The longevity of digital editions: are we wasting taxpayers' money?

Despite its fragility and susceptibility to damage, paper is a remarkably durable information carrier (21). The disasters and wars that have affected Poland did not manage to destroy all of our printed collections, and volumes of historical source editions published in the nineteenth and early twentieth centuries remain available to researchers and students. We can expect them to remain accessible in another fifty or even one hundred years. We are, however, living through an information revolution that has also transformed historical research. Research results and historical studies are increasingly available in digital form. The same applies to source documents: digital editions now appear alongside printed editions and sometimes replace them entirely. This interactive presentation of sources significantly expands research possibilities, while online access brings them to a wider audience. But will digital editions prove equally durable? Will researchers one hundred years from now still be able to use the work of scholars preparing an edition of sixteenth-century correspondence today? The paradox is that historical sources have survived for centuries and are likely to survive many more, while their scholarly editions, often produced by teams of researchers and developers through expensive, multi-year grants, face a much shorter life expectancy (22). The ephemeral nature of digital objects raises questions within the scholarly community about stable citation and the status of digital editions as reference works.

Longevity is not solely a problem for digital editions. It also concerns research data management, software sustainability, and digital humanities infrastructure. Services, websites, and applications created by projects completed several or even a dozen years ago regularly disappear from the web. Some applications remain online but no longer work correctly, for example because external services on which they partly depended have ceased to operate (1). A digital edition is more than text. It also includes supplementary data, the source code of the application that presents the text, servers, and online services. Every component must survive if the original edition is to remain available in the form in which it was created and published, giving users the experience intended by its authors. Those intentions are themselves dynamic: digital publication makes continual change possible and encourages the adoption of emerging technical capabilities (25), which does not make stable long-term maintenance any easier.

The challenge of maintaining digital humanities projects over the long term was recognized more than a decade ago. It involves not only technical questions but also organizational and financial ones (2). A project may receive maintenance support for several years after its research phase, but long-term survival is realistic only when it is anchored in a stable organization or institution (9). Except in catastrophic circumstances, a printed book issued in enough copies simply continues to exist and remains accessible. A publication delivered as a web application requires continuous support and care. In practice, the longevity of a digital edition means active maintenance, which goes beyond preserving digital objects as files in the sense defined by the OAIS (Open Archival Information System) standard (3). Information systems age quickly. Rapid technological development, as well as evolving security threats, makes it difficult to provide a stable environment for web applications. New versions of databases, web servers, programming languages, and frameworks introduce not only features but also security patches that need to be installed. New releases can be incompatible with older ones, making upgrades and maintenance considerably more complicated and expensive.
From a user's perspective, an application may continue to run correctly on its server, yet a prolonged lack of updates can make it incompatible with current web browsers and create usability or security problems.

How serious is the disappearance of digital editions? Michael Kurzmeier examined the issue in a presentation at the 2023 IIPC General Assembly and Web Archiving Conference (5). His study used Patrick Sahle's catalogue, which then contained more than 800 editions (6). Eighty-five per cent were accessible, a fairly optimistic result, although the study primarily checked URL status rather than whether the editions still functioned. An earlier 2019 poster based on the same catalogue estimated the average lifespan of an edition at 8.5 years (32). Another study estimates that digital editions disappear at a rate of five per cent per year and that between 25 and 50 per cent of editions published before 2000 are no longer available (7). The same article argues that copies captured by the Internet Archive's Wayback Machine (8) cannot necessarily be considered preserved digital editions. Web archiving usually captures web-facing files such as HTML and images, but not the XML files that often constitute an edition's data layer. Preserving an edition in a form different from the original may also flatten it, as happened when the XML edition of the Kodeks dyplomatyczny Wielkopolski, originally distributed on CD-ROM, was converted to HTML (10, 11).

Technological ageing is a different phenomenon from an edition disappearing from the internet. An application may continue to work as originally designed while no longer being suited to modern devices and software. Default font sizes may be too small, content may occupy only a small part of today's high-resolution displays, and the interface may no longer match users' expectations. The edition remains available and functional, but becomes less convenient to use (13). Technological obsolescence often contributes to a project's eventual disappearance, although it may interact with many other factors: growing technical debt, project members leaving the institution responsible for maintenance, or even the closure of the institution itself (15). Organizational changes or a new service address alone can make an edition difficult to find. Link rot is one of the persistent problems affecting online content; another is content drift, where an edition changes its structure or content after it has been cited (26).

How can a digital edition be preserved?

In 2022, the German Institut für Dokumentologie und Editorik published a manifesto on creating and maintaining digital editions. Among other points, it argues that editorial work does not end with publication, that digital editions must be preserved permanently and kept accessible, and that sufficient human, institutional, and financial resources must be allocated to long-term archiving and availability (14).

The Endings Project at the University of Victoria in Canada addresses precisely this problem: the long-term preservation of digital humanities projects (4). Its premise is that our ability to create digital content now exceeds our ability to preserve it. Digital humanities therefore needs methods and technologies that secure and archive projects while keeping them usable over time. The project produced recommendations and strategies for software development and digital project storage that aim to preserve not only files, but also the dynamic character of digital projects. Its principles concern:

  • data stored exclusively in open standards and version-controlled repositories;
  • documentation, including thorough documentation of the data models used;
  • software with few external dependencies, preferring static websites and avoiding server-side software where possible;
  • avoiding niche and fashionable technologies;
  • a simple application structure;
  • clear and consistent release management for the product, such as a digital-edition application.

In 2019, Oltmanns defined four levels of digital-edition preservation (1):

  • 1. Continuous maintenance: fully maintaining the working digital-edition application, its environment, and infrastructure. This is the most demanding level because it requires funding and organizing what is effectively the application's continuing technical development so that it keeps pace with changes in its technological environment.
  • 2. Application preservation: retaining the application environment in a virtual container, such as Docker or a complete virtual machine. Over a limited period, perhaps five to ten years, this may keep an edition running without modification or upgrades, although it raises security concerns.
  • 3. Application-data preservation: ensuring survival at the data level only. The data remain accessible even if the edition itself no longer works. They should be made available in accordance with FAIR principles and maintained when necessary, for example by migrating them from an obsolete format to a current one.
  • 4. Bitstream preservation: storing frozen copies of the application code, data, and configuration, preferably in several locations, so that the edition might be reconstructed in the future.

Two main approaches emerge from these recommendations: preserving the complete working application that constitutes the edition, or preserving its most important component, the data. A reasonable compromise may be to retain the data, application source code, configuration, and detailed documentation in the hope that the edition's functionality can later be reconstructed. Knowledge itself can disappear, however. In several decades it may be difficult to find specialists familiar with a particular technology, as illustrated by NASA's challenges in operating the Voyager probes (17).

Reducing technical complexity and publishing a static site are ways to lower the risk of disappearance and extend the period for which an edition remains a complete working application (23). One approach has been to rely on basic technologies such as HTML, XML, and simple JavaScript. This usually requires compromises in functionality and decisions about which features are essential; search and filtering are typically critical (27). The Endings Project proposed one possible solution in its staticSearch generator (29). Perhaps, as Roopika Risam writes, "Minimal computing is the answer. Minimal computing is easy..." (28). A minimalist approach may also be more ethical because it consumes fewer environmental resources.
A typical contemporary digital edition nevertheless contradicts the principle of minimalism. It consists of many layers: data, a database, server software, a user interface, a search mechanism, and network infrastructure. The most minimal preservation strategy retains only the underlying data, which could later support a new edition, perhaps with a similar form and functionality.

TEI, an XML format first introduced in 1988, is the standard data format for scholarly editions. It was designed to produce transcriptions independent of any particular application and resistant to technological change. The TEI Consortium describes TEI as the most systematic attempt to develop standards for scholarly heritage in a changing digital culture (18). TEI can preserve text, semantic data, and the scholarly treatment of a manuscript, thereby retaining the core of a digital edition. It is also a vast and complex standard, which makes appropriate normalization and documentation practices particularly important (20). Like software, standards such as TEI evolve. The transition from TEI P4 to P5 proved challenging for many projects based on the older version (24).

The rapid evolution of the software used to deliver editions makes long-term maintenance difficult. Such software is often developed as a custom project. Publicly available code and documentation help, but the resulting technical barrier still threatens longevity. TEI Publisher therefore became a popular platform for creating and presenting editions. It is an open, community-supported tool based on TEI and XML. Its goal was to standardize digital-edition projects and make software components reusable instead of requiring every project to build them from scratch, thereby addressing stable maintenance. Standardizing the publication platform does not, however, solve software ageing automatically. TEI Publisher itself changes and requires migrations and upgrades. It also relies on eXist-db, an XML object database that is neither simple nor widely understood. As a popular framework, TEI Publisher is likely to be more durable than bespoke software created for a single edition (19), but it imposes a particular model of publication that may not suit every editor.

The most durable digital edition will not be the one with the most impressive application, but the one that clearly separates its data and presentation layers and makes it possible to preserve both, even by different methods. Established standards and extensive documentation matter for each layer. Availability and maintenance should themselves be treated as proper outcomes of a digital-edition project, not as incidental technical details (7).

Could the solution to the longevity problem lie in hardware rather than software? Joris van Zundert's QUEDEE project begins with this premise. Today's digital infrastructure and the cloud services supplied by research and educational institutions may not last forever; using them as the foundation of a digital edition may be a mistake. QUEDEE is a long-term experiment asking whether a digital edition can be made to resemble a book in durability rather than a typical web application. The edition would therefore be a physical device, not merely an application: a Raspberry Pi with a small touchscreen. It would also provide access through its own Wi-Fi network as part of a strategy for controlled degradation. If the display failed after many years, remote access could continue. Duplicate memory cards would store the edition, while built-in solar panels would allow independent charging. The proposal is deliberately provocative and may broaden the debate over whether a digital edition can become a durable cultural object comparable to a book (30).

This project raises a broader issue. The problem is not only the longevity of individual digital editions, but also that of the digital infrastructure on which they depend. Research grants often fund maintenance of their outputs, including digital editions, for a fixed period. But what happens after those three to five years? A research institution cannot function without sustained funding; without stable funding it cannot maintain its digital infrastructure; and without stable infrastructure there can be no durable digital editions. Infrastructure here means not only hardware, software, and electricity, but also skilled staff capable of maintaining and caring for it (31).

Conclusion

Do the many maintenance problems mean that digital editions are not worth creating? Certainly not; that is not the purpose of this discussion. Long-term survival remains a genuine challenge, but principles and guidelines derived from more than two decades of experience can help. A digital edition is not merely a digitized version of a printed one. It can and should offer more, using technology to enrich the source and make the scholarly work easier to use. This utility has been cited as a reason why electronic editions may displace traditional ones (16). Yet digital editions are currently far less durable than their paper counterparts. A digital edition is usually maintained in one location, on a server, from which it can reach a worldwide audience. A printed edition exists in many copies, albeit a limited number, that readers must physically reach. Digitized copies of printed books, such as PDFs, now provide an intermediate form. If the server hosting a digital edition is shut down or destroyed, the edition becomes entirely unavailable, or only fragmentarily accessible if at least its data were archived. If one printed copy is destroyed, other identical copies remain. There is also a financial difference: maintaining a completed digital edition can be expensive, while the cost of keeping an individual printed book is low (32).
Returning to the title of this article, creating digital editions is not a waste of taxpayers' money. Abandoning a traditional printed version may be, at least until standards exist that can secure a digital edition's survival for the next one hundred or one hundred and fifty years. For now, we must continue improving and applying methods and guidelines for creating long-lived editions. Better funding for maintaining the outputs of digital humanities projects would also help.

Piotr Jaskulski
(the views expressed in this post are those of the author,
not of the Digital History Lab as a whole)

Bibliography: