all digital preservation strategies require that the digital bits be available for future use. To this end, refreshing is identified as the best practice for long term machine readability. Refreshing can be implemented in a number of ways from immediately moving files onto a common media maintained for all digital objects (Cedars, 2002) to more traditional approaches emphasizing environmental stability and routine maintenance and migration of the media (Grout, Purdy and Rymer, 2000).

In comparison to strategies, there is less consensus in recommending specific media or formats. This can be attributed to the need to support the significant properties of a broad range of artifacts, often requiring specific formats and media. Even with more general issues such as whether or not to use a compressed format, there is little consensus. While some (NLC, 1998) are explicit about holding only non-compressed items, others (Grout, Purdy and Rymer, 2000) only recommend the format be lossless to retain maximum fidelity. One argument for why compression of any type (including lossless) is problematic for archival files is the introduction of an additional level of complexity (Besser, 2000) which only serves to increase the recovery/migration problem in the future. One recommendation that is agreed upon is the use of a standard format that is non-proprietary (NLA, 2002, Kenney, 2000) as these kind of formats are more likely to have a preservation path in the future.

One recommendation to handle media issues is to create backups (Cedars, 2002, Kenney, 2000, IMLS, 2001, NINCH, 2002) using more than one kind of backup software to write the copies so as to safeguard against software bugs. In this scenario, at least one copy should be maintained in an offsite location and the media periodically checked as per the refreshing methodology.

There is general recognition of the importance of metadata in an overall strategy for digital preservation (NINCH, 2002, Hodge, 2000). However, as there is no single standard for preservation metadata widely accepted, many organizations and projects (Cedars, 2002, Gatenby, 2000) have created their own schemes for local use. Fortunately, there are enough common factors between the local schemes that convergence should be possible (RLG/OCLC, 2001) with appropriate crosswalks to convert existing metadata to the emerging standards.

For resource discovery metadata and structural metadata, there are emerging standards which have some degree of consensus. Most guides that identify the need for structural metadata recommend the usage of the Metadata Encoding Transmission Standard (METS) (RLG/OCLC, 2001, Cedars, 2002, NINCH, 2002) while Dublin Core is often recommended for resource discovery (Hodge, 2000, IMLS 2001, Jones and Beagrie, 2002, Grout, Purdy and Rymer, 2000, Cedars, 2002). However, identifying these standards does not imply endorsement and the relative newness of METS (2002) means that implementation guides may not be available for some time.

The need for a persistent identifier to track the object is raised by a number of guides, with proposed systems including PURL (IMLS, 2001, Gatenby 2000, Hodge, 2000, Kenney, 2000), DOI (IMLS, 2001, Hodge, 2000, Kenney, 2000), ISBN/ISSN (IMLS, 2001) and local persistent identifiers (Cedars, 2002). What is interesting is the surprising number of guides (NINCH, 2002, Grout, Purdy and Rymer, 2000, Jones and Beagrie, 2002) that make no mention of the need for a system of persistent identification. Related to the persistent identifier is the need to ensure that the object has not been corrupted or altered. Unfortunately, while the need has been identified, little work has been done except for verification of authenticity at a bit level. For instance, recognition between authentication (integrity of the record) and authenticity (the quality and context of the record) is made (Jones and Beagrie, 2002) but typically only authentication of the object recommended with the common practice to calculate a value such as a checksum to ensure bit-level integrity (Kenney, 2000, IMLS, 2001, Jones and Beagrie, 2002). It should be noted that even for this recommendation, the issue of standardization is problematic as checksums can be computed through a number of algorithms and verification requires the algorithms be identical for the checksum to be useful.

The issue of intellectual property rights and rights management is consistently raised with the primary emphasis on ensuring that institutions have sufficient rights for digital preservation activities. Current and proposed changes to copyright laws may prevent preservation either by preventing copying or preventing modifying the digital object (Besser, 2000, Gatenby, 2000) and represent a possible future barrier to preservation. Ideally, an arrangement needs to be made with the copyright holder but as noted in the experience of the National Library of Australia (Gatenby, 2000), the amount of work this represents can be daunting.

Finally, there has not been much discussion in terms of how much these recommendations will cost or even the overall cost of digital preservation. The CAMiLEON project does identify types of costs but the general consensus is that future costs are currently unknown and cannot be well predicted.