Skeletal introductions

§ 12 The most important thing about EB 4.0, as with its predecessor editions, is that it is intended to function as a companion to and tool for the study of Beowulf rather than a summa of contemporary textual, linguistic, and literary thought on the poem. Neither EB 4.0, nor any of its predecessor versions, therefore, should be understood as a replacement for standard scholarly or teaching editions of the poem (e.g., at various times, Fulk, Bjork, and Niles 2008; Klaeber 1936; or, for students, Mitchell and Robinson 1998). While EB 4.0 has some of the standard trappings of such editions (e.g. a critical text, glossary, textual notes), it is also missing several other common features: its short introduction focusses primarily on meta-computational and -bibliographic aspects of the edition itself rather than historical and literary issues, while its substantial textual notes focus solely on codicological and paleographic issues without discussion of the formation of the editorial text.

§ 13 The limited introductory "Index and Guide" is perhaps the most important demonstration of this difference from more traditional critical editions. To begin with, the edition appears to mean "index" in the web-design sense of "default top-level page" (i.e. index.html) rather than the traditional print sense of "list of names and concepts in a work with references to the place in which they occur." As a "Guide" rather than an "Introduction," moreover, the document tends to focus on showing how an argument could be made using the provided tools and materials, rather than providing contextual information and complete arguments about the poem as a cultural and historical object.

§ 14 This does not mean, however, that the edition is agnostic about either arguments or context—the arguments it suggests could be made using the edition tend to line up with those its editor has made elsewhere. They are also often framed in opposition to what we are told (briefly) are the positions of other scholars. But in contrast to a traditional introduction, the edition does not provide readers with a broad context to help them understand these issues or arguments: when context is required, EB 4.0 tends to refer readers to external sources, often by the editor himself, some of which have been collected in an associated repository.

§ 15 An example of how the Guide handles this kind of thing can be seen in its discussion of tools allowing for the virtual reconstruction of readings from the famous palimpsest on f. 179r.

One of the most valuable and perhaps revolutionary aspects of image-based electronic editing of medieval manuscripts is that it permits more accurate evaluation of conjectural restorations of lost text. Before the digital age, conservative editors like Klaeber, who intended to restore damaged readings based on manuscript evidence, printed reconstructions they thought actually fit in the manuscript. Their editions in this respect were meant to conserve the manuscript record. Digital imaging makes it possible for editors to examine the validity of their decisions.

Figure 19. Lost Text in Manuscript

A case in point is the obliterated text between syððan and þ on fol. 179r10. Any attempt at restoration is complicated by the fact that some of the ink traces, as conclusively shown by an overlay in Electronic Beowulf 4.0, come from an offset from the facing fol. 178v. Digital technology allows us to subtract these false leads and arrive at a more plausible restoration.

Figure 20. Conjectural restoration that does not fit

The editor of the most recent print edition restores [bemað] in this space with unusual confidence. A digital restoration of bemað using the scribe's own letters from the same page shows, rather surprisingly, that it cannot fit in the space in the manuscript. The extra-wide letters m and ð take up so much room that the cross-stroke of ð ends up crossing the following þ.

Figure 21. Conjectural restoration that fits

The same method shows that the conjectural restoration [beget], which well suits the context and happily observes everyone's metrical rules, perfectly fits the available space in the manuscript. (Kiernan 2015 "Studying Beowulf")

§ 16 Professional Anglo-Saxonists probably will recognise that this passage is as much an extension of Kiernan's longstanding interest in this page and part of an ongoing scholarly debate as it is an illustration of how some tools in the Electronic Beowulf can be used to understand various parts of the manuscript. In these paragraphs, Kiernan is in fact taking issue with arguments about the correct reading of line 2217 (traditional numbering) made by Klaeber in his early-and-mid twentieth century editions of the poem and, although they are not mentioned by name, Fulk et al. in their recent revision of Klaeber's work (2008). As such, the passage is also part of a larger scholarly dispute—set very much in motion in recent years by Kiernan's own previous work (e.g. in Kiernan [1981] 1996)—about the significance and history of the damage to this page.

§ 17 A novice reader, however, may pick up very little of this: no intellectual context is provided for the discussion (other than the fact that some letters are missing for some reason from this location on this page); and the bibliographic context Kiernan provides is both very informal (no works are actually cited) and includes the name of only one of the editors whose specific emendations are being challenged.

§ 18 When the Guide does provide context, it tends to be by reference to Kiernan's own previous writing, collected in what appears to be an informal post-print repository on a personal university webspace (http://uky.edu/~kiernan/). Here, for example, is his discussion of "Thorkelin A" and "Thorkelin B," two important eighteenth-century transcriptions of the poem:

Thorkelin A, 1787

This first complete transcript of Beowulf was probably made by James Matthews, a member of the British Museum staff, for Grímur Jónsson Thorkelin in 1787. It is now Ny kgl. Saml. 513 in the Royal Library of Denmark. The Royal Library allowed the project to digitize the transcript in 1994 and to include it in Electronic Beowulf. A brief introduction to the use of Thorkelin A is available in the Transcripts & Collations section of this Index & Guide. For evidence that James Matthews was Thorkelin A, see Kiernan, "Part One: Thorkelin's Discovery of Beowulf," The Thorkelin Transcripts of Beowulf (Copenhagen: Rosenkilde and Bagger, 1986), pp. 1-41.

Thorkelin B, ca. 1790

Thorkelin himself made the second transcript, known as Thorkelin B, which is cataloged as Ny kgl. Saml. 512 in the Royal Library of Denmark. The Royal Library allowed the project to digitize the transcript in 1994 and to include it in Electronic Beowulf. A brief introduction to the use of Thorkelin B is available in the Transcripts & Collations section of this Index & Guide. Although Thorkelin dates his own transcript also in 1787, there is negative evidence that he made it no earlier than 1789, and most likely after April 1790, when he was appointed assistant librarian in the Department of Printed Books of the British Museum. For details supporting a 1790 date, see Kiernan, "Part One: Thorkelin's Discovery of Beowulf," and the "Conclusion" to The Thorkelin Transcripts of Beowulf (Copenhagen: Rosenkilde and Bagger, 1986), pp. 1-41, 151-55; for the less likely 1789 date, see Kiernan, "Madden, Thorkelin, and MS Vitellius/Vespasian A XV," The Library: The Transactions of the Bibliographical Society, VI, 8.2 (1986), pp. 127-32.

§ 19 Following the links to "Part One: Thorkelin's Discovery of Beowulf," "Conclusion," or "Madden, Thorkelin, and MS Vitellius/Vespasian A XV," will, in each case, take you to an "electronic version" (i.e. what appears to be a transcription in HTML rather than a digital offprint or preprint) of previous publications by the author on the subject.

§ 20 This is helpful in one sense—the Rosenkilde and Bagger volume in particular was published in an expensive limited run and can be hard to find for anybody not at a good research university or without access to an interlibrary loan service.

§ 21 But it is also bibliographically problematic. The reprints do not reflect in anyway the pagination of the original documents and are stored at what looks to be an unofficial repository without any indication as to their copyright status or their likely permanence. This means that they cannot be cited by others in this form with any confidence as to their longevity or even the detailed accuracy of their contents. In at least one case, indeed, the documents have been slightly improved over the original: Kiernan notes that he has replaced the black-and-white images in one of these reprints with colour, a move that improves their convenience but reduces their bibliographic fidelity. It also raises the question of whether typos and other minor errors could have been corrected in the case of other documents in the collection.

§ 22 In as much as they contain only Kiernan's articles, moreover, these references provide a one-sided view of the debate: readers are provided with convenient reproductions of the details of the editor's views on the Thorkelin transcripts, but nothing other than bibliographic references (found in these same reproductions) to the work of others writing on the same subject. In contrast to the summary of scholarly consensus and relevant bibliography one might find in a more traditional comprehensive edition, the "Index and Guide" to EB 4.0 is better understood more as a minimal jumping off point for access to the edition and for a few topics only, reader-directed research.

Detailed textual notes

§ 23 In contrast to its relatively skeletal introductory guide, the "textual notes" to the edition represent a considerable, original, and self-contained achievement. These notes seem to be "textual" exclusively in the paleographic and codicological senses rather than the editorial (i.e. they are about the transcription of the witness rather than also about the (re)construction of the editorial text). But they bring to bear the full power of Kiernan's many years of unrivalled and technologically sophisticated access to the Beowulf manuscript. In the same way that professional Anglo-Saxonists are able to benefit from the first-hand observations of a Ker or Wanley (Ker 1990; Hickes 1705; Tite 1984), here too we benefit from Kiernan's willingness to share his close and detailed knowledge of the Beowulf manuscript and his ability to use digital techniques to answer many long-standing questions about the poem.

§ 24 Improving on Wanley or Ker, moreover, Kiernan is able to provide photographic evidence for what he sees: the textual notes make liberal use of high quality detailed photographs of individual letters and, in many case, specially processed images that demonstrate Kiernan's textual arguments.

Figure 1: Textual notes to f. 179r showing UV and colour photography.

Figure 2: Textual note showing reverse image processing for f. 179r.

§ 25 In many cases, but not all, clicking on an image in the textual notes opens a separate window with the same note and a much higher resolution image (while this works with standard colour and ultraviolet photographs in the textual notes, it does not seem to work with some of the enhanced photographs, such as the one showing subtraction in Figure 2).

Figure 3: Textual note showing high resolution detail for f. 179r.

§ 26 These notes are a considerable achievement and represent perhaps the greatest value in the edition for a scholar of the poem. They provide unrivalled access to details of manuscript and Kiernan's ability to document his claims photographically represents a considerable improvement over previous, print era, approaches to the paleographic and codicological description.

Useful display options

§ 27 The focus on photographic evidence in Kiernan's textual notes reflects the "image-based" approach to the manuscript that informs his edition as a whole. While EB 4.0 is not the work to turn to if you want to know more about what people think of Beowulf as a literary text or cultural milestone, or to find a broad entrée in Beowulfian scholarship, it is a resource to come to if you want to see how the text is represented in the manuscript, evaluate somebody else's reading or argument about the text or manuscript of the poem against what the manuscript shows, or get (something) of a sense as how the text appeared to medieval readers. Students in our seminar greatly appreciated the way the edition allowed them to follow along in the manuscript as they prepared for class and compare editorial readings in Fulk et al. (our class text) against the manuscript evidence.

§ 28 Exactly how this evidence is represented has been a focus of the Electronic Beowulf since its first edition, and the question is also one of the main reasons early editions used Java so extensively. The standard view opens to the beginning of Beowulf (other texts and transcripts from the manuscript are available) with a reading text on the left and a manuscript image (at about 80% full size on one of our computers) on the right.

Figure 4: The opening layout of EB 4.0

§ 29 A pull-out menu allows the user to change the transcript to provide a "critical edition" (i.e. with emendations, conjectures, some diplomatic information, and links to textual notes). On the manuscript side a button on the top bar toggles an overlay that ties the same textual notes to the image as well.

Figure 5: Pull-out menu (centre, cf. Figure 4) with a critical text (left panel) and image panel showing location of textual notes (black boxes, right panel).

§ 30 A "View" button at the top menu bar opens a dialogue that allows the user to change the order and orientation of these panels. The panels can be reversed (e.g. with the manuscript on the left and the transcription or edition on the right) or replaced by a number of different views and transcriptions: facsimiles of Thorkelin A or B, the Connybeare collation, or the Madden transcription. The menu can also be used to arrange the panels vertically rather than horizontally, and control how the panels behave during navigation and whether they move in sequence or independently of each other.

Figure 6: Setting the viewing options.

§ 31 With some manuscript images, a second button at the top of the panel allows the user to choose "conjectural restorations" and (for f. 179r only) an "offset" view. In the first of these, images of letters from the same scribe(s) are superimposed over textual gaps in the existing page in order to suggest how a reconstructed reading might look (when these reconstructions appear, they tend to include the lost letters at the edges of the manuscript; although all pages have such lost letters, however, not all pages offer this "restoration" feature). In the second view, a reversed image from f. 178v is superimposed over f. 179r to show how the letters from the preceding verso affect our understanding of the letters missing on the recto. In each case, a slider along the bottom controls the opacity of the overlay, allowing the user to compare the existing and "restored" states.

Figure 7: Screenshots showing the Overlay button (left), the "offset" view of f. 179r (reversed superimposition) with opacity set to about 20% (centre), and the "conjectural restoration" view with opacity set to 0 (right).

§ 32 The interface is, as these screenshots suggest, quite polished. In particular, students appreciated the ease with which the text could be navigated and arranged. At the same time, however, there are some problems worth noting. Although the edition has been entirely refactored, some minor display issues with EB 3.0 discussed in Simpson (2012) persist, particularly the way in which navigation buttons move as the edition is paged through. In addition, managing the interface's options and tools can be a little bewildering at times: sometimes similar options are controlled and displayed from different places or types of interface (e.g. buttons, drop-down lists, pullout dialogues, pop-out windows and dialogues that open in new browser windows, pop-out windows and frames that are programmatically generated within the interface, and so on). In other cases, things that seem very different are accessed via the same button: "restorations" (i.e. image-based reconstructions of lost text) that apply to many pages in the edition are invoked using the same button that, on f. 179r alone, also opens an image in which a reverse image of the previous page has been superimposed (see Figure 7; Rosselli Del Turco 2011 provides a general discussion of these kinds of problems in edition user interfaces).

§ 33 A final problem involves the ability to cite "state." Because digital editions are dynamic, users need to be able to refer to the way a particular view of the text was arranged: the particular combination of facsimiles and transcriptions, or specific versions of an image for which multiple versions exist. Because navigation is handled programmatically using Javascript, rather than by URL in the browser window, there is no unambiguous way of citing such specific arrangements of elements. The viewer URL remains the same (http://ebeowulf.uky.edu/ebeo4.0/CD/main.html) no matter what changes are made by the user to the interface and this means that programmatically generated views (i.e. views, such as those in Figure 7, that are generated by Javascript and do not seem to exist as separate documents on the server) have no recoverable URLs: the only way of pointing readers to the overlay page for f. 179r, for example, is to explain how they can recreate it themselves (for a contrasting approach, inspired by previous versions of the EB interface, see Rosselli del Turco et al. 2013; Rosselli del Turco and Masotti 2013).

Grammatical and metrical tools

§ 34 A final set of options, greatly appreciated by students in the class, provide various metrical, lexical, and grammatical aids. Like the Critical Edition/Transcription toggle, these can be switched on and off from the central pull-out menu on the Critical Edition or Transcription views (see Figure 5 above). When toggled on, the aids appear as a "tool tip" (i.e. yellow overlay with relevant text) whenever the cursor is held over an individual word or letter. Some users found that these tips worked on both traditional (mouse and keyboard) and touchscreen environments (e.g. on a tablet or touchscreen computer), others had difficulty with them in a touchscreen context. In addition to adding metrical and lexical information to the tool tip, the "Metrical" and "Glossary" toggles also open additional windows containing a glossary and various metrical tables. Double clicking on a word in the poem opens the glossary window to the entry.

Figure 8: Screenshot showing tool tip with all grammatical, metrical, and lexical information turned on. Glossary and metrical windows are over the right panel and can be resized and moved.

§ 35 These tool tips were identified by students as being among the most useful aspects of the edition when preparing or reviewing for class: they provide a quick summary of the grammar of the forms in question, with a literal translation of the line and a gloss on individual words.

Sometimes circular search tools

§ 36 EB 4.0 also comes with a suite of search tools. There is a detailed explanation in the Index and Guide of the options available, which allow for searches by word, (sub)string, "alliteration" (a search for individual letters forming an initial rhyme found in certain metrical positions within a given line), or "apparatus" (which is less a search than a series of indices collating scribal abbreviations, emendations, and similar forms in the transcriptions and editions).

§ 37 The students on the whole found these searches difficult to use (one reported that she was never able to get results even when she was looking at the form she was searching for). This is in part because the searches do not work equally well under all conditions (the guide recommends searching when the text is in the left frame, for example).

§ 38 It is also in part because the list of options is eclectic. Word and string searches are both standard and relatively straightforward, even if students had a hard time getting them to work. But the "alliteration" and especially the "apparatus" searches reflect relatively specific research goals that privilege some views of the text over others: "alliteration" privileges identifying initial rhyme over all other metrical questions (e.g. potential end rhyme, number of alliterating syllables, accidental [non-stressed] alliteration, different grammatical constructions, and so on—though information about many of these can be accessed using other tools). Likewise, the "apparatus" search recovers information that the designers appear to have encoded specifically to be recovered by this search (i.e., in essence, the search engine equivalent of a circular argument). Although the edition is said to have been encoded "in conformance with XML" ("Seach Options"), neither the location of this XML nor its format and contents are described explicitly in the "Guide and Introduction": we found the file, which seems to use a custom schema, only after searching the search engine javascript file for the document call (the two XML files we found are http://ebeowulf.uky.edu/ebeo4.0/CD/search/ed.xml and http://ebeowulf.uky.edu/ebeo4.0/CD/search/ms.xml). The "apparatus search" appears to recover previously information encoded with the <res> element (i.e. encoded so that it could be found by this specific search).

§ 39 A final problem with the search function is that it is hard-coded into the edition. With the exception of the "apparatus" search (which is conceptually more a dynamically assembled table than an actual search), the search engine really performs a sub-set of the many string and regular expression searches one could do with a utility like grep. But because they are hard-coded and the XML is hidden from the view of those unprepared to look for it, they cannot be modified, require considerable explanation, and can't easily be circumvented using other tools.

Overdetermination

§ 40 The problems with the search function speak to a larger problem with the design of the Electronic Beowulf as a scholarly and teaching resource—a problem that extends back to the first edition and that has survived its repeated refactorings: in attempting to provide users with everything they need to explore the manuscript on their own terms, the edition works too hard at anticipating those terms and, paradoxically, constrains them unnecessarily.

§ 41 This is a common problem in the design of digital scholarly editions (see O'Donnell 2005 for an early discussion): given the flexibility, dynamism, and power of the new technologies (as well as a sense that their readers may lack crucial computational skills, see O'Donnell 2009), designers are often tempted to over-specify their editions, to understand them as a computational platform for the exploration of a text rather than a representation of the text in computable form (this distinction between "programmable" documents and "static" documents goes back to the very earliest web standards; see Connolly 1994 for a discussion).

§ 42 The result is that the edition can end up preventing users from doing things its designers have not anticipated. As mentioned above, the EB 4.0 search function provides a number of potentially useful tools, but omits others that different users researching different questions might find equally interesting or useful. By making it difficult to access its XML and using a non-standard document model, the edition prevents all but the most determined users from checking the editor's work or modifying or supplementing it to explore different research questions the editor has not already thought of. And while the viewer anticipates many of the different ways a user might want to arrange its component parts, it also fails to anticipate many others: a user cannot tile three or four versions of the text (both the "edition" and the "transcription" say, as well as the manuscript facsimile and one of the modern transcriptions), or, of course, compare the facsimile of the Beowulf manuscript against facsimiles of the manuscripts of other poems that the user might find relevant to answer some unanticipated research question. While a heavily determined interface is acceptable in an edition intended primarily for students, this same (over)determination makes the edition less useful for more advanced researchers working on more cutting edge problems.

§ 43 In some cases, it is possible to overcome this problem relatively easily. It is possible, for example, to access individual pages from the manuscript facsimile directly by URL on an image-by-image basis (i.e. rather than as a collection) by "right clicking" within a web browser and selecting an option such as "Open image in new tab" or "Save image as."

§ 44 In other cases, however, it is more difficult to extract such information. The textual notes, for example, are individual files organised using a difficult-to-predict naming convention, meaning that they are difficult (though not impossible) to access by URL. The special overlays and "restorations," on the other hand, appear to be generated programmatically within the interface, making them apparently impossible to access directly, outside the viewer the edition provides.

§ 45 Even text that is accessible can be difficult to use. The "edition" and "transcription" texts of the poem, for example, appear to be different views of a single HTML file (http://ebeowulf.uky.edu/ebeo4.0/CD/ed/ed.htm). While the fact that they can be referred to by URL makes them directly accessible (and hence repurposable), they also require considerable cleanup before they can be used for anything other than viewing in the user interface: the text of the poem is encoded as a table with individual lines as rows (line numbers are encoded as a separate cells that are sized absolutely using a width and height in pixels); the character encoding is Windows-1252 rather than the standard UTF-8; and there are many examples of invalid or non-standard element-usage in both the head and body: for example, several undocumented custom elements (r, hm, and bb), and a large number of unclosed elements, including 42 unclosed line breaks (i.e. html:br) at the document end. All of these elements and character issues caused trouble when we tried to load the document into standard XML editing software such as OxygenXML.

§ 46 Many of these coding techniques have been deprecated for years, particularly the use of tables to position text, proprietary character encodings, and absolute values for element sizing (see, for example, Nielsen 2000). Others, such as the use of undocumented custom elements for what appears to be the encoding of emendations and corrections, duplicate functionality available using well-established and well-documented standards-based approaches (e.g. Bauman and Burnard 2008).

§ 47 What is common to these techniques, however, is that they exist for the convenience of the programmer rather than the user: these 42 lines breaks at the end of the text, for example, do not describe the Beowulf manuscript in anyway. They presumably serve some positioning purpose in the edition interface. Similarly, the representation of the text of Beowulf as a series of rows corresponding to individual metrical lines privileges a modern view of the poem over all others (and makes these alternate arrangements (such a one arranged by manuscript line) difficult or impossible to arrange. In other words, in EB 4.0, the reliable, long-term scholarly description and representation of the text of Beowulf in computational form (presumably the ultimate goal of any digital edition) is being subordinated to the arbitrary immediate programming needs of a custom-made viewing application. This prevents the text of the edition from being reused in other viewing environments without considerable ad hoc revision (see again Connolly 1994), and it prevents the environment from being used to display other texts that have not been encoded using EB 4.0's custom markup system. Given that the edition has been refactored at least three times in the last 15 years due to changing technological standards, this emphasis on custom and ad hoc programming and encoding seems likely to harm the edition's long-term usefulness.

Archive and version control issues

§ 48 A final problem with the reusability in this edition has to do with archival and version control issues. In common with many Humanities web-based projects, EB 4.0 is remarkably difficult to cite in an unambiguous, archivally-responsible fashion. The edition is published across a couple of stand-alone subdomains and web directories at the University of Kentucky: the introductory Guide and Index and the edition viewer itself are hosted at http://ebeowulf.uky.edu/; the repository of associated texts referred to in the introduction are hosted at http://uky.edu/~kiernan/. Although logos for the University of Kentucky and the British Library appear on the title page of the edition, there is no explicit indication that the edition is in fact being curated or maintained professionally by a library or publisher: a university sub-domain can be established as easily by a university IT department as a university library—and taken away again in response to changing employment situations or administrative reorganisations.

§ 49 There is also no unambiguous identifier for the edition. As mentioned above, three of the previous four versions were published to CD-ROM or DVD. This means that all three can be unambiguously identified through their ISBN numbers. While all three have been deprecated for security and technological reasons, it remains possible to discover and cite them—as indeed, we had cause to do in the course of preparing this review.

§ 50 Version 3.1 of the Electronic Beowulf (the first version distributed via the web rather than CD-ROM, though it required the CD-ROM for installation), however, is very difficult to find. Published to the web without any permanent identifier or, apparently, commitment to long-term preservation, that specific version of the edition is known to Google from only two locations: a reference to the edition in the current EB 4.0 (i.e. at http://ebeowulf.uky.edu/gettingstarted/overview) and an apparent ghost reference at a suspicious file-sharing site, "bestsearchengineever.us":

Figure 9: Google Search Engine Result Page for search "electronic beowulf 3.1".

§ 51 The situation is actually slightly better than this implies, however, since the Internet Archive (which is not indexed by Google) has been taking snapshots of the http://ebeowulf.uky.edu URL since it was first established in 2011. These snapshots include a link to a copy of the EB 3.1 installer (https://web.archive.org/web/20140603150425/http://ebeowulf.uky.edu/eBeo3.1_Installer/EBeowulf3.1-installer-j7.jnlp) as well as links to a (broken) beta version of EB 4.0 and various versions of the current edition.

§ 52 The Internet Archive, however, should really be understood as a "last chance" archive. It is not intended to serve as a scholarly repository and it works best on sites that, unlike EB 4.0, use minimal, standards-based HTML markup (indeed, the Internet Archive FAQ page specifically mentions the use of Javascript as something that often prevents it from archiving webpages successfully; see Internet Archive 2016 "How do you index dynamic pages"). The Internet Archive is also quite unwieldy as a resource because it records website changes rather than releases—a fixed typo can be as significant to its algorithm as a completely refactored edition. Thus on July 4, the Internet Archive reported that it had recorded 86 snapshots of the top-level page (http://ebeowulf.uky.edu) between April 5, 2011 and June 24, 2016, even though this period saw the release of only four major revisions of the edition (versions 3.0, 3.1, 4 beta, and 4.0). Because the Archive does not archive server-side programming or go into password protected sites, moreover, these 86 snapshots still manage to miss some important revisions: the beta version of EB 4 required users to register and authenticate with a username and password; this means that it is not fully preserved in the Archive.

§ 53 Preservation in the Internet Archive, moreover, does not solve the problem of sustainable citations. Because online versions of the EB seem to be revised in place (i.e. without changing the URL), it can be difficult to match a specific citation to a historical version of the edition. Thus the chapter in the Guide, "Studying Beowulf," goes through several major and minor revisions between its first appearance (referring to version 3.0) in early 2011 and today (cf. https://web.archive.org/web/20110630174211/http://ebeowulf.uky.edu/studyingbeowulfs/overview, https://web.archive.org/web/20120615142836/http://ebeowulf.uky.edu/studyingbeowulfs/overview, and https://web.archive.org/web/20151120021942/http://ebeowulf.uky.edu/studyingbeowulfs/overview). Since all of these versions were originally found at the same URL (i.e. http://ebeowulf.uky.edu/studyingbeowulfs), which is what they would have been cited by, it can be difficult to know precisely which version a citing article means when it uses this URL. Even a "date accessed" value doesn't necessarily help here: snapshots are dated in the Internet Archive to the day the web crawler discovered the change rather than the day it was made—it is entirely possible that a citing article may have accessed a changed page before the Internet Archive recorded it.