Automatic Evaluation of Migration Quality in Distributed Networks of Converters
Migration has always played an important role in digital preservation. The most recent developments in this context introduce networks of distributed converters. In this paper we propose an extension to current migration networks to enable users to: a) determine to what extent the essential characteristics of digital objects have been preserved during a migration; b) generate detailed migration reports for inclusion in the objects' preservation metadata; and c) provide advice on the available migration paths or target formats that will best suit users' requirements.
A significant part of the intellectual output created today is being produced in some sort of digital form. The simplicity by which this type of material can be created and disseminated over current computer networks and the high quality of results are significant factors that lead to the massive adoption of digital authoring tools. However, digital material suffers from a structural problem that puts its longevity at risk. Although a digital document may be copied an infinite number of times without loosing quality, it requires a technological environment in order to transform its computational representation (i.e. bit stream) into something that human beings are able to understand. This dependency on technology makes this type of material vulnerable to the rapid and unpredictable obsolescence that affects all technological artifacts .
The usable lifetime of a digital document is, in general, several times lower than that of a similar document stored in an analogue medium. A paper document, for instance, can be preserved for more than 100 years and a microfilm, if stored in appropriate conditions, will still be readable in 500 years time . On the other hand, the life expectancy of a modern technological platform may be as short as 5 years [3-6].
Digital preservation is here defined as the set of processes and activities that ensure the continued access to information and all kinds of cultural heritage existing in digital formats . In this context, a digital object is "(...) an information object, of any type of information or any format, that is expressed in digital form" . This definition is general enough to accommodate both information that was created within a technological environment (born-digital objects) and information that was obtained from analogue sources (digitalised objects). Text documents, digital photographs, vector graphics, databases, video/audio sequences, virtual reality models, Web pages and computer games or software applications, are a just few examples of what may be considered digital objects.
This paper is organised as follows: section 2 provides some background on the most prominent digital preservation strategies; section 3 describes the proposed research where a special emphasis is put on the research questions, methodology and expected contributions; section 4 outlines a few points for discussion during the doctoral consortium; and section 5 summarises the proposed research.
2. Related work
Over the last decade, a considerable number of strategies have been proposed aiming at solving the problem of digital preservation and technological obsolescence. According to Lee et al., the available strategies may be aggregated into three major categories: emulation, encapsulation and migration1 .
2.1 Strategies for digital preservation
The emulation strategy involves the use of special software to reproduce the behaviour of a given hardware/software platform in a different technological environment . This strategy allows the user to interpret digital objects by running the same software that was used in their creation [9, 10].
Emulation plays an important role in preservation, especially in the interpretation of highly dynamic and/or interactive objects, such as software applications .
The encapsulation strategy aspires to solve the problem of digital preservation by storing together with the object, enough information about how it should be interpreted . This information may consist of a formal specification of the object's format that will enable the future development of viewers, emulators or converters. Lorie argues that this formal specification should be expressed in some sort of machine language compiled for a virtual machine [13, 14]. If the virtual machine is thoroughly documented, it should be possible to recreate it in a future computing platform.
The encapsulation strategy is generally oriented at collections of objects that are expected to remain unexploited for long periods of time. Using encapsulation on such collections may reduce the number of preservation interventions leading to a generalised reduction in preservation costs.
The migration strategy consists of a "(...) set of organized tasks designed to achieve the periodic transfer of digital materials from one hardware/software configuration to another, or from one generation of computer technology to a subsequent generation." . Contrary to other preservation strategies, migration does not intend to maintain the digital object in its original format. Alternatively, it converts the object from a near obsolete format to a format that users are able to interpret with their up-to-date computers. The main disadvantage of this approach lies with the fact that when an object is migrated, some of its properties may not be completely or adequately transferred to the target format, i.e. some data loss may be experienced. The reason for this is twofold: there may be structural incompatibilities between the source and the target formats, or the converter may be faulty and therefore incapable of performing its task appropriately.
There are a considerable number of migration variants. The most relevant ones comprise normalisation, migration on request and distributed migration.
Normalisation consists of the conversion of digital objects from a wide range of distinct formats to a smaller and more manageable set of formats. The rationale behind it is based on the idea that by reducing the number of formats in custody, subsequent preservation interventions will be simplified, as the same strategies may be applied to a higher number of objects. Normalisation is usually performed by the digital archive during the ingest process of new material or by the producer of the materials before the archival process is initiated [6, 15].
Migration on request aims at reducing the problem of data loss by applying all necessary transformations to the original object instead of performing such transformations on the derivatives that have resulted from previous migrations . In this strategy the focus is not on the preservation of the digital object but in making sure that the migration tool remains usable over time .
2.2 Distributed migration
The most recent endeavours in the field of migration propose architectures of remotely distributed converters. The Typed Objects Model proposed by Ocklerbloom describes a network of mediator agents that assist users in the discovery and execution of object conversions . Other initiatives explore similar concepts but resort to more widely spread technologies. The Lister Hill National Center for Biomedical Communications has developed a Web Service that enables anyone to convert objects from fifty different formats to PDF . Hunter and Choudhury propose a distributed network of Web Services that provide advanced migration services to users . This network is supported by a semantic layer that enables software clients to autonomously find and invoke desired migrations.
This type of migration introduces several advantages over more traditional solutions: the use of Web Services hides the complexity of specific converters as well as the peculiarities of the technological platform that supports them; the development of redundant services insures that the network remains functional during situations of partial break down and the existence of multiple migration paths enables this solution to cope with the gradual disappearing of converters. Moreover, this approach is compatible with several variants of migration, such as migration on request and normalisation. The creation of a global network of converters may also result in a generalised reduction in preservation costs. Any willing organisation will be able to gain profit by publishing their converters on the migration network and charging a small fee for its utilisation.
2.3 Evaluation of preservation strategies
Although the number of proposed strategies for digital preservation continues to grow, none of them has been adequately proved or universally accepted . A preserver's decision for or against a specific preservation strategy is usually accompanied by certain amount of discomfort. Unfortunately, his/her decision is not only conditioned by the lack of universal acceptance. Several factors, such as the satisfaction of the designated community, the characteristics of the collection an the preservation costs, must also be taken into account .Rauch and Rauber have built a conceptual framework that allows preservers to evaluate, compare and select preservation strategies according to their individual requirements [20-22]. Their work is based on the ideas of Utility Analysis , which was originally developed for infrastructure and economic research projects. The framework follows a process that is composed of the following steps:
3. Ongoing research
In a previous section we described how networks of distributed converters may play an important role in the automation of preservation processes. In such networks it is likely that different sequences of converters may originate objects in the same target formats. However, distinct migration paths are likely to produce considerably different objects (Fig. 2). Moreover, different users are likely to have different expectations regarding the results of a preservation intervention. For instance, some users may be interested in preserving most of the characteristics of a digital object, while others, with more limited budgets, may have to give up on some of the characteristics in order to reduce preservation costs.
In this context, we propose an extension to current migration networks by introducing a new set of services that will help users perform their preservation tasks more effectively. These services will enable users to:
In addition, the system will be able to scale smoothly by supporting the inclusion of new quality indicators as well as new migration services. Fig. 3 depicts the top level components of the proposed system. The Metaconverter component constitutes the interface between the user (i.e. the preserver) and the proposed system. It is responsible for the aggregation of information about available migration networks (i.e. service registry), identification of transitive conversions on those networks, invocation of appropriate converters and the registration of information about performed migrations (e.g. cost, time, reliability).
The Migration Adviser is responsible for generating appropriate advice on which migration paths or destination formats will best suit the preservation requirements of the user. It accomplishes this by confronting the weights assigned by the user to each of the supported criteria, with the information collected overtime in the Migration Knowledge Base about the quality of each migration path. These reports will be processed and stored in the Migration Knowledge Base. They will also be delivered to users, allowing them to acknowledge the quality of a specific migration and to document the preservation intervention. At the moment we expect to base these reports on the PREMIS Data Dictionary for Preservation Metadata .
The digital objects and their converted versions will be compared and evaluated according to multiple criteria. Rauch and Rauber [20-22] have identified several quality criteria for at least three types of digital objects: e-journal documents (63 criteria), audio files (136 criteria) and video sequences (324 criteria). Examples of such criteria may be expressed in the form of questions: how much of the content has been preserved?; which appearance items were not effectively preserved during the conversion (e.g. font style, colour, frame rate, bit rate)?; is the target object bigger than the source object?; what is the stability/support/openness of the target format?; etc.
The Migration Adviser is responsible for generating appropriate advice on the migration paths or destination formats that will best suit the preservation requirements of the user. It accomplishes this by confronting the weights assigned by the user to each of the supported criteria with the information collected overtime in the Migration Knowledge Base about the quality of each migration path.
3.1 Research questions
We claim that is possible to develop a system to assist users in the selection and execution of preservation interventions based on migration strategies. The proposed system will be able to operate autonomously and integrate seamlessly with current and future distributed migration networks. The system will be available as a set of services enabling any digital repository system to expand its functionalities by remotely invoking the externalised funcionality.The central research questions in this work are as follows:
These research questions are closely related to the Migration Evaluator and Migration Advisor components previously described.
3.2 Research methodology
We intend to prove the concepts expressed in our research questions (automatic quantification of data loss and automatic recommendation of migration strategies) by developing a prototype of the described system and by running a controlled set of experiments in order to empirically attest its viability.
3.3 Proposed experiments
Rauch and Rauber have conducted several experiments to assess the effectiveness of their framework for evaluating preservation strategies [20-22]. These experiments were conducted at different institutions that had collections of objects that they wanted to preserve. Using the evaluation framework, Rauch and Rauber were able to compare several distinct preservation approaches and rank them according to the specific requirements of each institution. Unfortunately, the collections of objects used in these experiments are not publicly available; therefore, we intend to reproduce some of these experiments in order to collect useful empirical data that will enable us to validate our system.
We expect to validate the concept of automatic quantification of data loss occurred in a migration by comparing the evaluation reports produced by the Migration Evaluator component with analogous work carried out by a group of human experts. The experiment consists of assembling a collection of objects, migrating these objects to different formats using distinct migration paths and having a group of experts look at the outcome in order to evaluate the amount of data loss that occurred in each of the essential characteristics of the migrated objects. Human evaluations will then be compared with the evaluations performed by the Migration Evaluator component.
The concept of automatic recommendation of migration alternatives will be validated by applying the evaluation framework proposed by Rauch and Rauber to a collection of objects and a set of available migrations, and comparing its results with the recommendations produced by the Migration Advisor.
3.4 Key contributions
The main contributions that may result from this research are:
By attending this Doctoral Consortium, we expected to obtain feedback from senior researchers on the following topics:
In this paper we propose a system to enable users to select optimal migrations or target formats from a distributed network of conversion services. The proposed system accomplishes this, by automatically comparing, according to multiple criteria, the converted objects with their original counterparts and by recording information about the migration process itself. The results of these evaluations will be accumulated in a knowledge base thus enabling the recommendation of migrations capable of fulfilling the preservation requirements of the user.
The documentation of preservation interventions is a fundamental tool assessing the authenticity of the preserved objects. For that reason, the system will be able to generate detailed migration reports in appropriate forms for inclusion in the objects' preservation metadata.
A note about the author
Before beginning his Ph.D., Miguel Ferreira worked with José Carlos Ramalho in the design and implementation of the DigitArq system; a system designed to help archives in the creation and management of finding aids and digital objects. This project was recently awarded with a national Information Society Prize (Prémio Fernandes da Costa). Part of the outcome of this project is currently being deployed in dozens of public institutions in Portugal.
After the DigitArq project, Ferreira started working at the University of Minho as a researcher in the field of Digital Libraries. His work was developed under the supervision of Ana Alice Baptista and resulted in the creation of several add-ons for the HP/MIT DSpace platform. Some of these add-ons are currently being integrated in the official distribution of DSpace.
Currently, Ferreira is working on his Ph.D. project within the Information Society Research Group. His work is being supervised by Ana Alice Baptista from the Department of Information Systems and José Carlos Ramalho from the Computer Science Department of University of Minho. The research group that he integrates has a wide range of research interests, covering works from computer science to social science. He is part of a team of four students who share doubts, knowledge and ideas as they pursue their individual research goals in the field of digital libraries.
The work depicted in this paper is in its development phase where the top level components will be constructed in the following sequence: first, the Metaconverter; second, the Evaluator and finally, the Advisor component. Some conversion Web services as well as proxies for existing migration networks are currently being developed by a group of graduate students.
In parallel with the development of the system, we expect to clearly identify the experiments that will be carried out for demonstration and validation purposes. The identification of these experiments as well as the definition of the significant properties that will be considered by the system will enable us to limit the scope of the ongoing research.
The author maintains fruitful connections with the Documentation Services of the University of Minho, where an institutional repository is currently in production , as well as with the National Archives  and the National Library of Portugal . Since these institutions hold massive collections of digital objects with increasing preservation needs, we expect to find interesting real case scenarios for testing and validating our research.
This work was supported by the FCT under the grant SFRH/BD/17334/2004.
1. In this paper, the term conversion will be used to address a one-step transformation between two distinct formats, and the term migration to describe transformations that involve multiple sequential conversions, i.e., composite conversions.
 Chen, S.-S., The Paradox of Digital Preservation. IEEE Computer, 2001. 34(3): p. 24-28.
 Burkel, R., The Role of Microfilm in Information Management. Information Management Journal, 2003. 37(1): p. 58-65.
 Task Force on Archiving of Digital Information, Commission on Preservation and Access, and Research Libraries Group, Preserving digital information: report of the Task Force on Archiving of Digital Information. 1996, Washington, D.C.: Commission on Preservation and Access. 59.
 Conway, P., Digitizing Preservation: paper and microfilm go electronic. Library Journal, 1994. 119: p. 42-45.
 Graham, P. Preserving the Digital Library. in Long term preservation of electronic materials - a JISC/British Library Workshop. 1995. University of Warwick: British Library Research and Development Department.
 Hedstrom, M., Digital Preservation: A time bomb for digital libraries. Computers and the Humanities, 1998. 31: p. 189-202.
 Webb, C., Guidelines for the Preservation of Digital Heritage. 2003, United Nations Educational Scientific and Cultural Organization - Information Society Division.
 Thibodeau, K. Overview of Technological Approaches to Digital Preservation and Challenges in Coming Years. In The State of Digital Preservation: An International Perspective. 2002. Washington D.C.: Documentation Abstracts, Inc. - Institutes for Information Science.
 Lee, K.-H. et al., The State of the Art and Practice in Digital Preservation. Journal of Research of the National Institute of Standards and Technology, 2002. 107(1): p. 93-106.
 Rothenberg, J., Commission on Preservation and Access, and Council on Library and Information Resources, Avoiding technological quicksand: finding a viable technical foundation for digital preservation: a report to the Council on Library and Information Resources. 1999, Washington, DC: Council on Library and Information Resources. vi, 35 p.
 Woodyard, D. Digital Preservation: The Australian Experience. in Third Conference Digital Library: Positioning the Fountain of Knowledge. 2000. Malaysia.
 Digital Preservation Testbed, Migration: Context and Current Status. 2001: The Hague.
 Lorie, R.A. Long Term Preservation of Digital Information. in First ACM/IEEE-CS Joint Conference on Digital Libraries (JCDL'01). 2001. Roanoke, Virginia, USA: ACM.
 Lorie, R.A. A Methodology and System for Preserving Digital Data. in Second ACM/IEEE-CS Joint Conference on Digital Libraries (JCDL'02). 2002. Portland, Oregon: New York: ACM Press.
 Hodge, G. and E. Frangakis, Digital Preservation and Permanent Access to Scientific Information: The State of the Practice. 2004, International Council for Scientific and Technical Information & CENDI, available from <http://www.icsti.org/digitalarchiving.php>.
 Mellor, P., P. Wheatley, and D.M. Sergeant. Migration on Request, a Practical Technique for Preservation. in ECDL '02: 6th European Conference on Research and Advanced Technology for Digital Libraries. 2002. London, UK: Springer-Verlag.
 Ockerbloom, J.M., Mediating Among Diverse Data Formats. In School of Computer Science. 1998, Carnegie Mellon University: Pittsburgh. p. 164.
 Walker, F.L. and G.R. Thoma. A Web-Based Paradigm for File Migration. In IS&T's 2004 Archiving Conference. 2004. San Antonio, Texas, USA.
 Hunter, J. and S. Choudhury. A Semi-Automated Digital Preservation System based on Semantic Web Services. In Joint ACM/IEEE Conference on Digital Libraries (JCDL'04). 2004: ACM.
 Rauch, C. and A. Rauber. Preserving Digital Media: Towards a Preservation Solution Evaluation Metric. In International Conference on Asian Digital Libraries. 2004. Shanghai, China: Springer.
 Rauch, C. et al. Evaluating preservation strategies for audio and video files. In DELOS Digital Repositories Workshop. 2005. Heraklion, Crete.
 Rauch, C., Preserving Digital Entities - A Framework for Choosing and Testing Preservation Strategies. In Institute for Software Technology and Interactive Systems. 2004, Vienna University of Technology: Vienna.
 Weirich, P. et al., Decision Space: Multidimensional Utility Analysis, ed. Cambridge University Press. 2001, Cambridge.
 Caplan, P. et al., Data Dictionary for Preservation Metadata. 2005, PREMIS Working Group (OCLC/RLG).
© Copyright 2006 Miguel Ferreira