Although informal conversations had been taking place for some time, it was in 1999 that Duke University Press began to explore seriously the possibility of developing an online version of The Collected Letters of Thomas and Jane Welsh Carlyle. Experts were consulted; challenges were debated and assessed; and a National Endowment for the Humanities grant was discussed, written, and awarded. With the project nominally funded, a plan for moving forward with the development of “eCarlyle” began to take shape, albeit a plan that was necessarily fraught with technological, textual, and financial indeterminacy.

One of the earliest conclusions reached in the development process was that neither DOS- nor HTML-centric approaches were going to survive the next phase of technological advances for electronic publications. So it was decided that the first twenty-nine volumes of the edition would be encoded in XML (eXtensible Markup Language), conformant with Textual Encoding Initiative (TEI) guidelines.

Although there was no way to anticipate what the final product was going to look like, or in what software platform the resource was going to “live,” these foundational decisions provided a flexible, forward-looking path that allowed the project to proceed. From the outset, this path was directed toward the creation of a fully searchable, user-friendly, and nonproprietary scholarly tool that could eventually be connected with other digital humanities resources.

Early in 2001, in spite of the technical and conceptual difficulties of developing a resource that would rely on technology that did not yet exist, Paolo Mangiafico (Perkins Library, Duke University), David Southern (managing editor of the Collected Letters), and Brent E. Kinser (coordinating editor of eCarlyle) sat in the courtyard of Brightleaf Square in Durham, North Carolina, and tried to anticipate the encoding requirements necessary to transform eCarlyle from ideal into reality.

They began by adapting the general rubrics of TEI and of Model Editions Partnership (MEP), a project led by David Chesnutt at the University of South Carolina, into a specific set of encoding guidelines for eCarlyle. On the basis of this TEI and MEP framework, the eCarlyle team identified and prioritized features of the printed text that would need to be given XML tags so that the resource could be made to display properly online in multiple browsing environments.

The eCarlyle guidelines quickly began to take shape. The format for the appropriate headers was defined, and it was decided that time and resources would be devoted to the tagging of all dates, salutations, closings, and signatures. Most important, the team agreed that it was imperative that every reference to another letter or text in the eCarlyle would be rendered as a hyperlink, which would create a vast web of interconnectivity within the resource.

Because there would be neither time nor resources to tag all features in the text, a wish list for future encoding was also established, which included the tagging of regularized names, ships, places, organizations, and institutions.

Looking back, it now seems clear that the decisions made at Brightleaf that spring represented an important shift in the editing of text. Issues of transmission—what the editor wishes the reader to know—had evolved into issues of functionality—what the editor wishes the user to do. In identifying the features to be tagged, the eCarlyle team was in fact deciding what they would teach the text about itself.

What then seemed of secondary importance now seems paramount, for the concept of textual self-knowledge lies at the center of electronic textual editing. What the text knows about itself defines the scope of its functionality, which defines its ability to invite discovery, an essential quality of electronic resources meant to serve as primary vehicles for scholarly endeavor.

Once the initial guidelines for the encoding, i.e. the XML tagging, of eCarlyle were established, a search was made for an appropriate company to perform the initial digitizing and encoding of the first twenty-nine volumes, and in 2001 DNC Data Systems of Mumbai, India, was awarded a contract. After an iterative process of refining the work of DNC came the task of proofreading both the digital text and its encoding. During the initial proofreading process, Kinser developed a set of editorial guidelines and assembled a team of copyeditors to assist with the immense task of establishing the reliability and the consistency of the text and the XML tags of eCarlyle.

Through a first editorial pass and then a second, the encoded volumes were prepared for migration to a display platform. Unfortunately, the only extant open-source search engine and indexing tool that came close to achieving the necessary level of functionality, Dynaweb, was approaching obsolescence. Fortunately, in January 2003, again with the inestimable help of Mangiafico, the eCarlyle files were mounted into Dynaweb, and they worked.

The prototype version of eCarlyle was an inelegant but mostly functional resource that quickly became a valuable tool for the few editors and devoted readers who were allowed access to it. The links among the letters were useful as predicted, and rudimentary style sheets brought a modicum of aesthetic appearance. However, little was done to improve the prototype, since it was never intended as the permanent home for eCarlyle. The press and the editors remained committed to the future and to the development of a fully functional version of the resource.

After the prototype was created in 2003, the major obstacle to completing the development phase became finding a replacement for Dynaweb. What followed, as the project was forced to wait on technology, was a three-year period of “tweaking” the encoding of the first twenty-nine volumes as others worked to develop technological solutions to the Dynaweb problem. In addition, volumes 30–32 of the print edition were encoded by DNC, then edited and proofread by Kinser. In 2006, finally, technology caught up with the vision. At first it seemed as though the open-source indexing and query tool XTF (eXtensible Textual Framework), designed primarily by Kirk Hastings and Martin Haye for the University of California Digital Library, would be used to create the new digital home for eCarlyle. Although XTF was not ultimately selected, its capabilities announced an exciting new stage in the project’s history.

Simultaneously, two independent XML “content servers”—Mark Logic Corporation and iFactory (RDW Group)—and HighWire Press of Stanford University Libraries presented solutions to the problems associated with the online display and the database management of eCarlyle. In June 2006 the contract to build a new digital home for the newly named Carlyle Letters Online was awarded to HighWire Press, and in July the Web design firm Methodfuel was brought onto the CLO team in order to render into HTML the visual designs conceived by art director Sue Hall and the joint Duke University Press–HighWire team.

During the following year, desired functionalities were defined, conceptual decisions were negotiated, and technological problems were overcome, and on 13 July 2007, at a meeting of Carlyle scholars at a conference in Philadelphia, the coordinating editor of the project announced that the Carlyle Letters Online had entered a beta-testing phase, with the hard launch scheduled for the end of August. After nearly a decade of good service, at the end of July 2016, the HighWire version of the CLO was decommissioned. With the end of the print edition (volume 50) in sight, it had become clear that in the interests of sustainability, editorial control, and connectivity with other resources, particularly in the context of the Victorian Lives and Letters Consortium, the time had come to migrate the CLO to a new platform and a new host at the University of South Carolina Center for Digital Humanities.

From 2016 to 2021, the CLO project continued to evolve and to improve. Team members in Columbia, Durham, and Cullowhee worked together to modify the infrastructure of the site in order to take advantage of a MEAN stack framework, which simultaneously brought the project in line with best practices for design and made it easier to maintain and sustain. New areas of discovery were added such as the Carlyle Photo Albums and the materials located at Duke's Rubenstein Library. Like its predecessors, however, this new version of the CLO did not represent a final product. Work continued to focus on the future and next phases.

Recently, in the context of a new and exciting technology partnership with the Center of Digital Humanities Research at Texas A&M University, yet another round of innovation and improvement has begun. With the migration of the CLO to the new Corpora Dataset Studio, the editors and their new partners at CoDHR have taken a crucial step in securing the long-term extensibility and sustainability of the project. The process of perpetually editing and improving the CLO remains a conflux of past, present, and future in the service of scholarly endeavor.

B.E.K.