Library Trends, v.54, no.1 2005 Practical Preservation: The PREMIS Experience Priscilla Caplan and Rebecca Guenther Abstract In 2003 the Online Computer Library Center (OCLC) and Research Libraries Group (RLG) established an international working group to develop a common, implementable core set of metadata ele- ments for digital preser vation. Most published specifi cations for preservation-related metadata are either implementation specifi c or broadly theoretical. PREMIS (Preservation Metadata: Implementa- tion Strategies) was charged to defi ne a set of semantic units that are implementation independent, practically oriented, and likely to be needed by most preservation repositories. The semantic units will be represented in a data dictionary and in a METS-compatible XML schema. In the course of this work, the group also developed a glossary of terms and concepts, a data model, and a typology of relationships. Existing preservation repositories were surveyed about their architectural models and metadata practices, and some at- tempt was made to identify best practices. This article outlines the history and methods of the PREMIS Working Group and describes its deliverables. It explains major assumptions and decisions made by the group and examines some of the more diffi cult issues en- countered. Introduction In 2003 the Online Computer Library Center (OCLC) and Research Libraries Group (RLG) established an international working group to de- velop a common, implementable core set of metadata elements for digital preservation. Most published specifi cations for preservation-related meta- data are either implementation specifi c or broadly theoretical. PREMIS LIBRARY TRENDS, Vol. 54, No. 1, Summer 2005 (“Digital Preservation: Finding Balance,” edited by Deborah Woodyard-Robinson), pp. 111–124 © 2005 The Board of Trustees, University of Illinois 112 library trends/summer 2005 (Preservation Metadata: Implementation Strategies) was charged to defi ne a set of metadata elements that are implementation independent, practi- cally oriented, applicable to all types of materials, and likely to be needed by most preservation repositories. In addition, it aimed at establishing best practices for the implementation of preservation metadata. The stated PREMIS objectives were to • defi ne an implementable set of “core” preservation metadata elements, with broad applicability within the digital preservation community; • draft a data dictionary to support the core preservation metadata ele- ment set; • examine and evaluate alternative strategies for the encoding, storage, and management of preservation metadata within a digital preservation system, as well as for the exchange of preser vation metadata among systems; • conduct pilot programs for testing the group’s recommendations and best practices in a variety of systems settings; • explore opportunities for the cooperative creation and sharing of pres- ervation metadata. It was intended that PREMIS would build on the earlier work of an- other initiative sponsored by OCLC and RLG, the Preservation Metadata Framework Working Group (OCLC, 2003). That group was convened in 2001–2002 to develop a framework outlining the types of information that should be associated with an archived digital object. Their report, A Metadata Framework to Support the Preservation of Digital Objects (OCLC/RLG, 2002), expanded the conceptual structure for the Open Archival Informa- tion System (OAIS) information model (Consultative Committee, 2002) and mapped preservation metadata elements to that conceptual structure. Although the framework proposed a list of metadata elements, it did not contain suffi cient detail for an implementer to actually use the metadata in a preservation system without considerable further specifi cations. The PREMIS working group was established to take the previous group’s work a step further: to develop a data dictionary of core metadata elements to be applied to archived objects, give guidance on the implementation of that metadata element set in preservation systems, and suggest best practice for populating those elements. OCLC and RLG established the working group in 2003, chaired by Priscilla Caplan of the Florida Center for Library Automation and Rebecca Guenther of the Librar y of Congress. Because the charge was practical rather than theoretical, members were sought from institutions known to be running or developing preservation repository systems within the cul- tural heritage or information industry sectors. Conveners paid particular attention to diversity of stakeholders. The group consists of representatives from academic and national libraries, museums, and archives; governments; 113 and commercial enterprises in six different countries. In addition, PREMIS includes an international advisory committee of experts periodically called upon to review progress and provide feedback. In order to accomplish as much of the charge as possible in a reason- able timeframe, PREMIS divided into two subgroups with different deliver- ables and strategies. The Core Elements Subgroup took responsibility for drafting the “core” preservation metadata elements and supporting data dictionary. The Implementation Strategies Subgroup was responsible for examining alternative strategies for the encoding, storage, and manage- ment of preservation metadata within digital preservation systems and for developing pilot programs to test the group’s recommendations in a variety of system settings. The work of both subgroups was conducted almost entirely by weekly conference calls, which was a challenge given that the group members were from time zones ranging from the western United States to eastern Australia. Fortunately, only one person had to get up in the middle of the night to attend! However, the sheer frequency of calls and the ambi- tious agenda created a sense of camaraderie among participants. Members quickly learned each others’ voices and mastered use of a wiki (a Web site that allows any user to add and edit content) set up for their use by the University of Chicago. The Core Elements Subgroup also held two face-to- face meetings to expedite their work. The two meetings, one in San Diego in Januar y 2004 and the other in Cambridge, Massachusetts, in August 2004, were highly productive and contributed to the sense of community among members. One of the group’s practices has been well received and might well be found useful by other initiatives. Ever y month a summar y of each subgroup’s activities is posted on the offi cial Web site at http://www.oclc .org/research/projects/pmwg/. For example, the Core Elements update for September 2004 reads: The group spent time discussing the differences between fi les and bitstreams and how the semantic units applied to them. It was pro- posed that there was a need for a new level called “fi lestreams.” This also related to previous discussions about embedded fi les. The group continued its discussion of environment elements and whether this information is dependent on fi le format information. It continued to defi ne what information is needed about the environment in order to render objects for the long term. Two new participants joined the group, one from DSpace and another from the Walt Disney Company. A workplan was developed to fi nish the data dictionary by December in anticipation of a fi nal PREMIS report by the end of 2004. Because of these updates, anyone interested in the PREMIS activity could follow the group’s progress, see what issues were under discussion, and simply be assured the work group was working. caplan & guenther/practical preservation 114 library trends/summer 2005 Implementation Strategies The Implementation Strategies Subgroup was charged with examina- tion and evaluation of alternative strategies for the encoding, storage, and management of preservation metadata within a digital preservation system. To fi nd out how preservation repositories were actually implementing pres- ervation metadata, the subgroup decided to survey repositories that were in operation or under development. Although their work was focused on metadata, the subgroup felt that the sur vey provided an opportunity to explore the state of the art in digital preservation generally, and questions were drafted to elicit information about policies, governance and fund- ing, system architecture, and preservation strategies as well as metadata practices. In November 2003 copies were sent by email to approximately seventy organizations thought to be active in or interested in digital preservation. The survey was also made available on the PREMIS Web site and announced on various discussion lists. By the end of March 2004, forty-eight survey responses were received from institutions developing or planning to de- velop a digital preservation repository. Sixteen of these respondents were contacted for more in-depth telephone interviews. Although several institutions known to be developing digital preserva- tion repository systems did not respond, the replies received appear to be reasonably representative of the state of the art in the winter of 2003–2004. Responses came from 13 countries and included 28 libraries, 7 archives, and 14 other types of organizations. Among the respondents were 10 national libraries and 6 national archives, showing heavy involvement in digital pres- ervation at the national level, particularly in Europe and Canada. Key fi ndings are summarized in the report Implementing Preser vation Repositories for Digital Materials: Current Practices and Emerging Trends in the Cul- tural Heritage Community (OCLC/RLG PREMIS Working Group, 2004), so they will not be repeated here. However, a few points are worth noting. First, there is very little experience with digital preservation. Twenty- two respondents claimed to have a preservation repository in some stage of production (as opposed to planning, development, or alpha/beta test- ing). However, only half of these appeared to have implemented an active preservation strategy such as normalization, format migration, migration on demand, or emulation. This list included four national libraries/na- tional archives and six institutions categorized as “other.” None was an academic library. This fi nding must color all other results, including those pertaining to metadata. Whatever practices were reported on the survey, apart from these eleven institutions the results refl ect repositories not yet in production or not yet implementing active preservation strategies. We do not have enough experience to determine whether the metadata these systems record or plan to record is adequate for the purpose. 115 Second, those engaged in digital preservation still lack a common vocabu- lary and, to a large extent, a common conceptual framework. Although most respondents claimed to have been informed by the OAIS reference model and to be at least partly compliant with it, there was substantial difference of opinion as to the meaning of OAIS compliance. Although OAIS has been praised for providing a standard vocabulary for basic repository concepts, it is clear that most of these terms have not been widely adopted in the com- munity, at least not in informal communications such as survey responses. In relation to metadata, most respondents were recording several dif- ferent types of metadata, and more than half were recording metadata in all of these categories: rights, provenance, technical, administrative, descriptive, and structural. Repositories appear to draw metadata elements from various schemes to suit their purposes. The Metadata Encoding and Transmission Standard (METS) (Library of Congress, 2005), NISO Z39.87 (Technical Metadata for Digital Still Images) (National Information Stan- dards Organization and AIIM International, 2002), and the OCLC Digital Archive metadata set (OCLC, 2002) were the only named schemes used by more than 20 percent of respondents. Overall, thirty-three different metadata element sets or rule sets were mentioned by at least one reposi- tory. In general, the survey shows a picture of a community trying to take advantage of prior work but not at the point of developing or settling on dominant standards. Core Elements Methodology The Core Elements Subgroup began its work by attempting to defi ne the word “core” for the purpose of developing a metadata element list and data dictionar y. After much discussion the group settled on a practical defi nition of core: those elements that a working archive is likely to need to know in order to support the functions of ensuring viability, renderability, understandability, authenticity, and identity in a preservation context. Ini- tially the group felt that all core elements should be considered mandatory by defi nition, but some fl exibility crept in with the acknowledgement that some elements are more core than others, and even necessary information cannot always be provided. The Core Elements Subgroup then started analyzing the recommen- dations of the earlier Preservation Metadata Framework Working Group related to Preser vation Description Information. This included “digital provenance,” or the documentation of events associated with the digital objects. Those members of the subgroup from institutions actively run- ning or developing preservation repositories mapped the elements from the framework to what was used in their own systems. It became clear that the elements detailed in the previous work (which themselves had been caplan & guenther/practical preservation 116 library trends/summer 2005 mapped to the OAIS information model) did not always correspond to elements implemented in practice and did not give adequate guidance on how to use them. However, the exercise was useful in providing a com- mon denominator for diverse implementations; the group discussed each element in conference calls to see where there was commonality in usage. Elements that emerged as being widely used across implementations were considered the beginning of a core element list. The group made the decision that the data dictionary it was developing would be wholly implementation independent. That is, the core elements would defi ne information that a repository needed to know, regardless of how, or even whether, it was stored. For instance, for a given identifi er to be usable, it is necessary to know the identifi er scheme and the namespace in which it is unique. If a particular repository uses only one type of identifi er scheme, say one that is internally defi ned and assigned, the scheme can be assumed, and the repository would have no need to record it at all. The repository would, however, need to know this information and be able to supply it when exchanging metadata with other repositories. Because of the emphasis on the need to know rather than the need to record or represent in any particular way, the group preferred to use the term “semantic unit” (meaning an atom of meaning) rather than “metadata element.” The data dictionary therefore names and describes semantic units. After drafting a preliminary data dictionary for digital provenance in- formation, the group began to consider technical metadata, or detailed information about the physical characteristics of digital objects. The group realized that it did not have either the time or the expertise to tackle format- specifi c technical metadata for various types of digital fi les. By scoping the work to include only that metadata applicable to all (or at least most) digital formats, the group was able to limit the work to a reasonable set of semantic units and leave further development to format experts. The group compiled a list of potential semantic units based on specifi cations for the proposed Global Digital Format Registry (GDFR, n.d.) supplemented by data elements used in the repository systems of members’ institutions. Each element on the list was then discussed at some length, and those found to be both useful and broadly applicable were added to the data dictionary. Data Model One of the hardest issues to tackle was the development of an accept- able abstract data model. A valid criticism of the earlier framework was that the document recommended metadata elements pertaining to many different types of things while giving no guidance as to what type of thing they applied to. For example, “Resource Description” included the subele- ment “Existing metadata,” an example of which was “a MARC bibliographic record.” Bibliographic records usually describe intellectual entities, such as books, sound recordings, and Web sites. Another element, “File de- 117 scription” (defi ned as “technical specifi cations of the fi le(s) comprising a Content Data Object”), would appear to apply to individual digital fi les. A third element, “Size of object,” might be taken to apply to the total size of a complex object (for example, a book made up of many page images) or to a single stored fi le. The lack of specifi cs as to what level of granularity of an object the elements applied to made the document diffi cult to actually use in metadata implementations. The data model was intended to accomplish three purposes. First, it would force PREMIS members to be rigorous in their thinking in the de- velopment of the data dictionary. Second, it provided a structure for ar- ranging entries in the data dictionary. Third, it would help implementers of the data dictionary understand how to apply semantic units. The data model was not, however, meant to imply any particular implementation of the semantic units in the data dictionary. In the PREMIS data model there are fi ve types of entities: intellectual entities, objects, agents, rights, and events. Although it is possible these defi nitions will change before the fi nal report, these entities are currently defi ned as follows: • An event is an action that involves at least one object, agent, and/or rights entity. • An agent is an actor associated with preservation events in the life of an object. • A right is an assertion of one or more rights or permissions pertaining to an object. • An intellectual entity is a coherent set of content that is reasonably described as a unit, for example, a particular book, map, photograph, or database. • An object is one or more sequences of bits stored in the preservation repository. There are four subtypes of the object entity: fi le, fi lestream, bitstream, and representation. The most diffi cult part of the development of the data model has been to appropriately identify, name, and defi ne these subtypes. Defi nitions in this article are slightly less elaborate than those in the actual data model, but they communicate the concepts effectively. Of the fi ve entity types, fi le is perhaps the most intuitive, as our defi ni- tion resembles that of common usage: a named ordered sequence of zero or more bytes known to an operating system and accessible by applications. Every fi le has a fi le format, defi ned as a specifi c pre-established structure of a computer fi le that specifi es how data is organized. A fi le may contain zero or more bitstreams and zero or more fi lestreams. A “bitstream” is defi ned as data within a fi le that cannot be transformed into a stand-alone fi le without the addition of fi le structure (headers, etc.) and/or reformatting in order to comply with some particular fi le format. A “fi lestream” is a contiguous set of bits within a fi le that can be transformed caplan & guenther/practical preservation 118 into a stand-alone fi le conforming to some fi le format without adding information or reformatting the bitstream. An example of a bitstream is an image embedded within a PDF; an example of a fi lestream is a TIFF image within a TAR fi le. A “representation” is the set of fi les needed to provide a complete and reasonable rendition of an intellectual entity. It can be thought of as the digital embodiment of an intellectual entity. Preservation repositories never store intellectual entities, but they may store representation objects. As an example, the fi nal PREMIS report is an intellectual entity. There will probably be PDF and HTML versions posted on the Web; many readers will download their own copies, but all copies will have the same authors, title, and content. If the report were archived in a preservation repository, at least one representation would be stored. This might, for example, be a single, specifi c PDF fi le. The PDF fi le will doubtless contain embedded graphics for tables and charts, which would be bitstreams. If the HTML version were archived, the representation might consist of three or four fi les—the HTML fi le and several GIF images. Perhaps the repository will want to bundle these fi les together for storage by creating a TAR fi le. That TAR fi le would then have within it three or four fi lestreams, which could be extracted into fi les at some later time. These distinctions are important because different semantic units of metadata apply at different levels. The intellectual entity may have an ISBN or technical report number, but the representation does not. The represen- tation may have an identifi er known to the preservation repository, but the intellectual entity does not. The fi le will have a fi le name and fi le format, the fi lestream will have a fi le format but no fi le name, and the bitstream will have no fi le name or fi le format, although it may have other format characteristics such as color space. The PREMIS data dictionary attempts to defi ne core semantic units pertaining to all subtypes of objects and events. Intellectual entities and agents are not addressed in any detail because they have been the focus of other metadata schemes and they do not present unique requirements in the digital preservation context beyond the minimum needed to establish relationships between these and other types of entities. At the time of this writing, the group was still exploring the extent to which rights and/or permissions should be described. Relationships are the other important part of the data model. Entities can be related to entities of different types (for example, objects can be related to agents) and to entities of the same type (for example, objects can be related to other objects). Just as there may be core semantic units generally necessary in the majority of preservation repository applications, there are core relationships that most preservation repositories will need to record. The relationships between objects, agents, and events constitute digital library trends/summer 2005 119 provenance. As Clifford Lynch wrote in “Authenticity and Integrity in the Digital Environment”: Provenance, broadly speaking, is documentation about the origin, characteristics, and history of an object; its chain of custody; and its relationship to other objects. The fi nal point is particularly important. There are two ways to think about a digital object that is created by changing the format of an older object . . . We might think about a single object the provenance of which includes a particular transforma- tion, or we might think about multiple objects that are related through provenance documentation. Thus, provenance is not simply metadata about an object—it can also be metadata that describe the relationships between objects (2000). Objects and Events Most of the semantic units in the data dictionary pertain to objects and events. Semantic units related to the object entity describe characteristics relevant to preservation management. It is assumed that data content ob- jects are held in the preservation repository and that associated metadata may be held in the repository, in external systems, or in both. Data dic- tionary entries for objects indicate the level at which the semantic unit is applicable: representation, fi le, and/or bitstream. Filestream is considered equivalent to fi le for the purposes of applicability. Semantic units associated with object entities include identifi ers, loca- tion information, and technical characteristics. In anticipation of the devel- opment of format registries such as the proposed GDFR, the data dictionary also contains semantics for referencing format registry entries. Similarly, it provides for basic software and hardware environment information and anticipates adding references to future environment registries. Figures 1 and 2 provide examples of entries in the data dictionar y. Figure 1 shows the defi nition of a “container” unit (fi xity), which has no data itself but serves to group together three related semantic components (messageDigestAlgorithm, messageDigest, and messageDigestOriginator). Figure 2 shows the defi nition of one of these semantic components, mes- sageDigestAlgorithm. Events are actions that involve one or more objects and may be related to one or more agents. The PREMIS report states that whether or not a preservation repository records an event depends upon the importance of the event in the context of that repository. It recommends using the semantic units related to the Events entity when recording actions that modify objects. Other actions, such as the copying of an object for backup purposes, may be recorded in system logs or an audit trail but not neces- sarily as an event. Most of the documentation about the digital provenance of objects is given in relation to events. Semantic units include event identifi er, event type (for example, compression, migration, validation, etc.), event out- caplan & guenther/practical preservation Semantic unit fi xity Semantic components messageDigestAlgorithm, messageDigest, messageDigestOriginator Defi nition Information used to verify whether an object has been altered in an undocumented or unauthorized way. Data constraint Container Object category Representation File Bitstream Applicability Not applicable (see usage note) Applicable Applicable (see usage note) Repeatability Repeatable Repeatable Obligation Optional Optional Creation/ Maintenance notes Automatically calculated and recorded by repository. Usage notes To perform a fi xity check, a message digest calculated at some earlier time is compared with a message digest calcu- lated at a later time. If the digests are the same, the object was not altered in the interim. Recommended practice is to use two or more message digests calculated by different algorithms. The act of performing a fi xity check and the date it occurred would be recorded as an Event. The result of the check would be recorded as the eventOutcome. Therefore, only the mes- sageDigestAlgorithm and messageDigest need to be recorded as objectCharacteristics for future comparison. Representation level: It could be argued that if a represen- tation consists of a single fi le, or if all the fi les comprised by a representation are combined (e.g., zipped) into a single fi le, then a fi xity check could be performed on the representation. However, in both cases the fi xity check is actually being performed on a fi le, which in this case hap- pens to be coincidental with a representation. Bitstream level: Message digests can be computed for bitstreams although they are not as common as with fi les. For example, the JPX format, which is a JPEG2000 format, supports the inclusion of MD5 or SHA-1 message digests in internal metadata that was calculated on any range of bytes of the fi le. See “Fixity, integrity, authenticity,” page 4-5. Figure 1. Data Dictionary Entry for Fixity library trends/summer 2005120 121caplan & guenther/practical preservation Figure 2. Data Dictionary Entry for messageDigestAlgorithm come, and event date/time. When properties of an object are the result of an event, this is considered event-related information, but in practice this may be recorded with the object or with the event. An example of a data dictionary entry for a semantic unit related to the Event entity is given in fi gure 3. PREMIS Report and Further Work The fi nal PREMIS report will go into greater detail about the fi ndings of the working group and will present a completed data dictionary with examples. In addition, it will include a glossary, a description of the data model, discussions of some of the more diffi cult or controversial semantic units, and other related information. As of this writing, the working group was still conducting work by conference calls and the data dictionary was not yet completed. The target date for completion is December 2004.1 Semantic unit messageDigestAlgorithm Semantic components None Defi nition The specifi c algorithm used to construct the message digest for the digital object. Data constraint Value should be taken from a controlled vocabulary. Object category Representation File Bitstream Applicability Not applicable Applicable Applicable Examples MD5 Adler-32 HAVAL SHA-1 SHA-256 SHA-384 SHA-512 TIGER WHIRLPOOL Repeatability Not repeatable Not repeatable Obligation Mandatory Mandatory Figure 3. Data Dictionary Entry for eventType Semantic unit eventType Semantic components None Defi nition A categorization of the nature of the event. Rationale Categorizing events will aid the preservation repository in machine processing of event information, particularly in reporting. Data constraint Value should be taken from a controlled vocabulary. Examples E77 [a code used within a repository for a particular event type] Ingest Repeatability Not repeatable Obligation Mandatory Usage notes Each repository should defi ne its own controlled vocabulary of eventType values. A suggested starter list for consideration (see also the Glossary for more detailed defi nitions): capture = the process whereby a repository actively obtains an object compression = the process of coding data to save storage space or transmission time deaccession = the process of removing an object from the inventory of a repository decompression = the process of reversing the effects of compression decryption = the process of converting encrypted data to plaintext deletion = the process of removing an object from repository storage digital signature validation = the process of determining that a decrypted digital signature matches an expected value dissemination = the process of retrieving an object from repository storage and making it available to users fi xity check = the process of verifying that an object has not been changed in a given period ingestion = the process of adding objects to a preservation repository message digest calculation = the process by which a message digest (“hash”) is created migration = a transformation of an object creating a version in a more contemporary format normalization = a transformation of an object creating a version more conducive to preservation replication = the process of creating a copy of an object that is, bit-wise, identical to the original validation = the process of comparing an object with a standard and noting compliance or exceptions virus check = the process of scanning a fi le for malicious programs The level of specifi city in recording the type of event (e.g., whether the eventType indicates a transformation, a migration or a particular method of migration) is implementation specifi c and will depend upon how reporting and processing is done. Recommended practice is to record detailed information about the event itself in eventDetail rather than using a very granular value for eventType. library trends/summer 2005122 123 Although the data dictionary is intended to be implementation neu- tral, for information to be exchanged between repositories there must be some standard representation. The implementation survey showed wide use of METS among implementers. The METS initiative intends to draft PREMIS-based XML schemas suitable for use as extension schemas for the digital provenance metadata section (digiprovMD) and technical metadata section (techMD) of a METS document. The digiprovMD will be based on the events section of the data dictionary. The new techMD section will complement the other format-specifi c technical metadata sections and will include general technical metadata that applies regardless of fi le format. It will be necessary to reconcile existing format-specifi c extension schema with this new general one, since some data elements that apply regardless of fi le format will already be included in defi ned-format specifi c techni- cal metadata extension schema (for example, MIX, the XML binding of the NISO/AIIM standard Z39.87, Technical Metadata for Digital Still Im- ages) (National Information Standards Organization & AIIM International, 2002). Opportunities for developing testbeds for implementing PREMIS- compliant metadata are currently under discussion, as are trials of the exchange of preservation metadata among repositories. It is unlikely that these will actually be implemented before the group is formally disbanded, so other mechanisms for continuing this work are being considered. Mecha- nisms for supporting the adoption of PREMIS metadata, gathering feedback and evidence of practice, and maintaining the data dictionary over time will also be necessary. The PREMIS Web site should be consulted for the status of these and other related activities. Note 1. Since this article was written, the PREMIS working group released Data Dictionary for Preser- vation Metadata: Final Report of the PREMIS Working Group in May 2005. It is available from the PREMIS Web site at http://www.oclc.org/research/projects/pmwg/. The Web site for PREMIS maintenance activity is http://www.loc.gov/standards/premis/. References Consultative Committee for Space Data Systems. (2002). Reference model for an Open Archival Information System (OAIS) (CCSDS 650.0-B-1). Retrieved March 8, 2005, from http://ssdoo .gsfc.nasa.gov/nost/wwwclassic/documents/pdf/CCSDS-650.0-B-1.pdf. Global Digital Format Registry (GDFR). (n.d.). Home page. Retrieved March 8, 2005, from http://hul.harvard.edu/gdfr/. Library of Congress. (2005). Standards: Metadata encoding and transmission standard. Retrieved March 8, 2005, from http://www.loc.gov/standards/mets/. Lynch, C. (2000). Authenticity and integrity in the digital environment: An exploratory analysis of the central role of trust. In Authenticity in a digital environment (Council on Library and Information Resources report). Retrieved March 8, 2005, from http://www.clir.org/pubs/ reports/pub92/lynch.html. National Information Standards Organization & AIIM International. (2002). Data dictionary: Technical metadata for digital still images (NISO Z39.87). Retrieved March 8, 2005, from http://www.niso.org/standards/resources/Z39_87_trial_use.pdf. caplan & guenther/practical preservation 124 OCLC. (2002). OCLC digital archive system guides: Digital archive metadata elements. Retrieved March 8, 2005, from http://www.oclc.org/support/documentation/pdf/da_metadata_ elements.pdf. ———. (2003). Preservation Metadata Framework Working Group. Retrieved March 8, 2005, from http://www.oclc.org/research/projects/pmwg/wg1.htm. OCLC/RLG. (2002). Preservation metadata and the OAIS information model: A metadata framework to support the preservation of digital objects. Retrieved March 8, 2005, from http://www.oclc. org/research/projects/pmwg/pm_framework.pdf. OCLC/RLG PREMIS Working Group. (2004). Implementing preservation repositories for digital materials: Current practices and emerging trends in the cultural heritage community. Retrieved March 8, 2005, from http://www.oclc.org/research/projects/pmwg/surveyreport.pdf. Priscilla Caplan, Assistant Director for Digital Library Services, Florida Center for Librar y Automation, 5830 NW 39th Avenue, Gainesville FL 32606, pcaplan@ufl . edu, and Rebecca Guenther, Senior Networking and Standards Specialist, Library of Congress, 101 Independence Ave. SE, Washington, DC 20540-4402, rgue@loc .gov. Priscilla Caplan is Assistant Director for Digital Library Services at the Florida Center for Library Automation, where she is managing a project to develop a digital preservation repository for the use of the public universities of Florida. She is the author of Metadata Fundamentals for All Librarians (ALA Editions, 2003) and numer- ous articles on digital preservation, metadata, reference linking, and standards for digital libraries. In addition to co-chairing the OCLC Working Group on Preserva- tion Metadata: Implementation Strategies, she co-chairs the NISO/EDItEUR Joint Working Party on the Exchange of Serials Subscription Information. Rebecca Guenther is Senior Networking and Standards Specialist in the Network De- velopment and MARC Standards Offi ce of the Library of Congress, in which she has worked since 1989. Previous positions included cataloger at the National Library of Medicine; cataloger in the Library of Congress’ Shared Cataloging Division/German Language Section, and section head of the National Union Catalog Control Section/ Catalog Management and Publication Division. Her current responsibilities include work on national and international information standards, including, among others, rotating chair of ISO 639 Joint Advisory Committee on language codes and a member of the NISO Standards Development Committee. Rebecca has worked in the area of metadata since the early 1990s, including maintaining a number of crosswalks be- tween various metadata schemes; participating in development of XML bibliographic descriptive schemas (MODS and MARCXML); serving as chair of the Dublin Core Libraries Working Group and as a member of the Dublin Core Usage Board; serving as a co-chair of PREMIS, an OCLC/RLG working group on preservation metadata implementation strategies; and participating in the Open Ebook Forum’s Metadata and Identifi ers Working Group, among others. She has published articles and made presentations widely on metadata and various standards-related efforts. library trends/summer 2005