When one sees an electronic edition one wonders how much work and effort has gone into it and the first question that springs to mind is could I do something like that? The short answer is: No, you can't. There are other kinds of electronic editions that a scholar could produce on his own, but the multi- witness editions of the Canterbury Tales Project are just out of the range for individual production. The proof of this is that the Project has developed, since its beginning, partnerships and collaborations at different levels with many scholars. Since its last expansion - with two new members - the Canterbury Tales Project is now producing our CD-ROMs more speedily. Of course, another important factor that has enormous weight on the Canterbury Tales Project's production improvement, is the fact that most conventions used for the conversion into computer formats have long been established and tested, and are now more reliable that ever before. An idea of how many scholars have participated in a particular edition can be drawn by looking at the opening pages of one of CDs. I first approached The Wife of Bath's Prologue on CD-ROM in December 1999. I was then fascinated by the idea of all the manuscripts of the Canterbury Tales being transcribed and published. I also held a very strict textual-critical position that made me into a severe documentary editor. This CD-ROM pleased me greatly - although not completely - and I was particularly impressed by some of the transcription policies of the Project. But what really engaged me was the idea that Collate 'the main program used in our collations' was such a wonderful tool that could help creating this kind of materials. I was quite ignorant then and, even if I understood what coding was and had tried to teach myself SGML, I was far from being able to truly understand the nature and complexity of the production of one of these CDs. What I did then is what most of my colleagues would have done: I read everything I could get a hold of in which members of the CANTERBURY TALES PROJECT had participated. Eventually, I came across a statement similar to the one found in The General Prologue CD-ROM: “"The computer collation program we are using (Collate) permits regularization as part of the collation process. This has the great advantage of allowing deferral of regularization until all the evidence of all the spellings in all the manuscripts at any one point is available. It also permits a complete record to be made of all regularization done during the collation. Collate can also generate regularized-spelling version of each file from the regularization process."” My interpretation was that Collate could automatically generate regularized texts from the unregularized ones. This is just shows how naïve and ignorant I was. Collate is a wonderful tool and helps our work in unimaginable ways, but it can not take the various spellings of Middle English and regularize them to a particular form. In fact, because of the nature of the language at that time, I am doubtful that any automated tool could do this. So how do the complex spellings of the manuscripts get transformed into a regularized collation? This is just done by hand. Word by word, line by line, 20.000 lines of verse in 88 fifteenth-century witness are lemmatized and regularized by the Canterbury Tales Project team. The process is intricate, requires a great deal of attention and grammatical skills that few people arrive with, and when it is over, every single one of the regularizations must be checked to make sure that most mistakes are eliminated. Both the lemmatization and regularization are building blocks of our spelling databases. An important question that one might want to ask to the members of the Canterbury Tales Project is why should anyone go through the process of creating very detailed transcriptions of the manuscripts, making sure that tails and flourishes are accurately represented in the key manuscripts, if later one has to spend money and time in taking all those differences out to produce the regularized collation? The answer is quite simple, the regularized and unregularized collations have very different function in the CD-ROMs. Probably, the main reason to have chosen to produce a regularized collation is the fact that only using this one can proceed to make an adequate stemmatic analysis. Since one of the main aims of the Canterbury Tales Project is to achieve a better understanding of the textual tradition of the Tales and the ways in which the manuscripts relate to one another, it seems clear why the role of the stemmatic analysis should be a priority. One of the most evident discoveries of the Project is the fact that spelling variants are uninformative while using phylogenetic programs - which are used as an important part of our stemmatic analysis - and that, in fact, these create 'white noise' and impair the results yielded by evolutionary software. Using examples drawn from my work on The Miller's Tale on CD-ROM I would like to demonstrate some of the problems that we face in our everyday work. Moreover, I will discuss other issues generated by editions of the kind produced by the Project, such as working with other people - nearby and far away, consistency, revision and responsibility. In cases like this, it becomes clear that the more people work in a particular edition, the bigger the need for strict rules to be applied in each particular case.

The Greek New Testament represents, by every measure, the Everest of textual scholarship. Firstly, it is simply the biggest: with over five thousand surviving manuscripts and other witnesses, from over two thousand years, it dwarfs every other textual tradition of a major western text (some Indian textual traditions, where copying onto palm leaf manuscripts persists to this day, are larger in terms of sheer numbers of manuscripts). Secondly, it is the most complex: beside straight-forward copies one has to deal with a vast spectrum of versions, in many languages and from many cultures, some of them now deeply obscure. There is also considerable citational evidence, where scraps of text are referred to in early Christian and other writers, some of which bears crucial witness to textual readings, even whole texts (the elusive 'Gospel of the Hebrews', for example), otherwise unknown. Thirdly, it is the most intractable. Many textual scholars have lived by a comfortable presumption that as one tunnels upwards in a textual tradition towards the origins variation will diminish, to the point where it may be eliminated altogether and a single perfect and original text (in this case, indeed, the Word of God) will stand forth. But the situation with the Greek New Testament appears to show the precise opposite: variation becomes greater, not less, as we move back in time, and the earliest substantial evidence we have from the second century shows witnesses which differ more from each other and from the late text than do later writers. Add to this that the Greek New Testament is, by some considerable margin, the most important single text of western civilization, and indeed the foundation of nearly two millennia of our culture, and we have a formidable task. Six centuries of textual scholarship has defined itself against this task, and the names of the editors who have struggled with this text and its problems is an impressive rollcall: Erasmus, Griesbach, Bentley, Lachmann, Tischendorf, Westcott, Hort, von Soden, Nestle, Aland. Sooner or later, every theory of textual scholarship, and every technological development, must test itself against the Greek New Testament. The early development of printing in the west (and perhaps even its invention) was driven by the church's need for uniform texts; one could argue that printing reached its first technical peak in the polyglot Bible of 1515. Now, it is the turn of computing technology and electronic publishing to confront this challenge. Over the last five years, the Institut für Neutestamentliche Textforschung, Münster, (INTF) has been progressively incorporating computer-assisted techniques in the preparation of the new Editio Critica Maior series of the Greek New Testament (ECM). The most recent volumes of this have used computer methods not only in preparation of the printed text, but at every stage in gathering the data on which the printed text is built. The manuscripts chosen for inclusion in the ECM apparatus are now transcribed in full, collated using Robinson’s Collate software (originally developed for medieval vernacular texts), and the collation output to a database where it is integrated with other evidence and the apparatus for the ECM print editions created. This work involved testing and development of computer techniques capable of coping with the special demands of the Greek New Testament. For example: Robinson has had to add many new facilities to Collate, and enable it to cope with collation of up to 500 texts at once. At the same time, the INTF has had to redesign how it carries out the work, rebuilding it around full transcription and collation of the manuscripts rather than manual excerpting of variants. So far, this work has been limited to the making of printed texts. Recently, and encouraged by the success of this process, the INTF and their publisher, the Deutsche Bibelgesellschaft (DBG) have determined on a more ambitious program: the making of an electronic edition of the 28th edition of the Nestle-Aland Greek New Testament. The Nestle-Aland text is far the most widely used text of the Greek New Testament: it is the text published in the United Bible Society publications, it is studied in seminaries and used in translations worldwide. Just to make an electronic version of this printed text and apparatus alone would be a signal advance, and in itself a considerable challenge. However, the INTF and DBG propose far more: the electronic version of the existing text should be interwoven with full transcripts of the key manuscripts (and, as time passes, more and more manuscripts), collations of these, and analytic tools including the facility to carry out dynamic comparisions of manuscripts, build stemmatic analyses from the comparisons, and much more. Beyond this, the possibilities of linking to on-line manuscript images and lexicographic resources open yet further perspectives. Much of this will be built on the existing Collate software, and over the next years Robinson and other Leicester staff (both from the Centre and from Scholarly Digital Editions [SDE]) will be working with Wachtel and other staff at the INTF and DBG to set up the work practices necessary to support this, and developing a series of prototype publications. With funding from the Deutsche Forschungsgemeineschaft and the DBG, the first of these prototypes will be publically available on the web in February 2003. We will present a version of this first prototype at the conference. All text will be encoded in XML, following the TEI guidelines. The first prototypes, at least, will use Anastasia, the electronic publication system developed by SDE.

Turning basic text into a simple publication can be a relatively easy matter. It could be as easy as using a word processor or typewriter, but what if the text is more complicated? For example, what if the text was written in a different language, or you needed to include information pertaining to the original physical document? These can be some of the obstacles in creating a digital publication. The first step in this process is to find a means of encoding not only the raw text but also the information relating to the document or parts of the document. If it was a simple document produced in Microsoft Word we could mark parts of the text as bold, italic etc., but this method provides nowhere near the flexibility or power which is required by most. HTML, HyperText Mark-up Language, is a step closer as parts of the document can be marked or tagged to describe the structure, such as the title or the document body. Still this method is inflexible and very restrictive in the information it allows you to show. What we need is a way similar to this but also allows us to define how and what information we need to keep. For their transcription work the Canterbury Tales Project chose XML (eXtended Mark-up Language), and before this its predecessor SGML (Standard Mark-up Language) as their format for transcribing the Canterbury Tales. This allows them to transcribe the manuscripts in almost infinite detail, from something as simple as marking a character as being a particular medieval style of letter to showing that a sentence was added at a later date. For them to create a way of encoding this selected information all they need do is use a tag, such as <added>, which itself could hold extra information about this occurrence of added text. So wherever there is a need to include information in a portion of text either an existing tag can be used or a new one created to suit their needs. This style of encoding can have potential dangers. When several people are working on a transcription, they need to make sure that they agree on which tags they are going to be using and in which circumstances they should use them, otherwise the resulting transcriptions would be inconsistent and impossible to use in a real publication. In an effort to solve this problem Peter Robinson, director of the Canterbury Tales Project, helped produce the T.E.I. documentation. This set of documents created the standard which the C.T.P. uses to encode the manuscripts they are working on, and by which all their transcribers work by in order to create a consistence format for their publications. So now we have a set of files with all the text and its relevant information encoded within, but this format is not going to be presentable to the public. What we need is to create a system that can turn this information into format that can be read and navigated with ease. To this end Scholarly Digital Editions created Anastasia, an application that takes the transcribers' XML files and produces intermediate files which enable it to search these XML files and can tailor its output to suit the needs of the publication. Anastasia acts like a web server so it can be used in conjunction with a web browsers such as Internet Explorer or Netscape. The benefits of a system such as this is that most users will have some knowledge of these applications so they will not need to learn a new system in order to be able to use the publication. Other digital publication applications have chosen to create their own interface for the users, and this may allow them to refine the way they present the information to produce a view of the information that is exactly what they need. But this may have the drawback of having a steeper learning curve when people start using these applications. An advantage of the Anastasia system is that it gives extraordinary control over the interface we present to users. This makes it possible to tune interfaces precisely as we want. We are currently working on integration of mySQL into Anastasia. The reason for this is that there are certain operations which are very well handled by databases (notably, sorting of results of searches on particular fields). We also intend to integrate certain of the Collate functions into the software. Producing electronic publications is a long and laborious process, not only in transcribing the manuscripts but also in deciding how to encoding the information and then how best to let people view the resulting publication. At least this part of the process can be simplified by collaborating with publishers and choosing "off the shelf" applications.