It is perhaps particularly obvious, especially in the case of Internet publication, that an item that is available online will have a potentially far larger audience than an academic volume available only from specialist publishers or university libraries. The Perseus Digital Library regularly reports a half a million hits a month, many times more than the number of people worldwide who consult Greek or Latin texts in the original in libraries.

Crane 1985-2007 (visitor statistics at http://www.perseus.tufts.edu/WebStats/).

An internet publication is likely to receive a larger readership than a conventionally published book in part because of the cost: most internet sites are accessible for free, and many of those that do charge have licenses to which many libraries subscribe. There is also the advantage of convenience, since a website can be consulted almost instantaneously from one's desktop computer, rather than having to go to a library, or a bookstore, or purchase from an online store and wait for the title to arrive. This also means that sites can be consulted much more casually, when a single reference is all that is needed. These issues have proved to be particularly important for scholars in, for example, former Eastern Bloc countries, whose academic libraries may not be as well stocked with obscure academic titles as they might wish. Very few libraries outside of the Western world hold copies of the Roman Society's ALA, for example, and colleagues in the East have been very grateful for the web site, which has allowed them to consult these inscriptions for the first time. (This would be even more true for titles of interest to scholars and readers in the third world, of course, with the proviso that the minimum requirement of having access to a computer and reliable internet access may itself be a barrier to many.)

Increased accessibility also means that an electronic publication may be consulted by readers who are not traditional academics, and who might never otherwise have seen books on this subject. Academics in subjects other than Classics or Byzantine Studies would probably never have walked down the aisle of the library in which sat on the shelf, but if they had an interest in, say, far eastern inscriptions, or in funerary verse across cultures, they might well stumble across the website with an Internet search engine. Electronic publication fosters not only outreach but interdisciplinarity.

Perhaps the most immediately evident difference between the content of the electronic and the paper first edition is the fact that there was virtually no limitation to the quantity of material that could be included. This was particularly important when it came to the photographs illustrating the inscriptions, their monuments, and their archaeological context. In the 1989 volume, there were forty pages of black and white plates with between two and four small images on each. The majority of texts were not illustrated, and many of those that were could not usefully be read from the photographs. In the electronic edition, in contrast, there was no publisher to impose page limits, and the cost of uploading more images is effectively zero, at least as compared to the cost of adding glossy pages to a book. As a result, most of the 250 inscriptions in the web edition are illustrated by between two and eight photographs each, with some having several dozen images including shots of the top and back of the stone, close-ups and ranging shots, pictures of the topographical context of the monument, etc. Images are both black and white and colour, with no implications on the cost, and are a mixture of digital photographs, scanned slides, and sketches or notebook entries.

Scale is also an issue for text, as publications in traditional book form include a whole host of space-saving strategies and abbreviations. There are conventions for the parsimonious expression of epigraphic and archaeological data, for bibliographical entries, for references in footnotes, and the like. Epigraphists and papyrologists have spent centuries developing the most efficient and effective conventions for indicating textual condition, damage, restorations, expansions, and so forth. All of these abbreviated forms of communication are very effective at saving space, and are easily comprehensible to the expert. To the uninitiated, however, a sequence such as:

wm, sep. fr. ?sarc. 0.35 x 0.60 x c. 0.12. T.Aph. II/III; av. 0.022

Again, once the restrictions of a page limit are removed this abbreviated text can be expanded, conventions can be glossed, descriptions of comments can be repeated where they are relevant, and cross-references can be made more self-explanatory. This is not necessarily to reject generations of scholarship and academic jargon, which is familiar to practitioners of our disciplines and serves a useful function of communication in addition to space-saving. Rather, by expanding, explaining, and illustrating these conventional sigla and abbreviations we are enhancing our scholarly publication by making it more accessible to outsiders. It is still useful to be able to see at a glance and from a single line on the screen that a given text is a white marble funerary fragment, perhaps but not certainly from a sarcophagus, with dimensions width 35 centimetres, height 60 centimetres, and depth circa 12 centimetres, found in the old Temple of Aphrodite and dated by lettering to the second or third centuries C.E. with letters an average of 2.2 cm high. But it is also useful to be able to spell out this information for the uninitiated or casual reader who would still find it of value. The publication will contain far more explicit text that could ever fit into a single-volume publication of nearly two thousand inscriptions.

As discussed above in the section on scale, an electronic publication can contain a larger selection of digital photographs, in colour as well as the conventional black and white to which a paper publication may be limited. This is not the extent of the use to which new media can be put in an online publication. Online texts may be illustrated not only with images of various sizes, resolutions, and colour depths, but with PDF documents, animated GIF images, audio and video files, Flash animations, QuickTime Virtual Reality objects, three-dimensional reconstructions, and even interactive animations like miniature video games.

Many of these multimedia possibilities would seem to be extravagant and frivolous for an epigraphic publication such as ; it is hard to imagine the value of audio or video recordings of ancient texts carved in stone. On the other hand it has repeatedly been demonstrated that interactive and animated virtual reality reconstructions of ancient monuments and archaeological sites can not only improve the presentation and dissemination of geographical and topographical data, but also enhance the research into and understanding of the sites by the archaeologists themselves.

For a single example, see Beacham 2000-2007.

The other example of electronic media being used to deliver epigraphic and archaeological data in ways that would not have been possible with traditional paper publication, is the case of plans and maps. In the project, findspots of in situ texts have been marked on a series of plans which are delivered via the web site in scalable form. These maps are layered and labelled with different kinds of findspots, and the spots themselves are hyperlinked to the respective editions of the inscriptions. Colleagues are currently working on an interface to deliver plans to the web browser via a Scalable Vector Graphics aware plug-in, which would allow dynamic zooming, scaling, and scrolling around the site to find inscriptions in their present or original locations.

Almost all of the inscriptions in the project in focus are from a single site, the city of Aphrodisias. In a wider-ranging epigraphic or archaeological project, such as the Bodel 2002-2007 or the proposed digitisation of

Now funded from 2007 and in preparation by Reynolds and colleagues at King's College London.

As well as more advanced multimedia content, perhaps the single most revolutionary aspect of publication on the World Wide Web (or to a lesser extent via other electronic media) is the concept of hypertext or hyperlinking. Both internal and external hyperlinks can transform the way a text or collection is published, and the ways in which both the author and the reader relate to other academic texts.

At the simplest level, internal hyperlinks make it easier for the reader to move from one part of the publication to another, follow cross-references, and forge multiple paths through the data provided. Whether the link is to a footnote or a pop-up glossary entry, a cross-reference to an appendix or a related inscription, a reference to a pertinent passage from a table of contents, index, concordance, or prosopography, or a thumbnail image linking to the full-size photograph, it is clearly much more user-friendly to follow a re-traceable path through a publication than to keep a thumb in each of several sections of a bound volume.

The author can also use hyperlinks and dynamic referencing to make stronger links between parts of the data, the commentary, and supporting materials. A single passage of narrative commentary, for example, can be linked simultaneously to each of the dozen or more individual inscriptions to which it refers (each of which are then linked back to the commentary), thus allowing movement in either direction from the historical discussion to the texts, or from the single text of interest up to the discussion, and possibly back down to the related texts. Alternatively, the inscriptions might all be hyperlinked to one another with a line of explanatory prose, and the shared commentary drawn dynamically into all texts via references—eliminating the need for repetition and the error-prone maintenance of multiple versions of a single piece of text. By offering all of these options in a publication, with guidance and suggestions from the author, one can allow sideways movement through a discussion and multiple entry points. Where once an academic may have written two or three different books directed at distinct audiences, a hypertext edition now allows countless audiences to create the reading experience that best suits their needs.

External hyperlinking is a potentially even more powerful tool. References to journals or articles that have electronic versions—such as those published by the , but also online versions of traditional journals such as the —can be hyperlinked directly to the publication, or possibly even the section or paragraph of the publication in question. This external hyperlinking is not only a matter of convenience, saving a reader time and effort, and reducing the need to go to the library to check each and every reference. It also supplies the reader of the electronic text with more direct access to the checking mechanisms on which replicable academic research depends; when I cite Smith 1999 as saying that a certain object is formally of the early Imperial period, the reader may want to check that this is in fact what Smith claimed, and on what grounds. Hyperlinking of references thus makes the electronic publication more transparent and more verifiable.

The final aspect of hyperlinking which enriches a digital publication is the concept of dynamic linking. A project that uses open standards to create electronic resources, as adheres to the usage of EpiDoc and the TEI, is able to make available source code and texts in such a way that they can be repurposed by other projects that are using (or are aware of) the same standards. In our web publication, the EpiDoc XML files of all two thousand inscriptions are available both in a single downloadable ZIP file (including the EpiDoc DTD), and as individual downloads linked from the bottom of each page (and with transparent and predictable URLs for dynamic linking). The files, licensed under a Creative Commons attribution license, may be downloaded en masse and incorporated into a larger project; an epigraphic corpus, say, or a more general collection of Greek and Latin texts, or a specialist corpus of texts relating to the provincial Roman economy. Such a project would then run these XML files through their own XSLT stylesheets or process, publish the inscriptions as part of their larger collection, index by XML elements or character strings (not necessarily restricted to the indices we have envisaged), or make the files searchable within a web or database search engine. Perhaps most interestingly, although I do not know of anybody who has done this with EpiDoc XML documents yet, the online XML files available on the InsAph server, or a specific subset of them, can be automatically read by a web service and rendered via stylesheets or other processes into a dynamically served web page, text excerpt, or index by another project.

This dynamic linking or live hypertext data sharing is possibly the most powerful potential of electronic publication, and it relies very heavily on the use of open standards by projects that hope to have any kind of relationship with each other. In addition to the use of robust technologies such as XML, markup guidelines like the TEI and EpiDoc, there are interchange and registry protocols such as the Canonical Text Services, both in progress and available, which make such data sharing far more feasible.

One of the most obvious advantages of an online website or database is that the data made available therein can be kept constantly up-to-date and relevant. An encyclopaedic site such as the Wikipedia or a corpus such as the Epigraphische Datenbank Heidelberg is not a fossilised print publication that appears once, sits on a library shelf and never changes (until a second edition may appear eventually with updates and corrections). Rather, when a revision, correction, addendum, or any new data need to be added to the database, it can be so added at any time and is available to the world almost instantaneously (albeit with some degree of editorial or review process in most cases). There is a sense in which these works are perpetually "in progress".

This can be quite a burden on the author or editor of such a site. The author of an academic work, who is often a working lecturer with other responsibilities, occasional research leave, and a range of research interests, has traditionally spent a year or two writing a book and then, upon its publication, moved on to another project. In the current academic model, most projects are not permanently funded and staffed, and the academics involved cannot afford the time to receive and acknowledge corrections, incorporate new scholarship, and update the site continuously. There are projects, so-called "living databases", which have such constantly developing status, but most projects still do not have this model.

A more significant problem arises when one considers the status of a work that has to interact with, and be reviewed within, the world of peer reviewed, cited, traceable, and replicable scholarship. If a later work of secondary scholarship cites an online project such as , perhaps arguing with an interpretation of the evidence or adding new data to the historical debate, then that citation needs to be followed back to the source by a reader. If the editor of the cited project reads this new article, agrees with the arguments, spots an error, or otherwise sees a need for change, and simply updates the original site, then the reader following the reference back from the secondary source will no longer see the text upon which the scholar was commenting. This is clearly an unacceptable state of affairs, even if the original text can be recovered from an Internet archive or cache in some form. Now it may be possible to cite a dated, frozen version of the site, as is the case with Wikipedia articles, for example, but the protocols for citing thusly are not common in traditional scholarly publication and referencing.

It is for a combination of these reasons, the burden on the editor and the need for stable and citable publication, combined with the current model for finite-term funding of research projects, that the project chose to make the online versions of Roueche 2004 and Reynolds 2007 traditional, one-off publications. This decision meant that if there are additions, emendations, corrections, or updates, these will have to wait for a second edition. (Minor corrections, on the level of typographical errors or broken hyperlinks, can be fixed, but they will be recorded in the project RSS feed.) There may be a more ideal solution to these problems, and we would welcome discussion of this point.

In the five categories above I have so far talked about some of the repercussions of delivering a project outcome via digital or online publication. In this final section I want to mention some aspects of a different kind of implication of digital projects: the consequences of performing digital research as well as electronic publication. I have already discussed the fact that in the project we committed to making available the XML source files which lie behind the web pages that make up the publication, and the fact that these files can be dynamically linked and repurposed by other projects and applications. The first repercussion of the availability of these source files is to enhance the level to which the research is replicable by other scholars. As well as being able to return to the primary data and the history of scholarship upon which our research is built in order to examine and verify the conclusions we draw, a scholar with access to source code can also verify and replicate the digital processes we have used to enhance this research. This allows the verification of statistical processes, the checking of both markup and generated indexes and tables, and, as mentioned above, the running of new processes, searches, or algorithms on the data. The inclusion of these possibilities, and the large number of photographs and other visual representations of the primary data, opens up the author of such a digital project to a new level of scrutiny, which may on the one hand cause some to feel nervous, but can only be good news for scholarship.

Even more central to the research process, however, is the fact that a true digital project is not merely the result of traditional classical research that is at the last minute converted to electronic form and made available online. Rather the XML files (in the case of and other EpiDoc projects, other data models for other types of project) that lie behind the publication, are the direct result of, and primary tools for, the academic research itself. These files contain the marked-up data, Greek or Latin texts, descriptions, editorial commentary and argumentation, references and metadata, all in machine-readable and actionable form. It is this single, structured collection of source data which is taken by the machine process and run through a series of XSLT stylesheets which generate, in turn, the web presentations of individual or groups of inscriptions, the contextual tables of contents, indices, concordances, prosopographical and onomastic tables, and so forth.

Furthermore, because the XML files and XSLT stylesheets have been developed and used from a very early stage in the project, these indices, concordances, and lists have been available to the researchers as a tool both for checking results, verifying the consistency of markup, and calling attention to patterns, unexpected gaps, or irregularities in the data. Because every word in the corpus, for example, may be tagged as soon as the inscription is transcribed, and the lexicographical index of lemmatised forms automatically generated from these tags, the researcher can look at the index (or other structured representations of this data) for comparisons, exceptions, or illuminating parallels in the corpus. While this may seem to be more work than an editor is accustomed to doing at this point, it should be noted that the work would have to be done eventually, usually at a late stage in the project, and that the automated indexing both saves time later on, and makes the indices more reliable and verifiable. The research process, as always to some extent, is iterative and not linear.