In June 2008, The Rossetti Archive “closed,” although the site remains accessible. What can a “closed” site reveal to scholars today? Much. As digital scholarship gains purchase in the field of art history, we should learn from pioneering projects such as The Rossetti Archive.

Edited by literary scholar Jerome McGann, the archive began in 1993 at the moment of public access to the worldwide web and when McGann’s home institution, the University of Virginia, founded the Institute for Advanced Technology in the Humanities. The project aimed to make the work of Victorian poet and artist Dante Gabriel Rossetti accessible via the web. At its core, it is a digital critical edition, although McGann argues it is much more.

The archive today is composed of five major sections: “Home,” an introductory overview; “About the Archive,” which explains the stages by which the project was completed from 2000 through 2008; “Exhibits & Objects,” collections of Rossetti materials brought together under such rubrics as Pictures, Material Design, and Double Works (reflecting Rossetti’s practice of producing pictorial and textual depictions of the same subject); “Search Engine,” the interface that allows structured searching; “Bibliography”; and “Nines,” the acronym for the Networked Infrastructure for Nineteenth-Century Electronic Scholarship, a scholarly organization that aggregates peer-reviewed digital projects dedicated to nineteenth-century text-based topics.

McGann chose the term archive, as opposed to critical edition, to title his project. He argues that an edition represents a closed environment whereas The Rossetti Archive, as a hypermedia environment, “has been built so that its contents and its webwork of relations (both internal and external) can be indefinitely expanded and developed” (McGann, Radiant Textuality: Literature after the World Wide Web [New York: Palgrave, 2001], 69). All the documents in the archive have been encoded, work that exceeds that of a print-publication index and allows for connecting between texts as well as images. As information-technology specialists Daniel Apollon and Claire Bélise observe, “In the online Rossetti Archive, priority is given to the intertextual and trans-textual dimensions. It is precisely in this domain that digital tools make it possible to innovate more radically, not only allowing one to escape from the material constraints of print technology and the boundaries of the book page but also opening new horizons for more eclectic projects that offer the possibility to ‘tag texts in several dimensions’” (Apollon and Bélise, “The Digital Fate of the Critical Apparatus,” in Digital Critical Editions, ed. Apollon, Bélise, and Philippe Régnier [Urbana: University of Illinois Press, 2014], 92).

Tagging, markup, or annotation of a document—allowing for categorization, documentation, and description—encourages enhanced analysis and searching (Pichler and Bruvik, “Digital Critical Editing: Separating Encoding from Presentation,” in Digital Critical Editions, 179–99). Complex decisions inform this work, and will guide how the site performs; therefore, a high degree of transparency about this markup work is desirable. The decision of The Rossetti Archive not to include detailed information about their methodological approach, beyond references to XML, HML, and Cascading Stylesheets found in the “About” section, is thus unfortunate. One must turn to print sources to discover choices made.

At the outset of the archive, McGann explains, “we arrived at a double approach: first, to design a structure of SGML markup tags for the physical features of all the types of documents contained in The Rossetti Archive (textual as well as pictorial); and second, to develop an image tool that permits one to attach anchors to specific features of digitized images” (McGann, Radiant Textuality, 69). SGML (Standard Generalized Markup Language) is a system or syntax for marking up documents, analogous to XML (eXtensible Markup Language) and MECS (Multi-Element Code System). In the 1990s, as these markup languages were in formation, TEI (Text Encoding Initiative), a subset of SGML, emerged as the most commonly used language for marking up texts, particularly for the humanities. McGann and his team chose not to use TEI because the essential problem of the archive—“overlapping structures of literary works and their graphic design features”—was not well suited for TEI. Instead they created their own tool in SGML (McGann, Radiant Textuality, 69). That choice, McGann makes clear, was fateful as the team soon confronted a limitation of SGML that does not easily allow a single entity to be characterized as two distinct types—one cannot simultaneously identify a document as both a page and a poem, for example (McGann, Radiant Textuality, 90, 97). This limitation made it difficult to link illustrations appropriately to the texts.

Nonetheless, the decision to code documents in SGML had payoffs. It ensured that the presentation layer (seen via web browsers) would be separate from the source material, which is, in essence, a marked-up transcription of a document (Pichler and Bruvik, “Digital Critical Editing,” 179). The Rossetti source files, begun in SGML, could eventually be converted to XML, another markup syntax, to facilitate implementation of a new search engine.

The project team approached images separately from texts. They faced a wide range of image types, including those directly affiliated with a text as published illustrations, those associated with a text through the double-work process, and those produced as independent works of art. The team was soon frustrated by the process of permissions and copyright and so focused initially on historic reproductions (photographs of Rossetti’s works made during his lifetime or shortly thereafter) (McGann, Radiant Textuality, 92). This decision threw fascinating attention on an arguably little studied aspect of Rossetti’s visual corpus. Today, the archive features high-quality digital reproductions of many works, but the presentation layer, which serves up images in a single-file procession, makes comparative analysis difficult. However, the project team did devise useful ways to help users navigate the images through three primary filters: alphabetical, chronological, and an innovative timeline.

The site currently asserts the highly recommended Creative Commons Attribution license (CC-BY), whereby the original author retains attribution, and, in addition, the NC designation, meaning material cannot be used for commercial purposes. Appropriate attribution by users could have been further encouraged by creating a built-in citation tool.

Another challenge The Rossetti Archive team faced was tagging images to be as searchable as texts and affiliated with texts in the linked method envisioned. Our field still lacks a viable iconological approach to image searching, although image recognition, as a subset of computer vision, is improving. While languages and syntaxes have emerged to support the markup of texts, no robust equivalent has emerged for images (Julia Thomas, “Getting the Picture: Word and Image in the Digital Archive,C European Journal of English Studies 11, no. 2 [August 2007]: 196). The Rossetti team originally envisioned a tool, Inote, which would allow their editors to attach textual annotations to designated areas of an image. (The Inote tool has since improved and Martin Holmes has developed an Image Markup Tool for TEI.) But, as McGann confesses, the team did not bring the same attention to marking up images as it brought to texts, and initially texts and images were not fully integrated within the SGML-marked database (McGann, Radiant Textuality, 95–96). The solution devised by The Rossetti Archive was to utilize the critical commentary that accompanies each image as an integration method. Searches for images are actually searches through the textual descriptions affiliated with images, which can constrain results. For example, using “palm” or “dove” in the free-form search of image records did not turn up Rossetti’s painting The Girlhood of Mary Virgin, even though they feature prominently in the painting, because palm is pluralized in the image record and the word dove does not appear. It also appears that a standardized or controlled vocabulary, providing nomenclature authorities, was not employed in the image records.

McGann, in his published accounts, is transparent about the project’s failures and, in particular, the limitations of its handling of images. The Rossetti Archive was a project assembled by primarily literary scholars and its impact, as measured by ways it spurred the development of other projects—such as Juxta, an open-source tool that allows for collation and comparison of texts, Collex, an online exhibits and collections builder, and the NINES scholarly community—has been primarily within the disciplines of literary and Victorian studies. But, as art history increasingly embraces the digital environment as the appropriate platform for projects such as catalogue raisonnés, it is vital that we harness lessons learned from adjacent disciplines.

Perhaps the most important lesson learned from The Rossetti Archive regards when to draw a digital project to a close. As McGann makes clear, creating the archive produced an initially unanticipated parallel venture: “to use the archive’s process of construction as a laboratory for reflecting on the project itself. That second purpose led inevitably to a regular set of critical inquiries into the basic organizing ideas of the archive and its procedures” (McGann, Radiant Textuality, 141). This pursuit threatens to become Edward Casaubon’s unfinished Key to All Mythologies (from George Eliot’s Middlemarch) and requires considerable discipline to circumscribe and delimit. Indeed, this is the great challenge of all digital projects, no longer confined between covers and stored on a shelf, but instead endlessly extendable and requiring constant upkeep to keep pace with changes in digital technology. How to exit gracefully? By leaving behind an accessible, critically informed, soundly structured, and well-documented intellectual contribution.