Quite by accident we are featuring two articles with a common theme, “Just Do It.” The Oregon and Rotterdam experiences in addressing digital preservation illustrate the strengths and the drawbacks of such an approach. They also underscore the importance of testing the limits of “conceptually strong” solutions through practical experience.
The University of Oregon Libraries, like many libraries and other cultural heritage institutions, began to create and provide access to a variety of digital content about a decade ago. Some of the earliest efforts began in Special Collections & University Archives with the digitization of portions of the library’s holdings of more than 400,000 rare photographs. Scanning and metadata collection were done with little awareness of, much less adherence to, emerging standards. Digital files were stored on a variety of media, with no systematic backups, no checks for file integrity or media degradation, and little internal consistency in file naming, creation of digital images, or collection of metadata. In the space of a few years, some files were no longer usable, due to degradation of storage media or inadequate collecting and tracking of metadata. Other efforts, such as the E-Asia Digital Library or digital collections of aerial photographs, were developed with an awareness of and attempt to follow standards but with little institutional buy-in or support.
In 2000-2002, the Libraries took steps to address some of these problems by pulling together a group of stakeholders to study digital library issues and needs. A number of results came out of this Digital Library Initiative [1], including:
Guidelines for the design and use of a mass storage unit for digital content.
Standards for providing access to digital collections.
A decision to purchase the CONTENTdm® digital asset management system.
Formation of a Metadata Implementation Group (MIG) and a Software Implementation Group (later combined into one group) to develop metadata standards for digital collections and to implement one or more collections using the CONTENTdm® system.
The assignment of a group of functional experts, named the Digital Collections Gatekeepers, to monitor digital collection activity, make policy decisions and recommendations, and keep Library Administration apprised of needs and developments.
In late 2002, the MIG began to meet and hammer out agreements regarding descriptive, technical, and administrative metadata for a collection of photographs that had previously been digitized. We all became conversant in Dublin Core, we reviewed standards and practices being developed and utilized worldwide, and we tested software. We planned to migrate the digital content into the CONTENTdm® system, rescan some content to ensure it met our standards for digital images, and apply the newly approved metadata standards to this collection.
In March 2003, our plans took a very different direction. In what was to become a characteristic abandonment of fledgling plans in favor of a quick response to a critical need, the library identified a completely different collection of materials for digitization and mounting in CONTENTdm®. We created and made this collection publicly available in record time (four months) to fulfill our obligations to a granting agency. We achieved this by calling on the expertise of four librarians in what was then the Catalog Department, working closely with staff in Special Collections & University Archives (especially the Image Services Center), and recruiting staff of the Catalog Department to learn how to scan glass plate negatives, apply descriptive and technical metadata, and load the digital objects and accompanying metadata into the CONTENTdm® system. While this was underway, the library also established an institutional repository, called Scholars’ Bank, using DSpace software. I was involved in both efforts as the chair or co-chair of the groups working to implement the systems.
The die was cast. In December 2003, the Catalog Department was rechristened Metadata and Digital Library Services (MDLS) in recognition of its expanded role of implementing and maintaining digital collections—in addition to cataloging and preserving analog materials. We are responsible for the digitization of items, collection of metadata, upkeep of the software, coordination with and marketing to internal and external partners, and developing and maintaining the public interfaces to the collections. In the summer of 2004, MDLS absorbed the staff of the Image Services Center, with whom we had been working very closely. However, within the library, no new classified or professional positions have been added—all work on digital collections is an add-on to other ongoing work.
I have been concerned for some time about the sustainability of our model for building and preserving digital collections. In the spring of 2004, one of my colleagues and I requested and received funding from our library to attend the weeklong Digital Preservation Management workshop presented by Cornell University Library. In July 2004, he and I were part of a group of twenty-four people taking part in the Cornell workshop. Attending the workshop served three main purposes: first, it confirmed our fears that our model of digital collection development and maintenance was not sustainable; second, it gave us a clear understanding of why it wasn’t sustainable; and third, it provided practical tools for developing a sustainable program and becoming a trusted digital repository [2].
The Cornell workshop presented digital preservation needs as the three legs of a stool, all of which are needed to keep the stool balanced: Organizational Framework (policies and planning), Technological Framework (hardware, software, technical skills), and Resources Framework (staffing, funding, space) [3]. The Cornell model identified key indicators of the stages of development for each leg, ranging from stage one to stage five. During the course of the workshop, my colleague and I independently completed an exercise to assess our institutional readiness for digital preservation. We both identified the University of Oregon Libraries as being at stage one regarding the Organizational Framework, stage two on the Technological Framework, and stage one on the Resources Framework.
Stage one development in terms of organizational infrastructure is often characterized by nonexistent or implicit policies. Key indicators of stage two technological infrastructure are being project-specific and reactive. Being at stage one in the resources framework indicates that an organization’s resources devoted to digital preservation are generally low, finite, and ad-hoc. In addition to helping us identify problem areas, the workshop also provided a number of tools to enable us to move forward to the next stages of development.
What have we done with our knowledge in the intervening fifteen months?
Organizational and Technological Frameworks
Recognizing that we could have little short-term impact on the resources devoted to preserving digital collections, we decided to focus our immediate attention on improving the organizational and technological frameworks as much as possible. Immediately after returning from the workshop, I arranged to meet with our Library Administration to outline the challenges and propose some strategies for addressing them. I prepared a two-page document for them: laying out a statement of need, defining digital preservation, listing the attributes of a trusted digital repository, and proposing a series of next steps [4].
In order to get administrative buy-in to this approach, I suggested that we could undertake this effort with no additional outlay of staffing or financial resources—and with no reduction in our existing workloads. Naïve as this may sound, this is actually a defining characteristic of the way we do things at Oregon. We don’t let our lack of resources stand in the way of moving forward on issues that we identify as priorities; we just do it.
I began by calling together our Digital Collections Gatekeepers group. We met on August 20, 2004—the first meeting of the group since it was named in the fall of 2002. Prior to that meeting, my colleague, Corey Harper, and I prepared some background documents and suggested some reading for group members. We also worked with a member of the Systems Department and revised the guidelines for the mass storage unit (MSU) that had been drafted initially as part of the Digital Library Initiative in 2000-2002.
At the first meeting of the group, we discussed the overall framework for preserving digital content, using the Digital Preservation Snapshot document that Corey and I had prepared. The group agreed on what the next steps needed to be. We also discussed what we meant by digital content and agreed to define it broadly to include any digital content that we acquire, license, or create and that we want to retain or preserve. We agreed that we wanted to look at including digital preservation strategies (such as LOCKSS) in our licensing agreements and to use the OAIS Reference Model to help us evaluate vendors from whom we license or buy content, as well as using it for self-evaluation.
We also discussed the role of the Gatekeepers and other individuals and units within the library. We agreed to hold open meetings once a month, as well as to conduct business between meetings via an archived discussion list. At this first meeting we also looked at digital collections that were already in progress. This review provided a clear indication of how quickly collections can proliferate and how important it was for this group to become involved in helping to set a realistic framework for new projects. We renamed the group the Digital Content Coordinators (DCC). The Digital Content Coordinators were subsequently charged by Library Administration to: 1) develop a robust digital preservation strategy for the UO Libraries and work with Library Administration and digital content providers to implement it; 2) discuss, review, and make decisions about policies and guidelines; 3) make recommendations to Library Administration about the resources needed to support digital collections (e.g., labor, equipment, software, etc.); 4) serve as advisers to other members of the library or campus community in the planning and creation of digital collections; and 5) make recommendations to Library Administration about which proposals it strongly endorsed, supported, or considered not ready.
In subsequent meetings, we began to discuss the policy framework, using the Cornell Action Plan for Developing a Digital Preservation Program [5] provided during the workshop. We managed to work our way through the first several sections of the Action Plan before our attention was diverted to other matters.
Since beginning our work in August 2004, we have [6]:
Developed a checklist of information needed for a digital content survey.
Developed a digital content survey questionnaire and a contextual document for it, laid out the costs and benefits, and tested the survey instrument on an existing digital collection.
Reviewed and made preliminary recommendations for file naming conventions.
Revised and approved the mass storage unit (MSU) guidelines and migrated existing projects into the MSU structure.
Developed a preservation policy statement for digital resources.
Drafted a mission statement that is now awaiting administrative approval.
Developed a standardized workflow for acquired digital content, outlining roles and responsibilities, and attempted to handle several newly licensed collections using the workflow.
Reviewed and began to document our backup procedures for digital content.
Began to explore the possibility of setting up offsite storage of a copy of our backups with a sister institution in the region.
Kept apprised of new developments in digital preservation.
In June 2005, I attended the fourth DELOS International Summer School on Digital Preservation in Digital Libraries in Sophia Antipolis, France. Anne Kenney, one of the trainers and creators of the Cornell workshop, spoke for half a day on the principles of digital preservation. It was extremely useful for me to be able to review Cornell’s model again and take stock of the progress we had made (or failed to make) at the University of Oregon in the eleven months that had passed since I attended the Cornell workshop. When I filled out the survey on our institutional readiness this time, it seemed to me that we had made some progress. On the Organizational Framework, I rated us at stage two, rather than stage one. There was even some evidence that we might be moving on to stage three. The key indicators of stage two in the Organizational Framework are development of general policies and planning with increased evidence of institutional commitment. At stage three, there are basic and essential policies in place. Our progress in this area has been the result of the work of the Libraries’ Digital Content Coordinators. Regarding the technological framework, I rated us between stages two and three, the improvement being due to the fact that we have made progress away from being purely project-specific and reactive to doing more assessment of our technology needs and becoming more proactive. This is due, again, to the efforts of the Digital Content Coordinators.
Resources Framework
In June 2005, when I rated the UO Libraries on the third leg of the digital preservation stool, the Resources Framework, I concluded that we were still firmly entrenched at stage one: generally low, finite, and ad-hoc financial commitment. This is true with regard to staffing, funding, and space resources. The Oregon model of “just do it” is about to do us in. Since the DCC began its work in August 2004, the digital content managed through CONTENTdm® has increased 1,452% and the digital content managed through DSpace has increased 1,119%. There is no record of how much digital content created and managed by other groups or individuals in the library using other software systems has increased in this same time period. There are, as yet, no formal business plans for any of these collections. Some small amounts of outside funding have been secured, enabling us to hire a few temporary staff for short periods to assist with some of this growth. However, the additional staffing does not approach what would be needed to plan for, create, and manage the substantial increase in digital content that we have experienced. Requests for new hardware and software to keep pace with the needs for new types of digital content to be delivered ever more quickly go into the same pool for consideration as the needs to upgrade staff and public workstations, to purchase new system packages for controlling licensed electronic resources, or for acquiring new servers for all of the library’s computing needs. There is never enough money to meet all of our needs and we frequently find ourselves robbing Peter to pay Paul. Our ability to make steady progress on the goals and objectives that we identified as being necessary to become a trusted digital repository is continually hampered by a resources framework that is insufficient for the task. We are also victims of our own success. Our colleagues’ and our users’ appetites for more digital content, with ever more sophisticated user interfaces, have outstripped our ability to create and manage the content well. In such an environment, we are hard pressed to plan adequately and to manage the content appropriately. It has become increasingly difficult just to follow the standards and policies that we have already developed.
Clearly, the next step must be to develop business plans and impose project planning on our digital collections work. Reflecting on our efforts and slight progress since attending the Cornell workshop, I am reminded of a slide that Anne Kenney and Nancy McGovern presented when they first addressed the five organizational stages of digital preservation. In discussing the need to acknowledge that digital preservation is a local concern, they showed a slide that said, “Hello, my name is Cornell and I’m a digital imaging junkie.” Paraphrasing the slide, the University of Oregon Libraries must acknowledge our need to devote adequate resources to the task at hand and say, “Hello, my name is UO and we can no longer ‘just do it’.”
In January 2004, the municipal archives of the Dutch harbour city of Rotterdam started an ambitious project to construct a solution for the long term preservation of its digital materials. Project planners asked the Netherlands Archives School for support, and a cooperative effort between the two organizations ensued. The Archives School brings both knowledge and training skills to the project, as well as research expertise. In the School’s strategy, teaching, research, and practical applications are closely connected.
The municipal archives of Rotterdam is not a small organization. It has a staff of over 100 and holdings of 17 linear kilometers of records, hundreds of thousands of photographs, maps, drawings, prints, and books. Like most Dutch archives, it endorses the Total Archives concept, aiming to document the local society and serving as a primary source of the town’s rich history, dating back to the Middle Ages. Today, Rotterdam is a modern city, with a population of around 700,000 and a port that boasts to be the world’s largest. Increasingly the town administration carries out its business electronically, and consequently, records are created in digital form. The archives anticipates ingesting electronic records in the near future; most of the photographs, television and sound recordings are currently delivered in digital format. Furthermore, like many cultural heritage institutions, the archives undertakes programs to digitize original analog materials in order to improve access and use of its holdings. All of these types of digital materials need to be preserved as valuable cultural assets.
Project Goals
The initial goal was to develop a low-cost digital repository (e-depot), based on state-of-the art theory, standards, and best practices. The underlying idea was that if it is impossible to build a digital repository for a reasonable cost then keeping digital materials would be an impossible task for any Dutch archive.
During the course of the project the objectives were broadened. The original goal remained the same, but equal importance was given to learning, improving methods, and changing work processes – that is, to building a competent organization that is able to deal with all kind of problems associated with long preservation and access to digital information.
First Year: Just Do It
For each of the two years that the project has been underway, project staff chose a different metaphor and associated approach. The first year theme was: “Just Do It,” learning by doing. None of the archives staff had any real experience with digital preservation, and the theoretical knowledge was limited. Since the very beginning, all departments have been involved in the project, including archives management, the map and print department, the library, preservation, and reference services. The heads of the departments and a project manager formed the project team, and senior staff control the project as a steering committee.
The approach was extremely pragmatic. Based on recommendations from HP Netherlands (the company sponsored the project by providing a server), DSpace was installed. This open source digital library software was already in use by six Dutch university libraries, including the library of Erasmus University in Rotterdam. The project team attended a DSpace demonstration at Erasmus University, and their digital library staff was helpful in answering questions. Also, prior to working with the software, the project team paid a visit to the UK National Archives, to see their system at work. This visit helped shape the discussion about the impact of the initiative on the organization as a whole.
Time Boxing
The project management method the team used during this first year was Time Boxing. They divided the work into two fixed periods, or Time Boxes. Each of the Time Boxes had a goal to achieve, but time was the dominant factor for project management. The first Time Box concerned installing, configuring, and becoming familiar with the software. Instead of the standard DSpace user interface, an open source harvester was installed: i-Tor, a Dutch product based on the OAI protocol. DSpace and i-Tor were installed without customization, except to add the municipal archives’ logo to give the project team the feeling of ownership. The first lesson learned was that the software was not too complicated, nothing to be afraid of.
One conclusion from the first Time Box activities was that the standard DSpace workflow was not suitable for an archive. DSpace expects authors to submit a publication for ingest into a digital library. Such a submission must subsequently be approved by the librarian. The workflow in an archive is different. It is the archivist who decides about capturing records – the ingest strategy is a pull, not a push, strategy. The workflow was seen as too bureaucratic for archives staff with sufficient level of responsibility. Fortunately, a few university libraries felt the same need to change the workflow and the costs of modification could be shared.
The second Time Box continued the testing of ingest and description of a variety of single documents. By the end of the year, the project team was competent to evaluate the DSpace and i-Tor functionality.
Migration
After examining the existing literature on long term preservation, the Steering Committee decided to adopt migration and standardization as major strategies. As much as possible, documents would be migrated to XML. The team acquired and tested an open source migration software utility, XENA, from the National Archives of Australia. The best result was that the team got hands-on experience and insight in the process. A second, less positive result was the finding that the software is still immature.
The project also borrowed much from the hardware architecture of the National Archives of Australia, with separate working environments for quarantine, storage, preservation and description, and access. The fully operational hardware infrastructure was installed in January 2005.
System Design
In this project, systems design follows development. This is a controversial approach, but it worked well in an experimental, prototyping environment. The systems design is heavily based on the Open Archives Information System Reference model, the InterPares Preservation task force model, and emerging metadata schemas. The overall design approach is based on the Yourdon method of structured systems analysis and design and consists of two series of documents (models) each describing a facet of the system. The first series form the Essential Model or Logical Design. Components of the Essential Model are the Environment Model (describing the interaction between the system and its environment), the Process Model (describing the systems processes), the Metadata Model, the Control Model (describing how the system will be managed), and the Knowledge Model; which is an extension to Yourdon, describing the rules and domain knowledge necessary for running the system. The second series is the Implementation Model or Technical Design. It describes the hardware and software environment, the processors, the software modules, and the human-machine interfaces.
In the first year, a draft of the Essential Model was completed. This model clearly showed that a digital depot was just one building block in the whole of the information architecture for the archives, reflecting the fact that senior management expressed the policy to integrate digital archives management into the normal business of the departments. The project team rejected the alternative strategy of establishing a separate digital archives.
Second Year: Island Hopping
The objective for the second year was to test the limits of DSpace. It was expected that by the end of the year a decision could be made as to whether or not DSpace had sufficient functionality to serve as the basis for the digital depot system.
The project team developed a different project management strategy, using the metaphor of Island Hopping. Several research areas were selected, such as bulk data, multimedia, websites, databases, relationships with records creators, hardware and software, policy, access (digital reading room), metadata, procedures, and systems design. A working group was established for each area. User participation in the project this year was increased from 12 to over 30. The team invited other archives to have staff participate in the working groups. Each group had the task to accomplish a predefined goal: find an answer to a research question; that is, “to conquer an island.” Some working groups were more successful than others and all had different approaches. For example, some groups tested software, while other groups analysed theories and best practices.
Results
The overall positive result was – I should say “is” – the enforcement of the learning capacity of the organization. Staff members realized that implementing digital archiving is not just a complex and multi-disciplinary task, but it changes the organization and is more of a continuous process than a project. Digital archiving requires life long learning from both individuals and organizations. It requires and enables cooperation. Departments that worked for decades within their own disciplinary lines discovered each other's methods and systems and were open to challenging their own practices and standards. The working group on access provides one example. The group developed functional and quality requirements for a digital (“virtual”) reading room. In the process, the group evaluated existing applications and Web pages and concluded that a redesign would be needed. This appeared to be a serious problem because the different departments not only used different software applications, but variations in metadata models hampered easily understood access to the collections. The group set out to develop a harmonized metadata model for descriptions, applicable to all kinds of materials held in the archives, including records, photographs, prints, drawings, books, magazines, sound recordings, moving images, etc. They started with the ISAD(G) and ISAAR(CPF) standards, endorsing the concept of authority control, a rather unknown concept at that time. The group was successful, and the logical metadata model could even be mapped onto EAD, EAC, and Dublin Core, opening the way for the development of an organization-wide XML meta-database of descriptions, based on open standards, and accessible to harvesters, browsers, and search engines. The implementation of the database should be complete by the end of 2006.
Working with one metadata model affects the applications that the archives is using for description and access. Most of these applications are specific to one type of material and use different metadata schemas. Discontinuing these is not an option, at least not in the short term. Output scripts will likely be needed to migrate the local databases into the generic XML meta-database.
DSpace and Other Software
During the pilot projects carried out by the working groups, the shortcomings of DSpace became apparent. First of all, the metadata schema is insufficient for archival purposes. DSpace supports qualified Dublin Core, which might be good enough for presentation and retrieval of Web publications, but does not meet archival requirements. Full, multilevel description is not possible, nor is authority control. However, DSpace is now available on an Oracle platform, which makes it technically possible for the Rotterdam archive to merge with existing metadata systems. This means intense customization of the software. A second problem with DSpace is the ingest process. DSpace uses an upload procedure that appears to be too slow for large files, such as multimedia data and multi-layered bulk. One of the project’s most important questions now is whether or not to continue with DSpace. At the moment, alternative open source software is under consideration. The project team works on a component based architecture for the archival institution as a whole. Much of the required functionality for managing digital archives is not different from managing analog materials, and much of it is already available in current applications. The repository function is just one part of it, and possibly, software is available.
How to Proceed?
Whatever decision about DSpace is taken, the work will go on. Next steps include a rigorous system design. Based on design decisions, software options will be evaluated. The archive will need to decide the extent to which to continue with open source software. The original choice for DSpace and other open source software was a pragmatic one. DSpace was available and proved to be a good tool for acquiring knowledge and experience. A fundamental discussion about open source software has not, as of yet, been undertaken.
Working with open source is not necessarily less costly than using commercial software. It requires working with communities, and there are problems with an archival institution entering into a community mainly consisting of libraries. For open source to work, it is absolutely necessary to establish an archival community to cooperate and to share experiences and costs.
The Electronic Literature Organization (ELO) is dedicated to new forms of literary projects now possible using existing and emerging computer-mediated technologies. Specifically, ELO considers electronic literature to be “…works with important literary aspects that take advantage of the capabilities and contexts provided by the stand-alone or networked computer.”
The organization sponsors several programs and events to promote writers and publishers of electronic literature and to develop a supportive infrastructure for their works, including a directory of electronic literature, educational events, and awards. An additional ongoing project is the Preservation, Archiving, and Dissemination (PAD) project. This component of ELO “…seeks to identify threatened and endangered electronic literature and to maintain accessibility, encourage stability, and ensure availability of electronic works for readers, institutions, and scholars.” This project has delivered two reports:
Acid-Free Bits: Recommendations for Long-Lasting Electronic Literature (June 2004)
Born-Again Bits: A Framework for Migrating Electronic Literature (August 2005)
Together these reports provide an interesting perspective from the community of authors and content creators that raises awareness and seeks to address the inevitable problems—inherent to many types of complex digital files—that threaten ongoing access to their own creative works.
Does the recent spate of natural and human disasters around the world carry any lessons for the stewards of electronic information in cultural heritage institutions?
Introduction
Shortly after the U.S. invasion of Iraq in March 2003, widespread ransacking of that country's museums and libraries—and the corresponding loss of cultural artifacts, monuments, and antiquities—was widely reported in the mainstream media. That's unusual. More typically, and quite understandably, discussion of loss associated with natural disasters, warfare, riots, and terrorist incidents focuses on loss of life and private property, environmental damage, homelessness, job loss, and economic dislocation. Loss of cultural heritage treasures, though commonplace under most disaster scenarios, rarely achieves more than anecdotal mention.
Nevertheless, once more-immediate needs are attended to, loss of cultural heritage adds to the suffering of survivors and to that of the human community at-large. As stressed by the International Committee of the Blue Shield (a network of international organizations working to protect historic sites, museums, archives and libraries in the event of natural and human disaster) in its statement on the impact of Hurricane Katrina,
the cultural heritage of a community reflects its way of life, its history and identity, provides the link between its past, present and future, and contributes substantially to its economic sustainability and welfare. A community cannot function effectively in the larger societal context if deprived of its identity through the loss of its cultural heritage.
In the aftermath of a major disaster, accounting for cultural heritage losses will inevitably take some time, but once accomplished, it can add a sobering tally to the already massive sense of devastation. A fine example of such a reckoning is Cataclysm and Challenge by Heritage Preservation, which discusses the impact of the September 11, 2001 attacks on cultural heritage in and around the World Trade Center and the Pentagon. Their report , published in 2002, makes for compelling but harrowing reading as it details the disintegration of numerous cultural treasures in the collapse of the twin towers and in five adjacent buildings that were subsequently destroyed by the resulting debris and fires.
Just a few excerpts:
A total of 60 nonprofit organizations, along with their records and archives, had offices in the Trade Center. Primarily family foundations and charitable trusts, their interests ranged from mountaineering to early childhood development to support for the Brooklyn Botanic Garden. (p.7)
In addition, 22 federal government departments and agencies had offices in the Trade Center complex. Included were the Secret Service, the Department of Transportation (Coast Guard), the Department of Defense, the Peace Corps, the Department of Labor, the Federal Maritime Commission, the Treasury Department, and the Department of Housing and Urban Development. All official records and office documents, together with computer hard drives, were destroyed. Although certain revenue collection agencies, such as the Internal Revenue Service and the Customs Service, are required to have back-up documentation stored off-site, the National Archives and Records Administration presumes vast numbers of other federal records were simply lost. (p.7)
It took only minutes for the inferno to consume the archives and records of the Helen Keller International Foundation. An estimated $4 million in equipment, records, and historical data was lost, including first editions of Keller’s works, priceless photographs, and many of her own letters. (p.8)
Also destroyed were important paintings, sculptures, monuments, and building facades. The report also details losses in the Pentagon's library and art collection.
Original works of art, historic buildings and monuments, and tangible artifacts such as letters, manuscripts, rare books, and museum specimens will always be vulnerable to permanent loss in the event of a catastrophe. By definition, a unique cultural artifact is irreplaceable if destroyed. However, a growing trend in cultural heritage institutions (libraries and archives in particular) is for valuable holdings to be in digital form, either as digitized versions of analog collections, or as born digital content. Furthermore, descriptive catalogs and related documentation are more and more likely to exist only in digital form.
Digital holdings are generally perceived as both a curse and a blessing—a curse for their fragility, vulnerability, and ephemeral quality and a blessing for their compactness, ease of transport, and perfect replicability. Most cultural stewards are well aware that the strengths of digital content can and must be used to compensate for its weaknesses. Usually this means making backup copies of all but the least valued digital content. However, natural and human disasters of the past few years have made abundantly clear that simply creating backups is not sufficient. The terrorist attacks of September 11, 2001, the Asian tsunami of December 2004, Hurricane Katrina (August 2005), and the earthquakes in Iran (December 2003) and southeast Asia (October 2005) all resulted in great devastation. Though they varied in the number of lives lost, the degree of property damage, the size of the area of major destruction, and the period of time that essential services were disrupted, all carried a similar lesson about the survivability of digital information: unless backups are stored off-site and at a sufficient distance from the primary facility, they may succumb to the same disaster.
Who Needs Off-site Storage?
Decisions surrounding the use of off-site storage, like most concerning digital preservation policies, should be driven by risk management procedures. Risk management actions are governed by risk analysis, which can run the gamut from a seat of the pants evaluation to a highly formalized process involving the use of paid consultants. At minimum, even the poorest non-profit cultural heritage institution can analyze basic risks from potential disasters. Is the facility located in a known earthquake, hurricane, flood, or tornado zone? Is it built to withstand damage from known dangers? Is it near an airport or chemical plant?
Even if there are no obvious known risks, all facilities face at least some possibility of catastrophic destruction. No building, regardless of its location or material and means of construction is totally free from hazards, such as fire and flood (from nature or infrastructure failure). Digital assets can be lost or damaged by vandalism, sabotage, and equipment failure. Therefore, off-site storage of valuable digital assets is a risk management strategy that all cultural heritage institutions should seriously consider.
A growing number of vital records are covered by statutory record retention mandates from national, regional, or municipal governing authorities. In the US, agencies such as SEC (Securities and Exchange Commission), EPA (Environmental Protection Agency), IRS (Internal Revenue Service), and OSHA (Occupational Safety and Health Administration), among many others, have record retention requirements. Retained records must be protected from destruction, and this usually means off-site storage.
Federal legislation in the US such as HIPAA (Health Insurance Portability and Accountability Act) and Sarbanes-Oxley (regarding fiduciary conduct and financial disclosure by public companies) has tightened records management requirements for certain types of organizations. HIPAAdvisory, a commercially-sponsored HIPAA site, provides a partial summary of US Federal Government record retention laws. Some of the tightening grew out of concerns resulting from the September 11 attacks and from major accounting scandals such as Enron and WorldCom. Most of these regulations do not apply directly to digital data in the collections of cultural heritage institutions, but if an institution is already legally required to invest in off-site storage for vital records, it makes sense to investigate whether digital collections can be included in the same program.
Some excellent resources are available to help organizations conduct their own risk analysis and risk management, including issues relevant to off-site storage. The Computer Security Resource Center of NIST (the US National Institute of Standards and Technology) produces a series of special publications to assist organizations with a wide range of computer security issues. Three key guides from NIST dealing with risk management for information technology are listed in the resources at the end of this piece. Additionally, HHS, the US Dept. of Health and Human Services, publishes a guide on the basics of risk analysis and risk management (see resources). Though geared to "electronic protected health information" as defined in the HIPAA regulations, it provides a simplified overview of risk analysis and risk management procedures that can be applied to any kind of electronic information.
Devising an off-site storage plan is only one part of disaster recovery planning. Although this piece will maintain a narrow focus on off-site storage issues, the resources listed below can be used for many other aspects of disaster recovery planning related to information technology.
Off-Site Storage Considerations
The September 11 attacks significantly increased interest in and awareness of off-site storage and facilities issues. This was due in part to the media attention paid to the devastating losses suffered by bond trading firm Cantor Fitzgerald in the World Trade Center attacks and the amazing story of its recovery, largely attributable to excellent off-site preparations.
Hurricane Katrina has had a similar effect, but for different reasons. September 11 convinced organizations to think about the prospect of sudden and complete destruction of their physical infrastructure. Katrina drove home the fact that, even with ample warning of possible devastation, the geographic scope of a natural disaster can be so enormous as to invalidate even careful advance disaster recovery planning.
Businesses, especially those that rely heavily on information technology, have become particularly aware of their vulnerability to heavy financial losses or complete failure, as a result of natural and human disasters. The industry that provides disaster recovery services (also called business continuity services) has grown tremendously since September 11, and each new major disaster convinces more organizations to address the issue seriously.
Though mostly non-profit, cultural heritage institutions can also face financial ruin from disasters, but may not be able to afford the most sophisticated and failsafe off-site storage options. Fortunately, there are numerous alternatives available, varying in price and effectiveness.
The first place to start is with prevention. Natural disasters can't be prevented, but the impacts of minor ones often can be mitigated, and many human ones can be precluded by taking appropriate precautions. The NIST Contingency Planning Guide for Information Technology Systems (see resources) has a useful list of preventive controls (pp.18-19). Be careful in the selection of heat and fire resistant cabinets for protection of backup media. Those designed for paper documents are inadequate for magnetic media, which become unusable at much lower temperatures than paper burns.
Cultural heritage entities within large institutions may have the option of storing backup media in another building on their campus. This will probably be relatively inexpensive, and provides fast access when the backups are needed, but may not provide sufficient distance to prevent common destruction or loss of supporting infrastructure (e.g., electrical power) in a widespread disaster.
Another popular but relatively inexpensive option is to maintain a reciprocal arrangement with a like-minded institution a reasonable distance away. (What's a reasonable distance? We'll address that in the next section.) Obviously the other entity must be one that will take the responsibility seriously and provide a level of security, environmental controls, and handling care comparable to one's own expectations. In the event of a disaster affecting one party, the other has to be in the position to handle its own IT operation and to service the other's needs, simultaneously. The responsibility should not be taken lightly.
For institutions creating backups on conventional tape and optical media, there are a variety of commercial vendors that will store your media off-site. Look for a data management firm, not a mere warehouse or storage locker. The facility should be a reasonable distance from the primary site, temperature and humidity controlled, secure, and staffed by people who know how to handle electronic media. The off-site facility's own vulnerability to disasters should also be assessed. Is it in an industrial part of town next to an oil refinery? That may negate any other perceived advantage.
Suggestions for criteria to use in selecting an off-site storage vendor can be found in the Abbey Newsletter, the NFPA (National Fire Protection Association) Journal, and the Quality Cities Magazine. The state of Delaware has made its request for proposal (RFP) for off-site data storage publicly available. The specifications (pp. 23-25) within the RFP provide an excellent checklist for anyone evaluating off-site data storage contractors. The NIST Contingency Planning Guide for Information Technology Systems also includes selection criteria.
An option that is growing in popularity is the use of online off-site storage. Instead of backing up to removable media and transporting it to an off-site location, backups are conducted across a network (using either dedicated lines or the internet) to a secure, remote facility where the backups are maintained on spinning magnetic disks. For a premium, it is possible to subscribe to a real-time backup service over dedicated fiber optic lines that mirrors all your most precious data and can restore every file and every transaction that was saved, up to the moment a disaster occurs.
More affordable online backup services perform once-daily backups using existing network bandwidth. Typical amenities include automatic data compression and encryption, 24/7 data availability, and retention of multiple historical versions of each file. These services are still relatively expensive, though most claim to be competitive with local tape backups stored in an off-site facility because they eliminate the costs of personnel to perform the backups, hardware and media to create the backups, and the cost of the off-site facility. Most also claim to be considerably more reliable than tape backups. A review of several of the better-known online backup services provides a good sense of their capability and cost.
Other models for online off-site backup exist, including some employing a cooperative sharing architecture. An example of a formal, large-scale network of multi-site, reciprocal storage arrangements in the academic library community is the LOCKSS Alliance.
Which off-site data storage approach is right for a particular institution? Chances are most will choose a combination of the options. Digital assets vary in the level of concern, care, and expense their protection merits. Digital assets that required considerable time and effort to create deserve a higher degree of protection than those that would be easier to replace or recreate. Collection inventories, catalogs, and transactions such as library circulation records are mission critical and irreplaceable. According to the Heritage Preservation report mentioned earlier, the full extent of the cultural heritage artifacts lost in the World Trade Center attacks will never be known because many institutions whose collections were destroyed failed to store copies of their catalogs or inventories off-site. Off-site storage of routine application software, operating systems and utility programs can ease the process of returning to operation following a disaster, but may be less critical than data files and other unique assets.
Ultimately, any solution chosen must be practical and affordable. No matter what compromises are involved, even a partial but well-thought out off-site storage plan will be better than none.
How Far is Far Enough?
Following September 11, a draft interagency white paper produced by the SEC, the Federal Reserve, the Office of the Comptroller of the Currency, and the State of New York Banking Department proposed guidelines to "assure the resilience of critical US financial markets in the face of wide-scale, regional disruptions." Among its proposals was that certain crucial entities (major banks and securities firms) be required to establish off-site backup facilities located a minimum of 200-300 miles from the primary site from which they could restore their operations in less than a day with no loss of data.
Even in the cautious environment of that time, the proposed requirement was widely seen as excessively burdensome and technologically unachievable. However, trying to get agreement on what is the "correct" distance to place an off-site backup facility from its primary site isn't easy. Greater distance decreases the likelihood of simultaneous loss of both facilities, but increases transportation costs and the time required to retrieve off-site backups.
The widespread devastation caused by Hurricane Katrina and the subsequent levee failures around New Orleans has lead some to reconsider how far is far enough. Many firms in New Orleans that were flooded could not retrieve their backups from off-site storage facilities across town because the flooding simultaneously made those facilities inaccessible.
Disasters vary considerably in the size of the area affected. Hurricanes, for example, will typically cause devastation over a larger area than tornadoes, so desirable distances between primary and off-site facilities should theoretically be greater in a hurricane-prone area than in a tornado belt. Nevertheless, the question remains, how far is far enough?
In order to gain more insight into this issue, in 2002 PreEmpt, Inc., a Texas-based business continuity firm, conducted a survey of members of the Association of Contingency Planners (ACP) asking "How far should an alternate or offsite storage facility be from the primary operations facility?" ACP is the trade association connected with the business continuity industry. For off-site storage facilities, the survey results indicated a range of an average of 24 miles (for the risk associated with proximity to a civilian airport) to 84 miles (for hurricane-prone regions), depending on the nature of the risk or threat.
PreEmpt repeated the survey in 2005 during the period when Hurricanes Katrina and Rita were active. For most of the threats assessed, the average distances deemed as safe by the ACP members increased by 50-100%, compared to the 2002 survey. Apparently, business continuity professionals perceive an increase in danger during the past three years. The surveys make interesting reading and include other questions related to off-site storage decision making.
The bottom line on the distance question is that there is no hard and fast rule, but a substantial majority of disaster recovery professionals feel that it is never acceptable for backups to be stored in the same building, or even the same campus as the original copy. For real safety, conventional wisdom is that even relatively low-probability, geographically constrained threats merit a separation of tens of miles between primary and off-site locations.
Conclusion
In a tumultuous world that is more and more heavily reliant on digital data for its day-to-day operations, business and government are embracing as a basic tenet of computer security the need to locate backups of digital assets in locations considerably removed from the primary site. Non-profit and cultural heritage institutions that have not adopted off-site storage strategies should seriously consider whether they can or should do so. Many different options are available, and not all are high-cost.
Many cultural heritage organizations need to shore up their overall disaster plans. Developing an off-site storage strategy for backups of critical digital assets would serve as one important step in the right direction. With the lead seasonal hurricane forecaster at NOAA's Climate Prediction Center in Camp Springs, Md. recently predicting ".. high levels of hurricanes—and hurricane landfalls—for the next decade or even longer," there is no time like the present for improving disaster preparedness.
Looking at the bigger picture, the off-site storage solutions discussed here offer protection only against immediate loss of data from a disaster at the primary site. They do not address long-term preservation functions such as file format and media migration, authenticity testing, or metadata creation. Much work remains to be done before comprehensive digital preservation plans and programs are common.
US Dept. of Health and Human Services, Centers for Medicaid and Medicare Services, "Basics of Risk Analysis and Risk Management," Paper 6, Security Series, June 2005. (See especially starting on page 3 with "Important Definitions to Understand" through to page 16.)
The Digital Curation Centre and the Wellcome Library will examine curatorial processes, formats, and practical tools and techniques to help guard institutions’ websites against preservation related risks, such as institutional change and technological obsolescence.
Arizona State University will host this interdisciplinary meeting focusing on technical, policy, and training issues important for managing the electronic records of higher education institutions.
This symposium, sponsored by the University of Michigan University Library and National Commission on Libraries and Information Science, will facilitate discussion about the implications of mass digitization projects for libraries, universities, government, information policy, publishing, and education.
The mission of this conference will be to create collaborative, cross-departmental, cross-community approaches to look at the impact that the digital environment has on library collections, access to resources, and organizations. This conference will also provide an online component (part of the standard registration or purchased separately for those who are unable to attend in person) that will provide web-based access to speakers’ presentations, discussion areas, blogs from round tables, and selected audio presentations.
This meeting is intends to bring together librarians, archivists, records managers, and other information professionals to share experiences—successes, failures, lessons learned—in their efforts to manage and preserve digital state government information.
This event includes a conference, workshops, and training sessions centered around electronic imaging and the visual arts. Target audiences include professionals from cultural heritage, IT, media, government, and media research sectors.
The Second IEEE International Conference on Document Image Analysis for Libraries (DIAL’06) will cover technical aspects of document image processing, including digitization, image restoration, text recognition, metadata extraction, document encoding, and file formats. This workshop will also attempt to “describe the state of the art, to identify urgent open problems in image analysis suited to [digital libraries] and to collect all available information on document digitization projects across the world.”
Visit the new DigiWik as a contributor (a wiki is a website that anyone can edit) or for consultation. DigiWik is designed to be a repository of digitization information for use by individuals, museums, libraries, researchers, and any other entities with digitization needs.
Internet Archive’s Archive-it The Internet Archive has launched a new service, Archive-it, a public website that allows subscribers to create, manage, and search their own web archives. The service has recently undergone beta testing by several library and archive organizations. Some of the collections created by the pilot participants are currently available for browsing, such as The Library of Virginia's “Virginia's Political Landscape, Fall 2005” and the Latin American Government Documents Archive (LAGDA at the University of Texas).
The Informing Science Institute has announced the Journalof Information, Information Technology, and Organizations, a peer reviewed journal, available online, free of charge. The journal aims to publish papers that address a balanced treatment of all three entities signified in its title: information, information technology (IT), and the organizational context.
The National Library of Australia (NLA) has completed their first harvest of the entire Australian Web domain covering the period of June-July 2005. The NLA contracted with the Internet Archive to crawl approximately 185 million documents from 811,000 host sites—a total of 6.69 terabytes. The project is documented in an internal report by Paul Koerbin, which is available online.
The National Council on Archives (NCA) (UK) has published a new document intended to be a resource for archivists’ for raising awareness about “why preserving electronic records is an urgent issue.” Written with the layperson in mind, this advocacy document targets an audience of non-archivists in positions of authority who may make policy or resource decisions that affect archival services.
Microsoft has announced that it will submit the Office Open XML file format technology to Ecma International, a European standards body. Office Open XML will be the default file format for the next version of Office.
The Council on Library and Information Resources and Digital Library Federation (CLIR) has released a new report, “Acquiring Copyright Permission to Digitize and Provide Open Access to Books.” The report details three cases in which Carnegie-Mellon University sought to secure copyright permission to digitize and provide open access to books with scholarly content. The experiences are discussed in light of U.S. copyright laws, licensing practices, and technological developments in the publishing industry.
The LOCKSS Alliance announced that it has reached 50 academic library members. Also, the Joint Information Systems Committee (JISC) has recently announced that it will deploy a UK LOCKSS Pilot program.
The Google blog has a recent post about the Google Print project, their effort to make public domain books available and searchable online. Scroll down to see a long list of trackbacks for more discussion on the project.
Digitisation in the UK: The Case for a UK Framework This recently released report is based on the findings of a survey of digitized materials in the UK, conducted by Department of Information Science at Loughborough University and commissioned on behalf of Joint Information Systems Committee (JISC) and Consortium of Research Libraries in the British Isles (CURL). The report recommends the creation of a UK-wide strategy in order “to avoid the duplication, gaps in provision and lack of coordination that have hampered public sector efforts in this area.”
The National Archives of the UK has announced the release of PRONOM 4 and a new software tool, DROID (Digital Record Object Identification), an automatic file format identification tool that works with PRONOM to identify specific file format versions of digital files.
At the ceremony held at the British Library, award judges stated they "were impressed by the work PREMIS has done in compiling a "data dictionary" identifying core digital preservation metadata, which they have supported with practical examples and a software protocol. A key factor in the decision was the international scope of PREMIS, and the consensus building and collaboration that is so crucial in so many digital preservation issues." Further, the awards ceremony singled out the work of PREMIS by stating it "goes a long way towards establishing an international open-source standard for handling meta-data, which will help libraries and institutions around the world to archive digital content - the volume of which is doubling every year." More information about the UK Digital Preservation Award and the awards ceremony can be found here.
The PREMIS data dictionary and schema work remain active through the PREMIS Maintenance Activity, based within the Library of Congress's metadata standards website, http://www.loc.gov/standards/premis/. Additional information about PREMIS implementation, including the opportunity to join the PREMIS Implementors' Group, can be found on the maintenance activity site. As additional tools are developed, they will also be available there.
RLG DigiNews (ISSN 1093-5371) is a Web-based newsletter conceived by the RLG preservation community and developed to serve a broad readership around the world. It is produced by staff in the Department of Research, Cornell University Library, in consultation with RLG and is published six times a year at www.rlg.org.
Materials in RLG DigiNews are subject to copyright and other proprietary rights. Permission is hereby given to use material found here for research purposes or private study. When citing RLG DigiNews, include the article title and author referenced plus "RLG DigiNews." Any uses other than for research or private study require written permission from RLG and/or the author of the article. To receive this, and prior to using RLG DigiNews contents in any presentations or materials you share with others, please contact Jennifer Hartzell (jlh@notes.rlg.org), RLG Corporate Communications.
Please send comments and questions about this or other issues to the RLG DigiNews editors.
Co-Editors: Anne R. Kenney and Nancy Y. McGovern; Associate Editor: Robin Dale (RLG); FAQ Editor: Richard Entlich; Contributor &Copy Editor: Ellie Buckley; Production: Jenn Colt-Demaree, Carla DeMello; Advisor: Peter Hirtle.
All links in this issue were confirmed accurate as of December 15, 2005.
Printable Version
In late 2002, the MIG began to meet and hammer out agreements regarding descriptive, technical, and administrative metadata for a collection of photographs that had previously been digitized. We all became conversant in Dublin Core, we reviewed standards and practices being developed and utilized worldwide, and we tested software. We planned to migrate the digital content into the CONTENTdm® system, rescan some content to ensure it met our standards for digital images, and apply the newly approved metadata standards to this collection. 
We created and made this collection publicly available in record time (four months) to fulfill our obligations to a granting agency. We achieved this by calling on the expertise of four librarians in what was then the Catalog Department, working closely with staff in Special Collections & University Archives (especially the Image Services Center), and recruiting staff of the Catalog Department to learn how to scan glass plate negatives, apply descriptive and technical metadata, and load the digital objects and accompanying metadata into the CONTENTdm® system. While this was underway, the library also established an institutional repository, called 
Recognizing that we could have little short-term impact on the resources devoted to preserving digital collections, we decided to focus our immediate attention on improving the organizational and technological frameworks as much as possible. Immediately after returning from the workshop, I arranged to meet with our Library Administration to outline the challenges and propose some strategies for addressing them. I prepared a two-page document for them: laying out a statement of need, defining digital preservation, listing the attributes of a trusted digital repository, and proposing a series of next steps
In June 2005, when I rated the UO Libraries on the third leg of the digital preservation stool, the Resources Framework, I concluded that we were still firmly entrenched at stage one: generally low, finite, and ad-hoc financial commitment. This is true with regard to staffing, funding, and space resources. The Oregon model of “just do it” is about to do us in. Since the DCC began its work in August 2004, the digital content managed through CONTENTdm® has increased 1,452% and the digital content managed through DSpace has increased 1,119%. There is no record of how much digital content created and managed by other groups or individuals in the library using other software systems has increased in this same time period. There are, as yet, no formal business plans for any of these collections. Some small amounts of outside funding have been secured, enabling us to hire a few temporary staff for short periods to assist with some of this growth. However, the additional staffing does not approach what would be needed to plan for, create, and manage the substantial increase in digital content that we have experienced. Requests for new hardware and software to keep pace with the needs for new types of digital content to be delivered ever more quickly go into the same pool for consideration as the needs to upgrade staff and public workstations, to purchase new system packages for controlling licensed electronic resources, or for acquiring new servers for all of the library’s computing needs. There is never enough money to meet all of our needs and we frequently find ourselves robbing Peter to pay Paul. Our ability to make steady progress on the goals and objectives that we identified as being necessary to become a trusted digital repository is continually hampered by a resources framework that is insufficient for the task. We are also victims of our own success. Our colleagues’ and our users’ appetites for more digital content, with ever more sophisticated user interfaces, have outstripped our ability to create and manage the content well. In such an environment, we are hard pressed to plan adequately and to manage the content appropriately. It has become increasingly difficult just to follow the standards and policies that we have already developed. 
Clearly, the next step must be to develop business plans and impose project planning on our digital collections work. Reflecting on our efforts and slight progress since attending the Cornell workshop, I am reminded of a slide that Anne Kenney and Nancy McGovern presented when they first addressed the five organizational stages of digital preservation. In discussing the need to acknowledge that digital preservation is a local concern, they showed a slide that said, “Hello, my name is Cornell and I’m a digital imaging junkie.” Paraphrasing the slide, the University of Oregon Libraries must acknowledge our need to devote adequate resources to the task at hand and say, “Hello, my name is UO and we can no longer ‘just do it’.”
Project Goals 
One conclusion from the first Time Box activities was that the standard DSpace workflow was not suitable for an archive. DSpace expects authors to submit a publication for ingest into a digital library. Such a submission must subsequently be approved by the librarian. The workflow in an archive is different. It is the archivist who decides about capturing records – the ingest strategy is a pull, not a push, strategy. The workflow was seen as too bureaucratic for archives staff with sufficient level of responsibility. Fortunately, a few university libraries felt the same need to change the workflow and the costs of modification could be shared. 
System Design
The objective for the second year was to test the limits of DSpace. It was expected that by the end of the year a decision could be made as to whether or not DSpace had sufficient functionality to serve as the basis for the digital depot system. 
Results
How to Proceed?
http://www.eliterature.org/
Digital holdings are generally perceived as both a curse and a blessing—a curse for their fragility, vulnerability, and ephemeral quality and a blessing for their compactness, ease of transport, and perfect replicability. Most cultural stewards are well aware that the strengths of digital content can and must be used to compensate for its weaknesses. Usually this means making backup copies of all but the least valued digital content. However, natural and human disasters of the past few years have made abundantly clear that simply creating backups is not sufficient. The terrorist attacks of September 11, 2001, the Asian tsunami of December 2004, Hurricane Katrina (August 2005), and the earthquakes in Iran (December 2003) and southeast Asia (October 2005) all resulted in great devastation. Though they varied in the number of lives lost, the degree of property damage, the size of the area of major destruction, and the period of time that essential services were disrupted, all carried a similar lesson about the survivability of digital information: unless backups are stored off-site and at a sufficient distance from the primary facility, they may succumb to the same disaster.
Even if there are no obvious known risks, all facilities face at least some possibility of catastrophic destruction. No building, regardless of its location or material and means of construction is totally free from hazards, such as fire and flood (from nature or infrastructure failure). Digital assets can be lost or damaged by vandalism, sabotage, and equipment failure. Therefore, off-site storage of valuable digital assets is a risk management strategy that all cultural heritage institutions should seriously consider.
The September 11 attacks significantly increased interest in and awareness of off-site storage and facilities issues. This was due in part to the media attention paid to the devastating losses suffered by bond trading firm Cantor Fitzgerald in the World Trade Center attacks and the 
For institutions creating backups on conventional tape and optical media, there are a variety of commercial vendors that will store your media off-site. Look for a data management firm, not a mere warehouse or storage locker. The facility should be a reasonable distance from the primary site, temperature and humidity controlled, secure, and staffed by people who know how to handle electronic media. The off-site facility's own vulnerability to disasters should also be assessed. Is it in an industrial part of town next to an oil refinery? That may negate any other perceived advantage.
Even in the cautious environment of that time, the proposed requirement was widely seen as excessively burdensome and technologically unachievable. However, trying to get agreement on what is the "correct" distance to place an off-site backup facility from its primary site isn't easy. Greater distance decreases the likelihood of simultaneous loss of both facilities, but increases transportation costs and the time required to retrieve off-site backups.
The bottom line on the distance question is that there is no hard and fast rule, but a substantial majority of disaster recovery professionals feel that it is never acceptable for backups to be stored in the same building, or even the same campus as the original copy. For real safety, conventional wisdom is that even relatively low-probability, geographically constrained threats merit a separation of tens of miles between primary and off-site locations.