Open Infrastructure and the Threat of “Vanishing” Journals: Leveraging Open Knowledge Commons, Open Source Software, and DIY Solutions to Preserve Humanities and Social Sciences Research

The Threat of “Vanishing” HSS Resources

The first issue we aim to counter is the threat of “vanishing” or “disappearing” HSS resources, particularly those that are open access. This threat persists despite the important work that is being done by preservation networks or schemes, as discussed below. Additionally, though, this threat speaks to—and is an example of—the more general issue of the preservation of knowledge over time and across media. How might researchers leverage digital research infrastructure and open source software to better preserve the knowledge they produce? And why is this work of preservation especially imperative when it comes to open access research in the humanities and social sciences?

Although preservation in the broader sense is not the subject of the current study, it is worth nothing what this term entails in our own (academic) context: as Martin Paul Eve suggests, “Preservation activities cover everything from running server infrastructures to storing extra copies to ensuring that material is sent to archives at the time of publication” (2024, 4). Of note here, both digital infrastructure and the storage of extra copies of research materials by these infrastructures are important elements of what the Digital Preservation Coalition defines as “long-term preservation” (quoted in Eve 2024, 4). And while the objectives and means of preservation and archiving may overlap—frequently enough that the two terms are sometimes used interchangeably—we understand preservation to mean a wide range of activities (e.g., creating backups, migrating materials to new storage sites) that may or may not lead to archiving. Both are concerned with safeguarding physical or digital materials, but we use “preservation” to refer to activities geared towards safeguarding materials while also keeping them accessible for active research in the present; by contrast, we use “archiving” to refer to activities that are comparatively less interested in providing ready access but more interested in long-term preservation and accurately preserving the original contexts and/or physical media associated with an object (e.g., through detailed description, recording provenance).

Admittedly, there may not always be a clear line delineating preservation and archiving. Our provisional definitions here are meant primarily to distinguish our own activities—which were focused on duplicating digital research materials while providing ready access to them—from those of trained archivists better able to preserve these materials “in perpetuity” and in line with exacting cataloging rules or archival description standards. However, because digital preservation aims to provide easy, sustainable access to data, it can also be exacting, and have a high bar for participation, in the sense that it must grapple with challenges posed by variables such as media degradation and differences in data formats and operating systems (Johnston 2020). Additionally, and more relevant in the case of our own preservation project, studies by Katrina Fenlon et al. (2025) and Nick Ruest et al. (2021) discuss how digital preservation often requires collaboration with user communities or collaborators to figure out their data needs and build reciprocal relationships to support sustainability.

For its part, as Canadian digital research infrastructure, the HSS Commons aims to participate in long-term preservation for open access resources by storing additional copies of the resources, whether or not they are already backed up elsewhere. And there are obvious, well-documented benefits—including discoverability, community building, among many others—to preserving or indexing materials in multiple places, beyond just the benefits we had in mind for our project: to build community around specific sets of scholarly resources and to promote serendipitous discovery of open access research in the HSS Commons environment. Indeed, the importance of preserving research by backing it up in multiple places is the driving idea behind so-called preservation systems or networks such as LOCKSS (“Lots of Copies Keep Stuff Safe”).

Preservation schemes or systems such as LOCKSS (https://www.lockss.org), Controlled LOCKSS (CLOCKSS; https://clockss.org), Portico (https://www.portico.org), and the Public Knowledge Project (PKP) Preservation Network (https://pkp.sfu.ca/pkp-pn/) serve a vital role: they ensure that digital content is stored in duplicate and safeguarded in the event that, for example, a publisher folds; a natural disaster destroys or incapacitates a server or other digital infrastructure; data are compromised, lost, or intentionally destroyed as a result of war or geopolitical tensions; or files become inaccessible due to technological obsolescence.⁴ Both LOCKSS and the Directory of Open Access Journals (DOAJ) recommend journals or other entities archive their content “in more than one place, ideally, in at least three” (DOAJ, n.d.; LOCKSS Program, n.d.). However, preservation networks do not necessarily address two concerns we flagged during the course of our own research into options for preserving Iter journals and other HSS materials: discoverability and possible costs.

For starters, most preservation networks do not address the issue of discoverability, since they actually operate as “dark archives” (i.e., archives that cannot be accessed by the public and are kept primarily as a backup solution in case the original archive becomes inaccessible or other “trigger events” occur).⁵ One might reasonably ask, then, How does one address preservation while also improving discoverability? This is the question we asked when working with Iter, since our intention was not simply to preserve its journals but to do so in a way that would increase the visibility of, and prospective new readers’ engagement with, the journals.

Moreover, there is the possible issue of costs associated with preservation activities. Some preservation networks (PNs) have membership requirements and fees, which can be a major obstacle for journals. To mitigate this concern, the DOAJ has reduced costs through Project JASPER, launched in 2020 (DOAJ, n.d.). And there are undoubtedly other ways to mitigate the larger problem of PN membership fees or costs associated with preservation efforts and infrastructures.

These possible limitations of PNs take us back to the issue of so-called vanishing or disappearing resources. Several critics rightly observe that the phenomenon of vanishing digital publications remains a very real threat (Lightfoot 2016; Laakso et al. 2021; Eve 2024). Eve (2024) reports, “A significant portion, approximately 28%, of academic journal articles with DOIs appear entirely unpreserved in the archives […], endangering both persistent identifier systems and the chain of verifiable citation that they are meant to underwrite” (17). But this figure is for articles across all disciplines. HSS and open access journals are particularly at risk of disappearing: in “Open Is Not Forever: A Study of Vanished Open Access Journals,” Mikael Laakso and co-authors remark that “social sciences and humanities (SSH) journals represent the largest share of vanished journals in our sample (52.3%)” (2021, 1106). Compounding the problem, “OA journals are enrolled in preservation schemes at an alarmingly low rate, with 4 in 10 journals indexed in the DOAJ reporting enrollment in at least one preservation or archiving scheme” (Laakso et al. 2021, 1100). Taken together, these numbers underscore that open access HSS research is disproportionately at risk of disappearance, occupying an uncomfortable position in a Venn diagram of academic journal vulnerability. The question of why HSS and open access journals are at particular risk of disappearance deserves further study, including in our own national context, but some of the likeliest explanations—such as the lack of significant, stable, or long-term funding for HSS, which we discuss below—are so well documented that they have almost become truisms. More helpfully, perhaps, one might also point to the technical expertise and time commitments involved in responsible, ethical, community-engaged preservation and archiving of digital research.

Journals hosted or published primarily through PKP’s Open Journal Systems (OJS) software and other platforms that facilitate archiving through PNs are not immune, either—which is relevant in the case of our own project, since Iter’s journals are hosted through the University of Toronto’s Journal Production Services (JPS), an OJS instance.⁶ Bronwen Sprout and Mark Jordan observe that “the [Global LOCKSS Network, or GLN] preserves content from around 200 OJS titles (out of approximately 10,000)”; this alarmingly low preservation rate for OJS journals is due to the GLN’s prioritization of other content and due to the small size (and presumably limited resources) of many OJS journals (2018, 247–48). Even the PKP PN, which was created to fill this gap for OJS journals, does not provide a foolproof solution; it is up to individual OJS journals to enable an optional LOCKSS plugin to have content backed up in this way.

The fact that individual OJS journals must enable a plugin to be enrolled in a preservation scheme is a small example, but it helps prove a larger point made by many critics: that “there is no consensus over who should be responsible for archiving scholarship in the digital age” (Rick Anderson, quoted in Eve 2024, 4). This lack of consensus has contributed significantly to the problem of vanishing open access resources. Elaborating on this dilemma, Laakso et al. (2021) situate it in relationship to other issues affecting open access journals:

Although the necessary infrastructure exists, at least to some extent, questions as to what content to preserve and who should be responsible for its preservation remain unresolved. Current practices for selecting content for preservation can disadvantage OA content since aspects such as journal impact factors or the invested cost for content acquisition often drive such decisions. Especially in the case of small and independent OA journals, which face financial and technical barriers to preservation arrangements, it seems that the opposite approach for content selection is needed—one that also includes the most vulnerable journals instead of prioritizing prestige. Indeed, preserving the “long tail” of scholarly literature might be one of the most pressing challenges the scholarly community is facing. (Laakso et al. 2021, 1102)

Yet this “most pressing” concern is amplified further by threats to the long-term survival of preservation networks themselves. Rather ironically, even preservation networks fail or are in danger of shutting down (DOAJ 2018; Laakso et al. 2021, 1109). Clearly, more must be done to preserve not only those resources most at risk of vanishing but also the networks and infrastructures designed to preserve them.

The Need to Improve the Quantity and Discoverability of Open Access HSS Research

Our project also aims to make more open access HSS research discoverable online, specifically through a centralized subject repository. Several popular, discipline-agnostic digital commons host a significant amount of HSS research as well as research from other disciplines, but, as the example of Academia.edu attests, they may do so without regard for robust or standardized metadata (Dingemanse 2016). Many digital research commons of this kind are also “walled gardens”: closed, for-profit spaces that exploit user data and operate in a manner largely antithetical to the spirit of the Open Access Movement, in spite of official statements to the contrary.⁷ In these cases, discoverability of HSS research, rather than quantity, is sometimes the real issue, since their bespoke metadata is not concerned with interoperability or with conformance to other aspects of the FAIR (Findable, Accessible, Interoperable, Reusable) guidelines for data management.

Of course, a considerable amount of HSS research is also hosted in institutional repositories (IRs), which serve vital preservation, archival, and research-sharing functions. However, some of these distributed IRs are closed to those without institutional affiliation, and, as Brian Clark notes, IRs have been plagued by other issues, including a lack of faculty and student buy-in. Clark writes, “reasons [for this lack of buy-in] range from not knowing the IR exists in the first place, not understanding its purpose or benefits, concerns over copyright and plagiarism, or workflows that require too much time and input from faculty” (2023, 743). For the project outlined in this article, we chose to adopt a more active deposit model to populate the HSS Commons, which is built around a subject repository rather than an institutional one. That is, the deposit model we used is not passive, but active—and also community-engaged, following what Clark calls “mediated deposit models” (743). Initially, though, we had tried other active or hybrid approaches, including inviting researchers in our larger Implementing New Knowledge Environments (INKE) Partnership to add their publications to the HSS Commons repository, offering hands-on assistance from graduate research assistants who were part of the HSS Commons team. While this approach worked, insofar as it helped us populate the site with a modest number of new publications, it also took a considerable amount of time and effort—largely because it involved manually copying publication metadata from CVs, personal websites, or existing repositories into corresponding form fields in the front end of the HSS Commons. For these reasons, we began to search for an alternate, semi-automated active deposit model, knowing that well-populated repositories tend to encourage further contributions and thus increase the repository’s visibility and potentially prestige (Schlangen 2015; Hwang et al. 2020; Butterfield et al. 2022; Clark 2023).

The Need for Free, Open, and Sometimes DIY Solutions

The third and final issue we identified is that, when funding for publication or long-term preservation is not readily accessible, free, open, and sometimes makeshift or idiosyncratic solutions may be required. This is a known issue facing small journals and libraries (Laakso et al. 2021, 1108), and it constrains the work of publishers and researchers in many parts of the world—even as it drives others to create innovative, equitable, and community-minded publication tools or workflows.

The need for free, open, or DIY solutions as a response to such constraints, or as a response to the issues detailed above, is not just—or not necessarily—a Global North versus Global South issue; it is felt across national, linguistic, and disciplinary boundaries. As Laakso et al. note, for instance, there is a “disproportionately low share of vanished journals from Latin America—where the principles of community and OA are embedded into academic culture” (2021, 1108). In Canada, as elsewhere, funding within and for the humanities and social sciences has been an ongoing concern: despite the promise of open access publishing, barriers to or misalignments in open access policy (CRKN 2025) and cuts to scholarly publishing initiatives or the academic libraries and funding agencies that support such initiatives have been a constant threat, creating or exacerbating existing issues related to the production, accessibility, and preservation of knowledge (Canadian Federation for the Humanities and Social Sciences 2023; CRKN, n.d.).

Still, open access is widely supported across academic disciplines in Canada, and many scholarly communities are working to champion infrastructures and institutional pathways to support open access publication. As the Canadian Research Knowledge Network (CRKN) notes, “Canada has a strong diamond open access ecosystem, particularly in the humanities and social sciences” (2025). In a similarly hopeful vein, Simon van Bellen and Lucía Céspedes observe that, of the 944 active Canadian scholarly journals they identified, only 6% are controlled by “the six major commercial publishers globally, which include RELX-Elsevier, Springer Nature, Wiley, MDPI, Taylor & Francis and Sage Publishing” (2024, 7). While the tendency towards non-commercial publishers seems to bode well for open access publishing in Canada, there are other elements of Canadian academic culture and publishing undermining open access’s widespread adoption—and in some cases these elements disincentivize open access publishing and the long-term preservation of scholarly data. In terms of journals, the CRKN (citing van Bellen and Céspedes) remarks that “the vast majority of Canadian journals operate independently of commercial publishers and have already embraced Diamond OA, that is OA without author-facing fees” (2025). However, even according to the calculations of van Bellen and Céspedes, only 61% of “national” journals are Diamond OA, and these figures do not account for the many “international” journals that—as their summary of Larivière and Warren (2019) acknowledges—continue to hold significant appeal to “Canadian scholars, especially in the SSH” (van Bellen and Céspedes 2024, 2). Moreover, van Bellen and Céspedes nod to the well-documented fragility and instability of not-for-profit journals: they identify “312 periodicals having ceased publication, equivalent to 25% of the total number of journals having been active during this period” (12), noting also that “achieving financial sustainability remains a challenge for many journals, particularly Diamond OA ones” (16). To add to these concerns, policy issues complicate and, in some instances, obstruct the advance of Canadian open access publishing—to say nothing of shortcomings in Canadian funding programs. Van Bellen and Céspedes point out, for example, that “a subscription journal currently stands a better chance of being funded than a Diamond OA journal” (16). Despite Canada’s strengths in open access publishing, then, the fact remains that a successful open access model needs support at multiple levels—for example, culture, society, policy, funding, infrastructure—to remain viable as well as resistant to threats of obsolescence and disappearance once it has been implemented.

For the United States and its trading partners, various threats to academic publishing and activity have recently become a painful reality for many in HSS and beyond, as the recent tariff announcements and devastating cuts to the National Endowment for the Humanities both illustrate (Aktorosian 2025; Palmer 2025; Wulf 2025). Moreover, wars, climate disasters, and other violently disruptive events have led to creative solutions to preservation and archiving—such as those employed by Saving Ukrainian Cultural Heritage Online (SUCHO; https://www.sucho.org) and detailed in its members’ DIY Web Archiving zine (Dombrowski et al., n.d.)—demonstrating how DIY preservation methods might be adapted as part of a preemptive response to any number of crises.

Beyond these more extreme and exemplary examples, though, more mundane factors such as interoperability also impact preservation efforts in vital ways at local, national, or even international levels. For example, Clara Turp et al. observe that in Canada, “There is a need for repositories to move beyond stand-alone systems by promoting interoperability. Systems are slowly evolving away from the stand-alone model, whether through metadata aggregation services, discovery layers, or linked open data-based approaches” (2020, 1). Interoperability and the need for robust publication metadata linking distributed repositories were relevant in the case of our work with Iter too—work carried out as a research experiment, in which innovation was required less as a response to immediate geopolitical crises than to structural and infrastructural constraints.

But what does “innovation” mean in this context? As the next section elucidates, the workflow we developed involves the use of custom scripts that we wrote, the customization of open source software to extend its out-of-the-box functionality, and also scraping metadata rather than accessing it via application programming interfaces (APIs) due to technical as well as other limitations (such as those discussed below). It is innovative and DIY in part, though, because our team is not trained in preservation work of this kind and therefore was, at the outset, largely ignorant of existing workflows or methods used by other repositories; as it turns out, we were re-inventing a wheel, using whatever tools, materials, and digital skills were at hand. Still, while we later discovered that there are certainly more powerful and efficient ways of handling data migration and mapping than scraping, and that projects such as the SSH Open Marketplace (2025) have used more sophisticated methods to achieve similar metadata-ingestion ends, these often require much greater investment in personnel, infrastructure, and software. By comparison, our method—leveraging OpenRefine’s easy-to-use graphical user interface rather than requiring more advanced programmatic methods—may be useful as a customizable DIY solution that nevertheless provides remarkable data cleaning affordances primarily through widely accessible open source software.

1. Creating a Target Resource Spreadsheet

Working with Iter and others, we started by identifying all resources (at the journal issue or monograph level) that we wanted to migrate to the HSS Commons repository. We organized this information in a spreadsheet with journal titles, issue numbers or book titles, the URL for each issue on the JPS website, and the publication’s Creative Commons license type in separate columns (figure 1). For project management purposes, we also included columns to record the ETCL team member assigned to the migration work for a particular resource as well as to indicate when that resource had been successfully migrated.

Figure 1.

Screenshot of target resource spreadsheet

2. Creating an OpenRefine Project

For the next steps in our workflow, we used OpenRefine, “a powerful free, open source tool for working with messy data: cleaning it; transforming it from one format into another; and extending it with web services and external data.” Of note for anyone interested in these kinds of tasks, OpenRefine has active user and development communities as well as excellent official documentation.

In OpenRefine, we first created a new project based on the journal issue URLs from the spreadsheet. We started with URLs, since these were needed to populate the project with metadata obtained through scraping. Since some journals had published material with multiple Creative Commons license types, and since we received journal- and issue-specific notes from Iter, we also decided to create separate OpenRefine projects, grouped according to issues within a journal that used the same Creative Commons license.

3. Scraping Relevant Metadata

OpenRefine does not function exclusively as a scraper for external websites, and many users adopt it primarily for other purposes, but we found that—thanks to its built-in fetching function and additional support for HTML parsing—it can be quite an effective tool for harvesting open data, even though APIs and other programmatic methods for collecting metadata were theoretically available to us (including within OpenRefine, which also works with APIs). Despite the availability of an “OJS Scraper for R” package (gastonbecerra, n.d.), for instance, we preferred OpenRefine so that we could scrape, clean, and export the metadata within a single graphical user interface. This OpenRefine- centric process also allowed us to easily customize the scraping to accommodate any differences specific to JPS or other OJS instances.

Each of the journal issues we migrated from JPS has its own landing page on that journal’s JPS website, with additional landing pages for individual publications within each issue. Crucially for our purposes, then, both of these types of landing pages contain predictably structured HTML and metadata that can be scraped in an automated fashion. The predictable structure of the JPS system allowed OpenRefine to use our list of journal issue URLs (from step 2) to scrape the landing pages for each issue and then scrape the landing pages for each of the issues’ individual publications to extract metadata.

Within OpenRefine, we accomplished this by scraping the source code for each journal issue’s landing page by using the built-in “Add column by fetching URLs … ” function. Next, we used a General Refine Expression Language (GREL) statement to parse the landing pages’ HTML and identify all individual publications within each journal issue so that their metadata could be scraped and used to populate new columns using a series of targeted GREL statements.

The success of this process relies on the predictable structure of an individual publication’s metadata within JPS. For example, if one examines the source code for Eva Pelayo Sañudo’s “Multicultural Little Italy: A Literary Comparison of Canadian and US Urban Enclaves” (published in volume 34 of the journal Italian Canadiana and accessible at https://jps.library.utoronto.ca/index.php/italiancan/article/view/37450), the metadata information contained in the HTML and scraped by OpenRefine is encoded in hidden meta elements such as the following:

<mml:meta name=“citation_journal_title” content=“Italian Canadiana”/>

<mml:meta name=“citation_journal_abbrev” content=“Italcan”/>

<mml:meta name=“citation_issn” content=“2564–2340”/>

<mml:meta name=“citation_author” content=“Eva Pelayo Sañudo”/>

<mml:meta name=“citation_author_institution” content=“University of Oviedo, Spain”/>

<mml:meta name=“citation_title” content=“Multicultural Little Italy: A Literary Comparison of Canadian and US Urban Enclaves”/>

Again, because these HTML elements represent the publication metadata in a predictable, structured, machine-readable way, we were able to use OpenRefine to parse the HTML, extract the relevant metadata, and create separate columns for each of the metadata values we needed for our project. To create a new column and populate it with the title for each publication, for example, we used the following GREL snippet:

value.parseHtml().select(“meta[name=citation_title]”)[0].htmlAttr

(“content”)

However, once we had extracted all of the metadata we required, we then needed to clean and transform the data to ensure consistency and prepare it to be exported to the HSS Commons. This step was the most time-consuming part of the entire process, and the nature of the data cleaning and transformation work was slightly different across the Iter journal datasets. Typically, though, this data cleaning and transformation included using filtering/faceting techniques,⁸ GREL, or—in some cases, where the implementation of GREL or other programmatic solutions were prohibitively time-consuming—manual correction to perform the following kinds of tasks:

• Removing whitespaces from the beginning or end of various text-based fields (e.g., authors, publication titles, abstracts);

• Removing instances of multiple spaces within text strings;

• Excluding tables of contents, front matter, back matter, journal covers, and other journal materials we did not intend to republish;

• Replacing “null” values (e.g., adding a journal title to a publication’s list of subject tags, to increase a publication’s discoverability and prevent issues in the case of required metadata information in the HSS Commons repository);

• Splitting comma- or semicolon-separated author lists, reversing the given name(s) and last name(s), and creating a semicolon-separated list to be used in citations;

• Prepending the titles of book reviews with “Review of” to distinguish reviews from other publications and prevent confusion regarding authorship;

• Creating a PDF URL column, by identifying JPS galley links, to be used in a later step to batch download all publications; and

• Creating citations of the original JPS publications so that these could be displayed to users in the HSS Commons with an accompanying note acknowledging their provenance and linking back to the journal (to promote the journal, increase discoverability, and provide clear information about the publication of record).

Initially, we performed all data transformations one by one during the raw data cleanup phase through an iterative process, as we learned more about GREL and learned how best to scrape, clean, and transform metadata. Later, though, we created a transformation template, which we could apply all at once using OpenRefine’s extremely useful “Apply Operation History” feature (figure 2). This feature allowed us to perform all of the following actions, which OpenRefine encodes in JavaScript Object Notation (JSON) statements, in a single step: visit and scrape each issue’s landing page, visit and scrape any publications’ landing pages linked from main issue pages, scrape each publication’s metadata, construct a list of publication PDF paths, clean all of the metadata, and create new metadata columns to assist with the export process.

Figure 2.

OpenRefine’s “Apply Operation History” feature

4. Publishing Records on HSS Commons

Publishing the metadata from step 3 involved (1) creating an XML file containing information in a format compatible with HSS Commons’ custom schema, (2) downloading the PDF files for each of the records, and (3) uploading and publishing both record metadata and PDFs in the HSS Commons.

To create an XML file containing the metadata for each of the publications, we harnessed OpenRefine’s powerful “Templating export” feature (figure 3).⁹ By doing this, we were able to export each record as part of an XML file that conformed to the Commons’ required schema for batch ingestions. This process involved mapping metadata fields onto the XML schema as well as adding in a custom notes field to be displayed on the HSS Commons. The notes field acknowledges Iter and the journals in question for permission to carry out this work and indicates the provenance and permissions for each publication.

Figure 3.

OpenRefine’s “Templating export” feature

Once the XML file was ready, we used a custom Python script, written in XML and GREL by our developer, Archie To, to batch download all of the publication PDFs and insert their file paths into a second version of the XML file.

To upload and publish the records on HSS Commons, we first created an HSS Commons “project” for each journal within the larger Iter Community “group” and logged in with the Iter user account (figure 4). We then uploaded the PDFs for that journal to its project-level repository. Next, we batch imported the metadata by uploading the XML file to the administrative back-end interface of the HSS Commons.

Figure 4.

An Iter journal’s “project” landing page in the HSS Commons

To improve this process in the back end, Archie customized the PHP code governing HUBzero’s built-in “Batch Create” feature to automatically publish the records in the HSS Commons repository—a creative technical intervention that saved us from having to manually confirm and publish each of the more than 6,000 records, since the default batch-ingestion function saves items with a status of “Draft” rather than “Published” (figure 5). Additionally, this batch creation and publication process was unique insofar as—according to Clark (2023, 748)—many metadata-ingestion workflows of this kind do not also involve the harvesting and republication of publications’ primary files. When records had been successfully created in the HSS Commons, they became publicly accessible immediately via the site’s repository (figure 6).

Figure 5.

The customized “Batch Create” function in the HSS Commons back end

Figure 6.

A public-facing publication landing page in the HSS Commons

Limitations and Challenges

Our workflow was deliberately designed to use open source tools or software, to make it more accessible and potentially useful to others. Given our team’s own inexperience at the outset of this project, we are optimistic, too, that it can be applied in a variety of open access publishing and open infrastructure contexts. That said, we did document a number of limitations and roadblocks that, in this same spirit of openness, we thought it worthwhile to pass on to others who may be interested in this kind of preservation work.

First, our workflow was only designed to scrape articles from OJS-based journals—though it is, admittedly, a very widely used platform. But, more to the point, our workflow is tailored for ingestion into only one platform, the HSS Commons, and the batch-publication function on our site uses a somewhat idiosyncratic XML schema. Adapting the custom export template for other schemas or file types would require anyone not already familiar with XML to learn some fundamentals (or perhaps use an AI-assisted programming tool).

Second, any customization of our workflow—particularly the web scraping and data cleaning steps—requires some knowledge of OpenRefine, GREL, GREL- supported regex, and the OpenRefine variables used both in GREL expressions and in the templating exporter. Data transformations can be performed using Python/Jython or Clojure instead of GREL, but GREL is the default option. And while excellent OpenRefine tutorials such as the Programming Historian’s “Fetching and Parsing Data from the Web with OpenRefine” (Williamson 2022) are freely available, certain cleaning and transformation scenarios may present a challenge to those with minimal GREL or programming experience. During our work with Iter, for example, we struggled to transform a concatenated author list (e.g., “Graham Jensen; Sajib Ghosh”) into a list in which the authors’ names were inverted (e.g., “Jensen, Graham; Ghosh, Sajib”). After some research, we arrived at the following interim solution using GREL:

forEach(value.split(“; “), v,

v.match(/(.*)\s(.*)/).reverse().join(“, “)).join(“; “)

However, this transformation was not able to properly handle names with suffixes, such as “Jr.” or “III.” When doing similar preservation work later, for other open access journals, and after many more failed attempts to transform names such as “Martin Luther King Jr.” to “King, Martin Luther, Jr.”—as one might expect to see in an academic paper’s references section—we finally decided to return to the problem with the help of Chat GPT-4o. It came up with a solution that, while extremely effective in dealing with the suffixes “Jr.,” “Sr.,” “Junior,” “Senior,” and Roman numerals II through XX, we may never have been able to produce on our own:

forEach(value.split(“; “), v,

 if(

 or(

 v.split(““)[v.split(““).length() - 1] == “Jr”,

 v.split(““)[v.split(““).length() - 1] == “Sr”,

 v.split(““)[v.split(““).length() - 1] == “Junior”,

 v.split(““)[v.split(““).length() - 1] == “Senior”,

 v.split(““)[v.split(““).length() - 1] == “II”,

 v.split(““)[v.split(““).length() - 1] == “III”,

 v.split(““)[v.split(““).length() - 1] == “IV”,

 v.split(““)[v.split(““).length() - 1] == “V”,

 v.split(““)[v.split(““).length() - 1] == “VI”,

 v.split(““)[v.split(““).length() - 1] == “VII”,

 v.split(““)[v.split(““).length() - 1] == “VIII”,

 v.split(““)[v.split(““).length() - 1] == “IX”,

 v.split(““)[v.split(““).length() - 1] == “X”,

 v.split(““)[v.split(““).length() - 1] == “XI”,

 v.split(““)[v.split(““).length() - 1] == “XII”,

 v.split(““)[v.split(““).length() - 1] == “XIII”,

 v.split(““)[v.split(““).length() - 1] == “XIV”,

 v.split(““)[v.split(““).length() - 1] == “XV”,

 v.split(““)[v.split(““).length() - 1] == “XVI”,

 v.split(““)[v.split(““).length() - 1] == “XVII”,

 v.split(““)[v.split(““).length() - 1] == “XVIII”,

 v.split(““)[v.split(““).length() - 1] == “XIX”,

 v.split(““)[v.split(““).length() - 1] == “XX”

 ),

 v.split(““)[v.split(““).length() - 2] + “, “+

 join(slice(v.split(““), 0, v.split(““).length() - 2), ““) + “, “+

 v.split(““)[v.split(““).length() - 1],

 v.split(““)[v.split(““).length() - 1] + “, “+

 join(slice(v.split(““), 0, v.split(““).length() - 1), ““)

 )

).join(“; “)

This seemingly niche example demonstrates not only the potentially surprising complexity of any project dealing with metadata transformation challenges but also the power of GREL and AI-assisted programming tools to come up with DIY solutions to them. Although our experiment also sparked discussion about the ethics and financial, environmental, or other costs of generative AI (which are well beyond the scope of this article), we know that many information professionals and developers carrying out preservation work of this kind are likely much more well-equipped than we were to deal with such roadblocks without the use of third-party services or tools.

Similarly, we are fully aware that, in some cases, collecting metadata via API instead of scraping may be more efficient, more accurate, or—particularly in the case of much larger scraping projects—the option officially recommended by repositories and their system administrators.

Third, and finally, because we were dealing with resources in English, we did not consider the challenges posed by multilingualism vis-à-vis data repositories. Turp et al. (2020) have summarized some of the challenges involved when attempting to map metadata from one repository onto another, or to reconcile controlled vocabularies between different languages. Drawing on their work on the Canadian Federated Research Data Repository (FRDR), they point out that “[u]sers searching in French, for example, might be interested in searching across all the datasets, regardless of which language is selected in the interface” (Turp et al. 2020, 6). In our case, such considerations were beyond the scope of our project, and we have not yet created bilingual—let alone multilingual—crosswalks.

Further Theoretical Considerations and Next Steps

Returning to the theoretical contexts for this project, I would suggest that, despite the limitations of our approach, we accomplished what we set out to do—and we did so in line with our lab’s emphasis on openness and community. Still, the work of preserving HSS as well as open access research must go on. As Laakso et al. note, “[T]his issue [of vanishing OA journals] should be considered as an ongoing process that will continue unless we fully commit to preserving the scholarly record. Successfully solving this issue will require the active involvement of the scholarly community as a whole and solutions as diverse as scholarly research itself. While the current system places the responsibility for preservation mainly on OA journals alone, other actors (e.g., funders, academic institutions, authors) play a vital role in facilitating this process and in mitigating losses” (2021, 1108).

Our efforts to help Iter preserve its journals focused around JPS, an OJS instance that provides optional access to the PKP Preservation Network via a plugin. But our workflow could be adapted quite easily, in the future, to migrate other journals not enrolled in preservation schemes and therefore at even higher risk of disappearance. Tom Cramer et al. echo some of these sentiments; they concur that “digital preservation can never be a solved problem. It is work that does not finish” (2023, 312). Maybe the point, then, is that we are trying, and we are using open infrastructure to creatively address a shared problem that is everyone’s and no one’s responsibility.

Moving forward, we intend to migrate other journals to the HSS Commons as we improve both our workflow and the repository’s compliance with reusability, interoperability, and other FAIR (Findable, Accessible, Interoperable, and Reusable) principles. Indeed, reproducibility is one of the key benefits of an OpenRefine-centered workflow, as Turp et al. (2020, 4) have also noted. Taking advantage of the “Apply Operation History” feature described above, we have already reused and refined our Iter workflow to preserve other open access journals such as Canadian Food Studies.

Other next steps might be to create a similar workflow that uses OJS or OpenAlex API to retrieve open access journal metadata from within the OpenRefine environment; use the Python-based Sickle (n.d.) library to retrieve metadata from Open Archives Initiative sources; and experiment with sunilnatraj’s (n.d.) AI/LLM extension for OpenRefine for data transformation and analysis. Once we have improved our repository’s support for publication language specifications, as planned, we are also interested in actively preserving more non-English HSS research to improve the diversity of research materials within the repository and its usefulness as a multilingual resource, in line with our ongoing translation of the HSS Commons’ interface into languages other than those already available on the site (English, French, Spanish, Bangla, Portuguese, and Ukrainian).

In addition to these ideas, we have already started to act on our initial research into vanishing HSS and open access scholarship by using resources such as the Keepers Registry (https://keepers.issn.org/keepers) to identify at-risk journals that could be backed up in the HSS Commons repository and showcased using the site’s group, project, and collection features. As Sprout and Jordan explain, the Keepers Registry serves as a terrific resource for this kind of work by “providing information about which journals are preserved by which archiving services and highlighting those journals for which no arrangements exist” (2018, 252). As we consider what to preserve in the HSS Commons, we would like to leverage resources of this kind to guide our efforts so that we can continue to play a small role in preserving open, HSS scholarship for future generations while also building out our repository in ways that serve the present and always-changing needs of our community.

Finally, as we carry out this ongoing work, we acknowledge the political and critical dimensions of preservation and archiving (which cannot be summarized adequately here but are, nevertheless, worth mentioning as one of the considerations informing this project).¹⁰ As Samantha Cross remarks, summarizing Randall Jimerson, the assumption that archiving is a neutral act “deters active archiving and reduces archivists to passive recipients. In reality, archivists have the potential, if not the responsibility, to act and explore other options of collecting and serving their communities” (2017, 1). For us, our project was a deliberate intervention intended to promote open HSS resources originally published by our Canadian research partners. Inevitably, though, it also involved making choices that privileged certain journals, disciplines, or even languages over others, thus reflecting the academic interests and biases of our own research team as well as those of the larger national and international research communities in which we are embedded. When engaging in this preservation work, we have therefore tried to harness its potential for positive transformation of HSS communities and infrastructures; yet we acknowledge the responsibility that accompanies preserving when one interprets it, as we do, as a political act. To engage in what Cross calls “active, deliberate archiving” (2017, 2), one needs to grapple with myriad questions: What gets preserved? Is it open (or should it be)? Which research areas or researchers does this process (de)prioritize? How might it subtly re-direct conversations, emerging methodologies, or disciplinary fields? And so on. In these ways and others besides, we regard even our own modest efforts to republish and increase the visibility of resources through an open digital knowledge commons as acts that inevitably shape the academic landscape—hopefully to its betterment and to the betterment of those who occupy it.

Again, the kind of exploratory, DIY methods outlined in this article are meant only as a stopgap or complementary solution to the kind of large-scale re-evaluation called for by Laakso et al. (2021); Cramer et al. (2023); and other scholars, librarians, and tech experts already active in this area. Nevertheless, while that re-evaluation unfolds—and as the complex and sometimes arduous associated processes involving changes to policy, preservation initiatives, and related infrastructures gradually evolve—we hope that our patchwork project impresses the importance of preservation, including but not necessarily using the repository, software, and methods foregrounded here.