Acknowledging that the rapid increase of digital scholarship also requires new methods of preservation and sustainability for long-term access has been an ongoing topic of discussion since the early days of DH. Discussion on best practices for preserving and curating DH scholarship became more common in the mid-2000s with the impending issue of preserving complex digital objects. In their article, Hunter and Choudhury 2003“focus on the problem of preserving composite digital objects consisting of multiple media types” and introduce the concept of emulation, migration, and metadata as it pertains to complex digital objects, which includes DH. They state, “As scholars in the humanities create increasingly sophisticated multimedia research and teaching resources, they need to capture, collect, and create the documentation or metadata – descriptive, administrative, and structural – necessary to migrate, emulate, or otherwise translate existing resources to future hardware and software configurations” [Hunter and Choudhury 2003]. Cantara (2006) performed an early literature review of digital humanities metadata best practices and tools to automate that metadata, which is significantly different from traditional cataloging [Cantara 2006].

The bulk of individual digital objects being preserved are simple; images, text files, audio/visual materials, all of which we have established best practices for ensuring continuity. We have the Library of Congress Recommended Formats Statement to transform objects into the optimal format for long-term access. We use checksums to confirm fixity and ensure objects are unchanged and uncorrupted. We use audit logs like the PREMIS Data Dictionary for Preservation Metadata to document stages in an object's lifestyle, such as ingest, deletion, restoration, fixity change, virus checking, etc., and then use tools like the Metadata Encoding and Transmission Standard to encode the metadata for broader understandability and interoperability with common systems. We maintain copies, mirror the information technology method for redundancy. All of these practices are guided by digital preservation auditing metrics, most notably the ISO 16363:2012 Space data and information transfer systems — Audit and certification of trustworthy digital repositories (TDR) which also provides guidance on governance, security, and financial sustainability. Good and best practices for common digital objects are straightforward and have well-established guidelines, but when combined into a complex digital object like a website, interactive resource, or software, the basic digital preservation principle of ongoing preservation becomes increasingly complicated.

Another problem at large in hosting digital projects lies within the general scalability and sustainability of the services offered. Conway (2010) explored the concepts of digitizing for preservation and digital preservation and the transition from traditional preservation to digital preservations, leading to dilemmas in what objects are treated differently and how they are treated in a digital environment. Further, Vinopal and McCormick (2013) state that “these services should promote the development of reusable tools, platforms, and methods, and facilitate the creation of preservable, reusable scholarly content to ensure the long-term value of and access to the institution's research.” Their article provides an overview of various researcher needs and the relationship between the services and Libraries of a given institution and provides a Proposed Model for Digital Scholarship Services that includes DH scholarship and how DH programs fit into other institutional services. Leslie Johnston of the National Archives and Records Administration expresses similar thoughts on strategies in preserving DH specifically, stating “sustainability and preservation depend on active management of a project. Digital humanities projects correctly put the highest focus on the content and scholarship that they present. But these projects should not let their technologies go stale” (2013) by relying on technologies and plugins that no longer receive maintenance or community support.

Project failure can occur in many ways, from lack of funding to complete, inaccessible data, little to no preservation or access storage, and lack of technical training, among others. While a project can fail for a number of reasons, we searched for examples of DH projects that had failed due to lack of preservation or sustainability planning. Surprisingly, there is little record of specific projects that have been discontinued or failed as they are no longer accessible. A study conducted by Timothy Vines and his team surveyed 516 articles published over the course of ten years and found that after the first two years of publication, access to scientific data falls by seventeen percent every year [Vines et al 2013]. Their summary was that “broken emails and obsolete storage devices were the main obstacles to data sharing” [Vines et al 2013]. This study examined accessibility and sharing and can be applied to other forms of digital content in that broken links and obsolete storage are obstacles to digital maintenance. In terms of DH, there are some but few examples of failed projects. James Cummings, for example, provided a list of projects he personally supported that had ultimately failed, five of six of which included lack of sustainability planning, funding, or limited technology in their reasons for failure (2020). Jasmine Kirby also discusses the failure of the Sophie 2.0 project at the University of Southern California, an international collaboration with over a million dollars in funding, noting “the focus was on building the tool with the most features and not something that actually could be sustained or even worked” (2019). Even a 2013 JISC study surveyed DH Center Directors regarding long-term maintenance of hosted projects and found difficulty in keeping these sites live [Maron and Pickle 2013]. The study suggests that centers can turn to museums and libraries for long-term hosting and access, but also notes that “many have yet to develop a formal digital strategy” (2013).

In contrast to unsuccessful projects, many long-standing DH projects share specific aspects that led to success. The accessibility of the raw data as a component of the project regardless of the platform and the ability to explore it independently appears to be the key to the success of these long-standing projects. These projects all ensure the data is platform-agnostic and manipulable in other platforms with multiple tools and exists in formats that are interoperable. Many of these projects also contain corresponding documentation to describe the data as another component of the work. For example, the Walt Whitman Archive is built on TEI/XML and simple HTML. The interface has not been updated in some time, however, the success of this project is based on the range and depth of the content, and the accessibility of the raw data. All of the manuscripts, correspondence, and most other content is available for direct download from the Git repository in their original TEI/XML markup. Similarly, the William Blake Archive provides a short API code to query the entire archive independently. Finally, the Charles Algernon Swineburne Project provides basic TEI/XML and XSLT files to explore and recreate the archive.

Comparing these successful and unsuccessful projects prompted us to review other guides on sustainable websites in beginning our case studies. There are several guides on creating sustainable websites available through the Library of Congress’ “Creating Preservable Websites” page, the blog post “On Preserving Digital Culture and Digital Projects” and “Digital Humanities for Tomorrow: Opening the Conversation about DH Project Preservation” (2016) from Digital Humanities at Berkeley, and Stanford Libraries’ “Archivability” guide (Stanford Libraries). This guidance provides some foundational best practices and again acknowledges the need for sustainable and preservable websites, but does not provide a comprehensive solution. Additional guides that are more thorough and include practical information such as the Socio-Technical Sustainability Roadmap at the University of Pittsburgh and the Digital Project Preservation Plan: A Guide for Preserving Digital Humanities [Miller 2019] have been recently developed. However, Kilbride (2015) notes that while the standards have been created to fill the gaps in digital preservation as it pertains to DH, these best practices have not been widely adopted.

The literature review of resources, guides, and projects indicate that increasing access to DH scholarship is a mix of infrastructural and developing best practices, as well as providing education for scholars to learn to adopt best practices. The review also supports the theory of supporting the components of projects for long-term preservation and access in regards to common preservation practices and what has allowed long-standing DH projects to thrive.

Virginia Tech is home to dozens of DH projects ranging in legacy to recent, hundreds to thousands of items, and on a variety of platforms. Recently the Libraries have received requests from faculty and administration to develop sustainability plans for these programs and to promote greater use through easier access to data, formal analytics and branding, and advertising of these projects. The Libraries approached this request by experimenting with the concept of DH as data with comprehensive exports of project content, the creation of necessary metadata, and the creation of project documentation into a single data set.

There are several possible repositories and storage locations for this content. Projects that contain humanities content with additional scholarship and fit the criteria for the institutional repository, VTechWorks. While projects that contain mapping datasets fit the criteria for the Virginia Tech Data Repository. Additionally, any project that is a general product of Virginia Tech research could be ingested into our newer platform the Virginia Tech Digital Libraries Platform. For the both of our case studies, we found them most suitable for the Virginia Tech Data Repository as the final preservation datasets contained a layered file system with a mix of documentation and smaller datasets.

We explored two case studies in the last year. Social Networks in Georgian Britain (SNiGB), a new project branching off of a long-standing project, Lord Byron and His Times (LBT), to enhance the exploration of the content containing over 40,000 items. And the second project, Redlining Virginia, is an Omeka-based project with approximately 110 items that is represented in multiple instances outside of the Omeka site.

We employed data curation methods that resulted in described datasets and content packages consolidated to be portable, explorable, and reused. The two case studies we have conducted so far have produced another access point for the raw data, content, and code, as well as provided another method of tracking impact through the Virginia Tech Data Repository (VTDR) when the Libraries do not have the ownership to apply analytics to a live project. This accomplishes both goals of providing a permanent access location for data regardless of whether the project as a whole was successful (an example of Redlining Virginia is in Figure 1), and of providing a preservation environment for DH data. Impact can also be measured through the number of views and downloads shown in the VTDR interface.

VTDR is hosted on Figshare, which is backed up automatically to the digital preservation service Chronopolis and VTDR manually deposits all datasets into the Academic Preservation Trust digital preservation service of which VTUL are members. We successfully ingested each project’s data into a stable storage location with enough context to reconstruct the entire project in its integrity and to explore individual components. This was accomplished by organizing each individual component into its own folder with associated metadata packaged into a single folder with collection-level documentation and guides on where and how to locate additional resources (an example of Redlining Virginia file structure is in Figure 2, Figure 3, Figure 4, Figure 5, and Figure 6).

Metadata development for SNiGB was straightforward as the raw data is a series of small XML files stored in zipped folders. Redlining Virginia contained georectified maps displaying the evolution of redlining over time. These maps, a combination of JPG, TIF, XML, TFW, OVR, and Shapefile bundles, required more research into appropriate metadata schemes. The final schema chosen was the Content Standard for Digital Geospatial Metadata modified for our georectified maps. From a technical standpoint, researching and developing this metadata was the single most time-consuming phase of this case study. We applied a newly developed Digital Exhibits Metadata Application Profile designed by the Metadata Coordinator and a Preservation Profile designed by the Digital Preservation Coordinator and Digital Preservation Technologist that will also be applied to additional DH projects. Applications of both of these Profiles can be viewed in the Virginia Tech Data Repository preservation packages for Redlining Virginia.

One limitation we did not explore is discoverability. The new documentation developed for these case studies is not discoverable in and of itself and is only found through the data repository via the associated title, keywords, and descriptions. While we established another location for access, we did not have any autonomy in the metadata development to be more efficient in search engine capabilities.

This practice is imperfect. We can only enhance and preserve what we are given or what can be found and exported. From these case studies, it is clear that defining a minimal standard of what is required from the project and researchers, and what is reasonable and possible for the Library to manage, is necessary.

Time and bandwidth were also challenges in these case studies. Two library faculty members dedicated over forty hours of time each to develop and implement workflows for each project. Given the dozens of projects that will require similar efforts in preservation and without considering additional enhancement and large-scale migration, this is not a reasonable task. The immediate outcome from these two case studies was increased evidence and advocacy in VTUL for a permanent support system for digital scholarship in the form of a standing committee . With sponsorship from the Associate Dean and Information Technology Executive Director and support from Library stakeholders and department heads, the newly established Digital Scholarship Technical Review Committee is designed to serve as an advisory and implementation body for both legacy and new Digital Scholarship and web-based projects hosted by VTUL. This committee is currently developing and will implement workflows, policies, and technical infrastructure for hosting digital scholarship, and will also include the appraisal and prioritization of migrating and developing projects. The committee will determine what qualifies as a digital scholarship project that the library is prepared to support, create migration procedures, identify responsible individuals, and implement workflows to transfer projects to the Libraries. This committee will also develop documentation and guidance for creating new sustainable digital projects.

The first qualitative result we found is that accessible and sustainable DH suffers from a lack of education for researchers and a lack of standards and best practices that are usable by novices. Kilbride (2015) suitably remarks: “Consequently there are two abiding risks for those humanities scholars trying to ensure a long-term future for their precious digital cargo: that the emerging practices of digital preservation are a poor fit to the changing needs of humanities; and that jargon and miscommunication confound thoughtful efforts to engage.” The overlap of humanities critical analysis skills and the technical skills to implement a successful and usable project is difficult to attain, particularly for projects hoping to continue beyond a few years. Foundational DH courses in the undergraduate and graduate programs tend to focus on the multiple definitions of DH, tools and techniques, and impact of providing unique public access to otherwise unavailable information. Sustainability, usability, and project management may be introduced, yet the number of DH projects created early in a researcher’s profession that eventually break or disappear is high.

What DH standards and best practices do we introduce to students and new professionals? A traditional example is Nineteenth Century Scholarship Online which provides a formal, peer-reviewed system for evaluating projects, but is limited to projects concentrating on 19th century humanities. Other similar resources are equally contextual and not universally applicable, such as the University of Nebraska-Lincoln’s “Best Practices for Digital Humanities Projects” guide that offers technical best practices, but does not necessarily provide information on documentation or preservation infrastructure.

Within the DH field it is often necessary to learn at least one or multiple technologies from digitization, processing and quality control, metadata, website development, text mining, and so on to create successful digital scholarship. With our two case studies, as well as many other projects at Virginia Tech, researchers are largely self-taught and use popular platforms like Omeka to execute their work. Sometimes these projects are completed without direction or collaboration with the library and are later brought to VTUL for maintenance and preservation. In the case of older projects, many of these were built before VTUL had developed a preservation program, meaning that even those with the best intentions may have received little direction in the area of preservation. Humanists are not trained in information technology, digitization standards, metadata standards, content management systems, storage management, and digital preservation, because that is not generally required in the field. There is a balance between what technology is community-supported, what is simple to adopt, and what is sustainable. What are considered foundational DH skills may not incorporate IT best practices, digitization guidelines, or project management. The problem is paradoxical - what services can a library provide to its faculty and students when there are no defined best practices to teach? The lack of consistency between institutions and initiatives in their approach to DH is partially what makes DH so dynamic, but also causes formalization to be even more difficult than in a more established field of study.

The second result we found was an inconsistent approach to project building, indicating a lack of DH project infrastructure available to DH-ers. Like many other digital scholarship projects, website creation, and digital resources, DH projects are typically not designed to be sustainable and preservable. DH has grappled with insecure and unsupported technology because many creators are drawn to trendy technology that may not be sustainable [Owens 2014]. DH is often driven by new or easy-to-use technology that may not have a built-in infrastructure yet, when the goal of the project should drive the technology choices. Butler et al explored this integral infrastructure issue in 2019 in the two-part blog series “Archiving DH”. Taking a new technology and applying it to brand new research is a unique benefit for exploring options, but the lack of infrastructure leaves these projects unsustainable over time [Butler, Visconti, and Work 2019]. In “Archiving DH Part II: The problem in detail”, Butler et al also discuss what defines “authentic access and use over time for digital scholarship” as websites cannot last forever without human intervention and maintenance. The authors discuss what the ideal point in the life cycle of scholarship is to introduce “preservation, sustainability, and portability” [Butler, Visconti, and Work 2019]. The piece on portability is highly relevant to the existing DH projects at VTUL as the Libraries receive many requests to transfer ownership of projects, and in some cases, this can mean transferring the platform of the project.

As a culture, DH is open access in that many of the projects are openly available online. Additionally, funding from sources like NEH require grant awardees to produce open access projects [Brennan 2020], while publishers like the International Journal for Digital Humanities are exclusively open access, and more often than not, DH is designed to highlight new content and research that only benefit from open distribution. However, while a project and its corresponding data may be openly available, if the code, algorithms, documentation, and metadata are not, how impactful and usable is the project in the long-term? Preservation is tricky, you cannot just “preserve it” because the essence of what matters about “it” is something that is contextually dependent on the way of being and seeing in the world that you have decided to textual components alone are not enough to accurately enhance, reproduce, or preserve a DH project.

There are also the typical digital object collection problems to consider regarding storage, access platform, and responsibility for maintenance. For traditional scholarship we have journal databases and institutional repositories, and for data we have databases and data repositories, but a best practice for storing and maintaining DH projects, which is often a combination of scholarship and data, is still evolving. The choice we made to store our DH preservation packages in the Virginia Tech Data Repository was based on the best choice from the resources we had available, not because it’s a perfect DH storage solution.

At VTUL we want to provide access to a preservation package that contains the text, context, code, metadata, and infrastructure to reproduce and reuse. This preservation package is designed to provide all of the components of a project and guidance on how to reuse, rebuild, and validate DH projects and data. In terms of digital preservation specifically, in the Open Archival Information System (OAIS) Reference Model, the Submission Information Package would be created by VTUL, the Archival Information Package is stored in a preservation system (Virginia Tech Data Repository), and the Dissemination Information Package is hosted in a publicly accessible location (the live website). In many Virginia Tech DH projects, and many DH projects in general, the main deliverable is equivalent to the Dissemination Information Package with little regard to raw data or preservation and sustainability from the lack of technical infrastructure.