What has gone right

§ 9 I do not think this is the case. We have seen several electronic editions which indeed do all that was imagined of them—by some of us anyway—ten years ago. We presumed then that we would find editions which presented the full text of important original sources alongside high-quality digital images: the Rosetti Archive ( http://www.rossettiarchive.org/ ), SEENET , the Canterbury Tales Project , and the Electronic Beowulf (among others) have achieved that. We presumed then that text could be presented in many forms, representing different levels of editorial intervention and of readerly interest: again, the projects published over the last decade have offered just this. We presumed then that we would be able to offer a multiplicity of new ways to present variation among texts, freeing readers from arid collations: the Canterbury Tales Project publications offer line by line comparisons between manuscripts, highlighting their differences, and much else. We presumed then that we would be able to offer readers a wealth of new tools to permit dramatic visualizations of relations among our texts and assist in their exploration: you can indeed find these now in, for example, the Blake archive ( http://www.blakearchive.org/ ), Estelle Stubbs' edition of the Hengwrt digital facsimile ( Stubbs 2001 ), Martin Foys' Bayeux Tapestry Digital Edition ( Foys 2003 ), and my own edition of the Miller's Tale on CD-ROM ( Robinson 2004a )—not to mention perhaps the most elaborate and ambitious of all current electronic edition projects, the 28th Nestle-Aland edition of the Greek New Testament ( http://nestlealand.uni-muenster.de/ ). We presumed then that these texts would be searchable, in many different ways—and indeed they are. In one important respect, moreover, digital editions actually have greatly exceeded our original hopes: the extraordinary progress in computing hardware and software over the last ten years has made it possible to give editions interfaces of remarkable beauty and flexibility.

§ 10 From these examples (and there are several others) I think it is fair to conclude that digital editions can and do fulfill the hopes we had of them. True, print editions have not stood still over the period: computer-based typesetting and new production methods make it possible to create beautiful new print editions, such as Jim Mays' Coleridge (e.g. Mays 2001 ), or the Cornell Wordsworth (e.g. Parrish 1977 ). But I believe that the digital editions we have made over the last ten years now far exceed their print equivalents in what they include, in their ability to shape themselves to the reader's needs, in the routes they offer to understanding highly complex webs of knowledge. There is good evidence, too, that others share this view: in recent years, major academic awards have been won by digital editions. In 1998, the Beatrice White prize for an outstanding publication in Medieval and Renaissance studies was awarded to my edition of the Wife of Bath's Prologue ( Robinson 1996 ), the first time the award was given to a non-print publication. In 2001, Kevin Kiernan's Beowulf ( Kiernan 1999/2003 ) repeated the feat.

What has gone wrong

§ 11 However, there is one crucial respect in which digital editions have not achieved our expectations. Some ten years ago, there were at least three major publishers offering to publish them: OUP, CUP, and Michigan. As I have explained above, all three have now stepped out of the field—a loss that is not made up for by the creation of SDE, operating out of my spare bedroom. We thought then that we had a sound publication model for digital editions: major publishers would publish them, just as they have always done for print editions. But this has not happened. Further, we now know anecdotally that many scholars remain sceptical of electronic publication. Combined with the movement by leading academic publishers away from this field, this scepticism leads rather easily to the opinion that electronic publication is not real publication at all. It is depressing to find cases where scholars do not use the digital editions one has gone to such trouble to make, even when they know of and have access to them. To give just one example: my edition of the Wife of Bath's Prologue and several later Canterbury Tales Project publications include Dan Mosser's descriptions of the Canterbury Tales manuscripts. These descriptions are the result of several decades of work by Professor Mosser, in the course of which he has inspected every manuscript and every complete incunable copy (and very many fragments too); consulted with every leading scholar; read every article of note; and built up a formidable expertise in palaeography, codicology, and watermarks. By all odds, these are not just the most recent, but also the most careful and comprehensive accounts ever made of the Tales ' earliest texts. Despite this, I have come across several examples of work, even by senior scholars with access to Professor Mosser's research, where these essential resources have not been cited.

Tools

§ 13 Which leads us to the paradox: if digital editions are so manifestly superior, then why indeed are we in the state of affairs described above? Why are so many scholars, and so many scholarly projects, still making print editions? I suggest that the answer to this lies almost wholly in the second aspect to this problem: the availability of usable tools. Over the past two decades I have made two of the leading tools for making scholarly electronic editions. The first is the collation software Collate , which I first wrote as a set of VAX routines in the 1980s, and re-wrote into a Macintosh program in the 1990s. The second is the XML publication software Anastasia , which I initiated in the mid-1990s. Several of the electronic editions named above depend heavily on these two tools. One can assert that it is indeed possible to use them to make digital editions which offer all we could hope for. But as their creator I think I am uniquely qualified to note that they are not easy to use: if everyone who wanted to make digital editions was required to use these two tools, very few digital editions would ever be made. Both tools require very high levels of dedication (though not particularly advanced computer skills). They can be frustrating in the extreme: perhaps refusing to work at all, or (just as bad) working in quite unexpected ways for reasons which are far from obvious. One can spend hours—days even—pursuing some slight problem with the programs' output. Typically, an editor using these tools will need large amounts of time, dogged determination, disciplined organization, and considerable help. It is not quite true that I am the only person in the world who can use Collate and Anastasia (it would be an interesting exercise to trace all the humanities software that has been used only by its creators)—but it is uncomfortably close to the truth. The wonder is not so much that rather few projects have managed to go the whole distance with these tools, as it is that any have been able to do so at all.

§ 14 There are other tools about. The success of XML in establishing itself throughout the business and academic communities as the universal encoding system has brought with it many new tools for preparing and publishing texts. Furthermore, for more than a decade now, the Text Encoding Initiative (TEI) has been preparing text-encoding guidelines that have achieved wide acceptance in the humanities computing community. Indeed, the TEI encodings are fundamental to several of the editions named above, editions which simply could not have been made without the TEI. Once more, I believe I am in a unique position to pronounce on these guidelines, and hence on tools based on the TEI scheme in terms of their appropriateness to the wide range of actual scholars who want to make actual editions. I was chair of the working group that was responsible for the chapter on textual apparatus in the first published TEI guidelines, and so wrote the draft which formed the substance of the chapter. Somehow, although I was not formally a member of the working group on manuscript transcription, I attended a meeting of this group and ended up writing a draft that became the substance of the transcription chapter too. I am therefore responsible for some 50 of the 1250 plus pages of the Guidelines as they were published in 1994. Further, these are the 50 pages of most concern to scholars making editions of texts based on primary sources. Several digital scholarly editions have indeed used these guidelines profitably, so it must be said that in terms of their immediate aim—to provide encodings which would support such editions—the guidelines were and are successful. But in terms of another aim, to provide a system which any reasonably competent humanities scholar can use (which, eventually, is the only aim that matters), the guidelines are a failure. One has only to see the look of dismay on a scholar's face when encountering their full horror for the first time to know this. One may contemplate, with equanimity, every complexity of Byzantine medieval military history but be quite defeated by the unfamiliar vocabulary of the mysteriously interconnected universe which is the TEI. All scholars bring the same two questions: where do I start? and where do I find what I need? But the answers each time are different, and even those expert in the TEI may struggle to find them—while engaging in intense theological disputes over the correct interpretation. Little wonder then that scholars choose to find other things to do—or to make print editions. I do not mean by this that the TEI is fundamentally wrong in its design and aims. In fact, the first priority had to be to make a system that worked, with usability a later concern. But this is the situation it has placed us in. Putting it simply: if a scholar has to understand how the TEI encodings work—and indeed, requires any more than a basic knowledge of how XML works—in order to make an electronic edition, very few will ever do so.

§ 15 Of course, the TEI and XML are encoding systems, not tools. One needs tools to put the encoding in and other tools to publish the encoded texts. Here, however, we are even worse off. There are certainly exceptionally powerful tools about and, with the widespread success of XML, many to choose from. But once more, they are not for everyone. One may take as an example the Oxygen system. This is widely used by IT professionals, reasonably priced, and well supported and documented. But a glance at the vendor's website shows that it is not for the average academic working in the average department, with little or no specialist support. The first features listed at the site on January 15, 2005 were Different perspectives: source editor, XSLT debugger, tree viewer/editor and Support for XML 1.0, XML Schema, Relax NG , DTD and NRL schemas . A scholar struggling to find time to work on an edition between teaching, administration, and what used to be called leisure is unlikely to go much further. Yet Oxygen is simplicity itself compared to some other XML editing tools: just look over the correspondence on the EMACS editor on any XML mailing list (e.g. on TEI-L for 14-17 January, 2005 [ http://listserv.brown.edu/archives/tei-l.html ]).

§ 16 Even if our average scholar successfully does master Oxygen or some other XML tool, he or she then needs to seek out a publication system that is able to handle reasonably large and complex documents with appropriate searching tools and interface customization. In the past, a publisher might be expected to do this (as, indeed, publishers have done for centuries for print editors). But, as we noted above, the supply of publishers willing to support complex electronic scholarly publications has dried up. So this means that our editor is likely to engage in some form of self-publishing, with whatever help he or she can muster. And that is not easy. Our editor will need, first of all, access to a web server with full permissions to configure as needed (something many universities, rather reasonably, routinely deny to individual academics). Presuming he or she is actually given permission, he or she will then need to configure the system. Here is a rather typical configuration put together by one humanities computing specialist: ^{[ 3 ]}

1. Install the Apache webserver, set up so that it can be seen by the outside world
2. Install the MySQL database
3. Install the PHP package within Apache, permitting queries of the MySQL database for search purposes
4. Install the Sablotron XSLT processor to transform the XML files to HTML when found
5. Prepare scripts to translate the XML into the database, and further scripts for managing searching and display
6. Set this up, make sure all the pieces are working together, keep it running.

§ 17 To be sure, many people have been down this path before. Many computer systems arrive with many of these pieces already in place; XSLT scripts to manage the crucial XML transforms are also widely and freely available. But even with all this—and with much better documentation and training than is now available—one cannot imagine the gulf being bridged between the technical skills possessed by most editors and those required by publication systems they are required to use. It could be argued that the examples I have been giving are of open source systems, which typically do require higher levels of computer skill. But even if he or she can afford a proprietary system (and that is a large if), our scholar is still going to need a high level of skill to set up and configure the software, and then customize the interface for his or her data. And even then, he or she may still find that the software cannot do everything required of it.

Publishers

§ 18 This gulf between the actual technical skills of scholars and those demanded of the publication systems should not come as a surprise. Translate the terms of this discussion to the print world. No-one would expect a scholar, having written a book, to set the type, make the paper, choose, configure, set up, run, and maintain the printing press, operate the binding machines, pack the books into boxes, store them, and finally take care of their marketing and distribution. These are exactly the tasks that, for five hundred years now, publishers have done for scholars. But the disappearance of publishers from the field of electronic scholarly publication has left us with a problem. Who, in their absence, is to do the equivalent for a digital publication?

§ 19 The answer that has emerged over the past decade is that these critical technical tasks—where they are done at all—are performed by institutional IT or humanities computing specialists. This means that the only scholars who are able to make digital editions are those few who are lucky enough to be at institutions which have capable experts. ^{[ 4 ]} And there are rather few institutions that do—probably less than a dozen in the English speaking world. It might be argued that this should be enough: after all, fewer than a dozen academic publishers are sufficient to publish the great majority of the world's academic print output. But partly for reasons to do with academic politics, partly for reasons to do with simple logistics, the few institutions with such capability generally are unable to collaborate with many outside academics—and in some cases, with very few inside scholars either. It is not quite true that one needs to be tenured at the University of Virginia to make a scholarly electronic edition. Once again, however, it is too close to the truth for the health of us all.

§ 20 Suppose, then, you are a scholar—as most of us are—at one of the many institutions which does not boast a team of experts in TEI encoding and software tools. What are your choices? You have three:

1. Become an expert in these things yourself. This will take time (very large amounts of time) that could have been spent on the scholarship itself
2. Form a partnership with a group in an institution that has the necessary expertise. This requires attracting the attention of that group, which you will only do if your project is of overwhelming interest or (amounting to the same thing) you have lots of funding. Most scholarly editions do not qualify on either count.
3. Decide that, after all, you will produce a print edition.

Goals

§ 25 What will help, then? Our goal must be to ensure that any scholar able to make an edition in one medium should be able to make an edition in the other. Further, that an edition in either medium should be equally assured of appropriate distribution: just as once a library has bought a print edition it can be used by any member of the library for years to come, so too should it be for electronic editions.

Tools

§ 26 Let us deal with each of these goals separately. The first is a matter of tools. Fundamental to the model of electronic scholarly edition as it has developed over the past decade is the inclusion of full transcripts of all witnesses to the text. These transcripts need complex structural and other encoding from which computer programs may generate full collations of all the witnesses. Further computer programs may offer different representations and analytic tools for studying the differences and similarities among the witnesses, and then present all these, typically alongside images of the manuscripts, within a single electronic interface. At present, to do all these tasks requires a range of different computer programs. A tool that brought all these into a single interface—that permitted submission of the transcripts, their collation, adjustment of the collation, presentation of collation analyses, presentation of the transcripts, and integration of editorial matter, all on-line—would go far to ease the way for scholars into electronic editions. One immediate benefit of this is that it would permit scholars to see, almost instantly, the effects of an editorial decision. At present, the long chain of software needed to make an electronic edition means that work is done in batch mode with each link of the chain being a discrete production stage: you do all the transcripts, then all the collation, then adjust the collation, then you put it all in a publication interface, and only then do you discover you made a mistake right back in the first transcription stage. So you go back, redo that, and go right through the chain once more to see if the fix worked. This wide gap between editorial decision and edition realization is perhaps the single largest problem with current systems (including my own) for making electronic editions. It is also the single area where systems such as TeX or the CTE are stronger: the editor makes a change and within moments, you can see the result.

§ 27 I am not the only scholar to recognize this weakness in current systems for preparing electronic scholarly editions, or even the first. At least two major efforts are under way which, to varying degrees, address the problem. The first is the Archway project, led by Kiernan with other scholars from the University of Kentucky ( http://beowulf.engl.uky.edu/~kiernan/ARCHway/entrance.htm ). Its declared aims are to produce a system for building digital libraries of image-based scholarly editions for the humanities , and to lay the groundwork for sophisticated technical tools to interpret, assemble, disseminate, and maintain image-based scholarly editions on a continuing basis . The second is the NINES initiative, led by McGann and others at the University of Virginia. Among the many aims of this project is the creation of JUXTA, a text comparison and collation tool ( http://www.nines.org/tools/tools.html ; see also http://www.nines.org/ ). Both initiatives, however, have several other aims and it remains to be seen how far the tools they make will serve the needs of those of us who want to make electronic scholarly editions. It will not surprise the reader, then, to learn that I am planning to make such a tool. Tentatively named EDITION (for Editing Digital Interactive Texts through Online Networks) this will amalgamate the collation algorithms, built up over the last two decades and currently contained within Collate , into the Anastasia electronic publishing system. To these we will add a CVS system for handling different versions of submitted transcripts, a relational database for storing editorial information relating to regularization and variant setting operations and phylogenetic and variant database tools for analysis of patterns of variation. The entire suite of tools will be bundled together so that scholars will be able to submit, validate, and collate transcripts, view and explore patterns in textual variation, and then see the collation, analyses, transcripts, and manuscript images all using a single web interface. Effectively, this will collapse all the many links of the chain by which we currently make editions such as The Miller's Tale on CD-ROM into a single operation. We are currently (January 2005) seeking funding for this, with the aim of commencing work on this in July 2005.

Publication

§ 28 Tools such as those described above might go some way to helping achieve our first goal: to bring it about so that scholarly electronic editions require no greater level of computer expertise than is required say for the CTE. But they will not satisfy the second goal: to provide a route to electronic publication which is as secure (and, therefore, likely to be as esteemed) as is now the case for print publication. Quite simply, making electronic editions and then just parking them on a website somewhere, hoping (firstly) that scholars will flock to your site and (secondly) the site will miraculously sustain and develop itself over the years, is not publication. There is no help for it. For people to learn about your site, and—especially—for the site to be kept alive as computer systems change, someone has to pay money. No granting agency will ever commit the funds to do this, in perpetuity—and nor should they. It could be argued that the move towards free publication in the humanities— free meaning of course that someone else pays—might provide a platform for these publications. But scholarly editions are highly complex publications, far removed from the pdf files which are the staple of the Open Archives Initiative (OAI) and its congenitors http://www.soros.org/openaccess/index.shtml . I am very sceptical that systems designed for mass publication of scholarly articles and the like could cope with the editions we want to make (I am actually sceptical of the whole OAI concept, but that is another matter).

§ 29 On the other side, it can be said that scholarly editions of the kind we are making are among the must have items of any reputable humanities school. If we make them as they can be made, with the interface they should have, they will be used at every level: for reading, for exploration, for teaching, for simple fun. A successful model of institutional subscription has been established for electronic materials of this kind: witness the success of projects such as Early English Books Online ( http://eebo.chadwyck.com/home ), the MEC, and online versions of the Oxford English Dictionary ( http://dictionary.oed.com/ ) and Dictionary of National Biography ( http://www.oxforddnb.com/ ). These show that academic institutions (and, we can suppose, individuals) will pay reasonable prices for high-quality academic electronic materials. It seems a fair hope, then, that scholarly electronic editions can be published, maintained, and extended by a subscription model. Once more, I will be exploring this later in 2005 with SDE as we begin to set up online subscription to our publications. Perhaps publishers will return to the field of electronic scholarly editions, after all.