Researching Long Term Digital Preservation Approaches in the Dutch Digital Preservation Testbed (Testbed Digitale Bewaring)

Maureen Potter
Digital Preservation Testbed, Netherlands
Maureen.Potter@ictu.nl


In 1996, the Netherlands Ministry of the Interior and the Ministry of Education, Culture and Sciences initiated a collaborative programme entitled Digital Longevity (Digitale Duurzaamheid). This programme, run in conjunction with the National Archives, sponsored Jeff Rothenberg's 1999 publication, Carrying Authentic, Understandable and Usable Records Through Time, which proposed establishing a testbed to carry out research into possible approaches for the long term digital preservation of archival records (1). The Digital Preservation Testbed (Testbed Digitale Bewaring) was born the following year.

This article introduces the work of the Digital Preservation Testbed. It first places the Testbed in context within the rest of the Digital Longevity programme and defines the scope and goals of the project. Our objectives and research questions are identified, followed by a review of the rigorous scientific approach that the Testbed takes in its experiments. The benefits of this are highlighted, as is the practical nature of the Testbed. Finally, the products and deliverables that are expected to emerge throughout the course of the project are discussed and identified.

Background and Scope

The Digital Preservation Testbed is part of a wider network of initiatives that the Dutch government has established to deal with the challenges posed by the electronic era. The Testbed belongs to the Digitale Duurzaamheid Programme, whose overall aim is to guarantee the accessibility of information held by the government in digital form (2). Three other projects complete the Digitale Duurzaamheid programme: the RecordKeeping System (RKS) project, establishing guidelines and providing advice to Dutch Ministries on the selection of an RKS; the Kwaliteitzorg, concerned with ensuring the quality of the records being produced electronically; and the Taskforce DigitaleDuurzaamheid, whose main aim is to raise awareness of the digital longevity issues throughout government. The goal of the Testbed within the Digitale Duurzaamheid programme is to help achieve the lasting accessibility of government information in digital form. The Testbed will provide advice that is tailored to the situation here in the Netherlands. Our focus is on the preservation of electronic records for the long term, and our strategy begins with preparing for the preservation of records from their point of creation. Our intention is to ensure the reliable creation and management of electronic records so that they are in a suitable state for long-term preservation action. The Testbed is running controlled experiments to explore options for long-term preservation approaches and the advice on these will be issued to the Dutch government later this year.

Our research is initially limited to four main alphanumeric record types: text documents, email messages, spreadsheets and databases, all of which are widely used within ministries and government organisations. Three preservation approaches are under consideration: migration, emulation, and XML, which are discussed in more detail below. Four record types and three approaches result in 12 possible combinations. This initial set concentrates our resources and limits what is otherwise an exponential and unstructured research area. Also, not every record type is suitable for every preservation approach. For example, we do not consider it to be worthwhile to attempt emulation for emails. Email packages rely upon standard exchange formats that enable email systems to be interoperable. The sender and receiver will often perceive the look and feel attributes of a message differently. The question then becomes: "what exactly am I trying to preserve?" You need not preserve something that was not present in the first instance. Emulation is thus not the best match for a preservation approach to this record type.

An integral assumption in our research is that different record types have different preservation and authenticity requirements. Records ingested into the Testbed are analysed in terms of the five attributes posed by Rothenberg for digital records: Context, Content, Structure, Appearance, and Behaviour. Authenticity Requirements are developed for each record type in terms of these five attributes and act as the success criteria for an experiment's preservation approach. In addition to this, other preservation and archival issues arise from these experiments. Our research thus extends to wider considerations and includes objectives beyond the success of a technical strategy.

Objectives

The objectives of the project are to provide insights into:

• Technical solutions for the preservation of authentic electronic records
• The effectiveness of current and potential preservation approaches
• Authenticity features of digital records
• Cost factors for storing, preserving and managing digital records and associated metadata
• Management processes and activities required to capture, generate and maintain metadata that support the ingestion of records and preservation of long-term access to authentic electronic records

Research Questions

The Testbed Research Framework translates these objectives into a clear set of research questions that are refined and updated throughout the duration of the project. These range from fundamental research questions that require comparing the results of large groups of experiments, to questions focusing on the role and significance of record features, attributes, and metadata that may be answered by individual and smaller groups of experiments.

Fundamental Research Questions include:

• What are the advantages and disadvantages of implementing each of the specified preservation approaches?
• What are the factors that affect the effectiveness or appropriateness of a particular preservation approach?
• What are the basic requirements for preservation functions?

These questions can be considered in light of cost, record type, authenticity requirements, and supporting resources, to name but a few.

The subset of Attribute research questions includes:

• What are the options for preserving record attributes?
• What factors affect the preservation of record attributes?

These questions consider attributes in terms of record type, software, preservation approach, metadata implementation, and preservation function implementation. Defining essential preservation metadata is also a priority of the project, and is included in other research questions.

Preservation Approaches

As the Testbed Research Framework notes, record keepers have not yet defined an explicit and definitive methodology for any of the preservation approaches we are considering. The Testbed is contributing to the delineation of various appropriate methodologies. As discussed elsewhere, there are various ways to implement a migration strategy (3). The same can be said for emulation and the vaguely defined "XML approach." Indeed, there are so many different ways to formulate a strategy that the boundaries between them can become indistinct (4). Let's begin by defining the various approaches: migration, emulation, and the XML approach (5).

The Testbed definition of migration is a relatively simple one: the transfer of records from one hardware and software configuration to another. This includes migration through and over generations of application versions, as well as across applications and operating systems, in practice often on either proprietary or part-proprietary software, but excludes media refreshing.

Closely related to this is the XML approach. Conversion to XML (or conversion to standards in general) can be considered to be a form of migration. However, XML is not tied to any particular software system and is often regarded as the most promising present day format for archiving and interoperability. It has a multiplicity of uses, and so deserves to be considered as an approach in its own right.

For the emulation strand of our experiments we will be working with IBM and will perform experiments with their UVC (Universal Virtual Computer) on archival records at the Testbed. The UVC approach is described in an earlier RLG Diginews article. The UVC is a variation of the emulation approach that addresses the problem of future interpretation of data files by writing a program to carry out the interpretation in the language of a Universal Virtual Computer. This strategy can be extended to archiving the program as well, making it more like the full emulation approach identified and proposed by Rothenberg (6).

Methodology

The Testbed is a practical project that runs controlled experiments in a secure environment to study potential preservation approaches, and to ascertain the effects of preservation actions on different file formats and record types. In order to provide valid and reliable results, the Testbed has established a rigorous experiment process, that clearly articulates the requirements for each experiment.

An experiment process is defined around a specific preservation approach and record type, e.g., migration of text documents, or conversion of emails to XML. Each process consists of 12 documented stages. Once a process has been defined and run for the first time, further experiments can be run on that process using its generic requirements and procedures.


For example: Experiment process 1 caters to the migration of text documents. The first five stages in the process are general stages concerned with the broad requirements of the preservation approach (migration) and record type (text documents). These stages define the exploration area, identify relevant background literature, specify authenticity requirements and evaluation criteria, develop an overall experiment process design, and estimate the required resource specifications. The remaining stages are then specific to an individual experiment; that is to say, they specify the records to be used in that experiment (e.g., Microsoft Word 95) and the specifics of the approach (e.g., migration to PDF). Further experiments can be run using the generic documents from the first five stages of the Process. Several further experiments can therefore be run using one experiment process. These may examine different formats of the record type (e.g., Word Perfect or Microsoft Word 2000 as text documents) or they may examine metadata issues.

There are several advantages to this approach. The controlled, well-documented experiment process allows each experiment to be easily reconstructed in the future and allows experiments to be re-run to confirm or check results, if necessary. The process lays out requirements to be considered at each stage, from preservation requirements to functional requirements, metadata requirements to authenticity requirements. The iterative nature of the process allows us to produce a base set of documents for each set of related experiments, thus eliminating any duplication. Grouping experiments in this way also helps us to consider the combined results of sets of experiments and the overall "success" of each preservation approach. The secure and controlled Testbed environment, in which all of the experiments are carried out, ensures that our experiment results are valid and free from errors from extraneous sources. We carry out regular control and null experiments to ensure this remains consistent.

The practical nature of the project brings other benefits as well. The experiment process covers all of the key points in the history of the record, from creation (by way of model records (7)) to capture, appraisal, and long-term storage. These aspects can have unexpected effects on the implementation and success of a preservation strategy. The Testbed considers all of these aspects within the scope of the experiment process, and they have yielded useful and interesting results.

Preliminary Results (8)

Our first experiments concentrated on the migration of text documents. This combination was chosen for the first experiments as the use of text documents is widespread throughout government and many organisations already carry out routine migrations when updating their computer systems. Our main goal in the early experiments was to examine and identify record features that changed as a result of the migration process.

Microsoft Word was identified as a good starting point. It is one of the most widely used word processing applications available and is used by many agencies to produce government records. Experiments have taken place on the migration of text documents through and over generations of Microsoft Word using both model and test records. Migration through generations refers to migrating through successive versions of an application, e.g., from Word 95 to Word 97 then to 2000 and then to 2002. Migration over generations skips the intermediate versions and goes straight to the current highest version, e.g., from Word 95 to Word 2002. We have experimented with migration through and over generations of Word 95, Word 97, Word 2000 and Word 2002. We have also experimented with migration of Word files to PDF 1.2, 1.3, and 1.4.

Results of the model record Word experiments showed that if the record had been created well initially, it stood a far better chance of retaining its features through and after migrations. Fields that update automatically (e.g., date fields) and that were not fixed after document creation wound up being updated every time the document was accessed, thus altering an essential content and context-reference feature. This is a problem whilst the records are still in active use, let alone when they reach the archives. However, most features migrated successfully. The position of the text on the page was sometimes different, but colour, paragraph and font formatting, bullets and numbering, inserted and well-formed tables, hyperlinks, pictures and diagrams were all successfully retained in the experiments we have carried out so far (9).

Use of records donated by government ministries took our investigations to the next level. We had not been involved in the record creation, so we could not be sure of how the features had been formed. One record, which at face value looked like a well-formed table, turned out to be composed of floating text boxes. Other records from different participants had designated "protected sections" in which automated fields had been used and then "fixed" in place. These sections included such essential metadata items as date, author, recipients, and a unique reference. Yet other records were composed on different computers with different settings, or included text that had been cut and pasted from a different application altogether. These "cut and paste" sections are affected differently than the rest of the document during and after a migration and can result in a change in the appearance of the record without adversely affecting its content.

These experiences allow us to examine unexpected record features and to assess more closely the ways in which records are created. As a result, we can formulate advice on the creation of records and use of record-creating applications, putting preservation concerns in place from the beginning of the records continuum.

The set of Microsoft Word experiments showed that generally, migration over generations was at least as reliable—and in some cases more reliable—than migration through generations. This may counter some of the scepticism about the costs of migration. Sceptics have argued that the recurring costs of migration will be too great to bear. Our results so far have shown that migration to each new version of an application should not be necessary, and we hope that experiments with other word processing applications will allow us to extend this hypothesis. The archival regulations of the Netherlands state that ministries are responsible for the authentic retention of their own records for the first 20 years, after which time a percentage of them are sent to the National Archives for long term archiving. The rest can be disposed of according to disposition regulations. It may be the case that Ministries can simply retain the documents in their original format, with maybe one or two controlled migrations, until the twenty years have passed. The Archives can then undertake more suitable long-term action concerning the current and future formats of the record.

This is simply one possibility that we are considering. We still have many more experiments planned, and are waiting until all of the results are in before we release full advice on the long-term preservation of records produced by government agencies. The combined results will allow us to weigh alternative approaches, in conjunction with metadata and authenticity requirements, and determine the best way to implement these approaches. There are many different ways to carry-out the same task and it is unlikely that a "one size fits all" approach will be suitable for different record types with different retention requirements.

Products, Tools, and Deliverables

In addition to the research results, the Digital Preservation Testbed will also develop a more concrete set of tools and products. These include the Testbed Research Database, which supports our experiments and acts as a valuable source of knowledge on many aspects of digital preservation. This database is being built over the course of the project and aims to collect and provide commentary on relevant digital preservation literature. The contents are not limited to publications, but also include public listserv messages, Web sites, presentations, and Testbed references. Wherever possible, we have gathered electronic copies of the research documents and stored them in the Testbed system for reference by the project team. These are supplemented by online resources (for which the URLs are checked on a regular basis) and printouts. The research database is easily searchable and will be a valuable record of the project, as well as a useful resource. An abridged version for online resources only is available on our Web site.

Other deliverables include white papers on each of our preservation strategies. The white paper on migration was published in December 2001, and the XML for preservation paper is scheduled for publication in late summer 2002. These white papers aim to provide a synopsis of current knowledge about each preservation approach, and to delineate ways in which the approach can be implemented. We will also deliver a technical report on the Testbed system itself, for which an extensive set of functional requirements is being developed. The Testbed Newsletter is produced quarterly and the Web site is updated monthly.

Thus far, the Testbed has completed groups of experiments on emails and text documents and is issuing preliminary advice on the short and long-term preservation of emails. Work will soon commence on spreadsheets and databases, for which advice will also be released. The Testbed project is due to finish in October 2003 but preservation research is unlikely to stop there. The Digital Longevity Program will continue to run and coordinate digital recordkeeping and archival efforts for the Government of the Netherlands.

Footnotes
(1) Jeff Rothenberg & Tora Bikson: Digital Preservation. Carrying Authentic, Understandable and Usable Records Through Time. (The Hague, 1999) (back)
(2) The Testbed and the Taskforce are also part of the ICTU, a non-profit organisation established by the Dutch government to house their e-government projects, including PKI (Public Key Infrastructure) and advies overheid (monitoring and advising on government Web sites at every level). See http://www.ictu.nl for further details. The close proximity of these projects allows them to easily collaborate and share information. (back)
(3) Testbed Digitale Bewaring white paper, Migration: Context and Current Status (The Hague, 2001). (back)
(4) See for example Kees van der Meer et al, Emulation and Conversion: Organisational and Architectural Overview of an Electronic Archive (Technical University of Delft and Utrecht University, 2001). This can also be seen by comparing current literature from around the globe dealing with preservation strategies and enforcement. (back)
(5) See the Testbed White Paper on Migration (op cit) for a more extensive discussion on each of these approaches. (back)
(6) Rothenberg & Bikson, op cit. See also Jeff Rothenberg: An Experiment in Using Emulation to Preserve Digital Publications NEDLIB Report Series (The Hague: NEDLIB consortium, 2000) and Avoiding Technological Quicksand: Finding A Viable Technical Foundation for Digital Preservation: A Report to the Council on Library and Information Resources. (Washington, D.C.: Council on Library and Information Resources, 1999). (back)
(7) The Testbed uses two sorts of records in its experiments. The first are model records, which are created in the Testbed to examine and evaluate the effects of preservation action on specific record features (e.g., user-defined and automatic fields, templates, font and paragraph formatting, and signatures). The advantage to using our own records for this purpose is that we know exactly which features are present in each record and where. This allows us to carry out highly focussed experiments on record attributes. It also serves as a good starting point for any new round of experiments. The second sort of records includes test records. These are obtained from ministries and other government organisations, and are used in the larger-scale experiments that address the fundamental research questions and authenticity requirements. (back)
(8) This section is intended to give the reader a flavour of the types of results we have gathered so far. Future reports from the Testbed will discuss our results to a greater extent than this introductory article. (back)
(9) The exact position of the text on the page can be affected by as small a thing as changing the printer or printer driver. This is not a change that has affected the authenticity of the records in any of our experiments to date, but it has resulted in documents containing several more pages than they did originally, especially if page breaks have been employed. (back)