Morton Paley's piece, "The Devil's Walk and The Devil's Thoughts," gives a clear account of the relationship of Shelley's satire to its Southey-Coleridgean model, so clear and so convincing that it leaves his respondent little to say. So, having little to say in direct response, I will instead comment upon the electronic edition, assembled by Neil Fraistat and Donald Reiman, which Paley depends upon so heavily. My comments will have little to do with the quality of Fraistat and Reiman's editing (although at least one typo may remain in their reading texts); rather, I want to talk about the electronic medium itself, in particular on-line, web-based scholarly editions, because I believe them to be, on balance, something of a mistake.
Fraistat and Reiman's edition brings together a wide range of texts and contexts related to The Devil's Walk: these include full texts of both the letter and broadside versions of Shelley's poem, scanned images of the printed broadside (an extremely rare item), full texts of Coleridge and Southey's The Devil Thoughts and its later revised versions (where the title changes to The Devil's Walk), Byron's The Devil's Drive, and a section on "Other Romantic Devils" (They are missing, however, the 1830 version of the poem, edited by H. W. Montagu and ascribed to Richard Porson, with 6 illustrations by Cruikshank, an odd omission, given the multimedia capabilities of the World-wide Web.). In addition, they provide the standard editorial fare of apparatuses, editorial headnotes, descriptions of copy texts, and the like. All this is very helpful, and has certainly never been done for this poem before: their edition clearly demonstrates how the electronic medium allows us to provide a wealth of editorial material, even for relatively obscure texts like this one, and show that they are far more interesting than we may have thought.
But Fraistat and Reiman's efforts have some drawbacks, from which I hope we can learn. What they present here is, essentially, a more-complete-than-normal annotated edition: that is, the only real difference between what they provide and what a book provides is that normally publishers do not allow us to include this much material about this kind of text. If the editors were independently wealthy and could afford a vanity press edition of their poem, there would be little difference in what they include, and in fact the book version would be considerably better. It would be easier to read: I don't believe that I'm the only one who prefers printing out on-line texts to reading them off a computer screen. It would be easier to negotiate and use: variants, for instance, could be placed at page bottom, rather than in a separate file, and readers could easily compare the variants with the reading texts, something which is difficult in the on-line version. Book format would even allow extensive use of parallel, facing-page texts, and the editors would not have to worry about setting up frames in HTML, or whether the reader's browser supports frames or not. And a standard table of contents and index is a superior way to describe the structure of an edition than the rather clumsy HTML table Fraistat and Reiman currently supply. My point is this: if we conceive of our electronic editions as merely expanded versions of traditional books, we end up with inferior products. Books do book-like things better.
The question, then, and the great editorial challenge, is how to prepare texts for the electronic medium in such a way as to exploit those things that computers do better than books, and differently. Computers, as we know, have the power to search, compare, and analyze, with amazing rapidity, tremendous volumes of data, in ways that just a few years ago would have been unthinkable. But to exploit this power we must move beyond the limitations of HTML markup and employ a tagging system that allows us to ask new kinds of questions of our texts and yield interesting results. I am referring, of course, to richer implementations of SGML, such as the Text-Encoding Initiative and whatever its descendants may be in the future.
When SGML was invented over a decade ago, its function was to allow us to describe the structural features of a document in such a way that any computer, running any operating system, could understand it, store it, and search it rapidly. Markup, the SGML people told us, must be descriptive: it should describe the actual structural features that any given text might have, and not concern itself with producing visual effects in software browsers. In the case of a poem like The Devil's Walk, such features might include stanzas, with meter and rhyme scheme appropriately indicated, individual lines within the stanzas, metrical patterns with the poetic line, even individual metrical feet. In the TEI scheme, for instance, the first stanza of Shelley's poem might be tagged like this:
<lg type=3D"stanza" rhyme=3D"abab"> <l n=3D"1">ONCE, early in the morning,</l> <l n=3D"2"=rend=3D"indent"><name>Beelzebub</name>=arose,</l> <l n=3D"3">With care his sweet person=adorning,</l> <l n=3D"4" rend=3D"indent">He put on his Sunday=clothes.</l></lg>In this example, <lg> is the tag for "line group" (to be used for both stanzas and verse paragraphs), < l> is the tag for "line," and the other entries are ways of more fully describing the structural and layout features of the stanza. With the full text so tagged, with additional tags added to describe meter, and with a relatively simple search engine, users could investigate a variety of the structural features of Shelley's poem, compare them with similar features in his model, and do so at lightning speed. In the matter of a few seconds, we could perform analytical operations on the text that would have taken hours (and, in longer texts, weeks and even years) B. C. (Before Computers). This is the kind of thing that computers do well: they manage and analyze large volumes of data quickly, if they have first been told what to look for.
By contrast, the markup of the first stanza of the HTML edition looks like this:
<TABLE BORDER=3D"0" cellspacing=3D7> <TR> <TD COLSTART=3D"1" ALIGN=3DRIGHT VALIGN=3DTOP= WIDTH=3D"25"><a= href=3D"variantsb.html#B1">01</a></TD> <TD COLSTART=3D"2" ALIGN=3DLEFT= WIDTH=3D"550"> &nbs= p; ONCE, early in the morning,</TD></TR> <TR><TD COLSTART=3D"1" ALIGN=3DRIGHT VALIGN=3DTOP= WIDTH=3D"25"><a= href=3D"variantsb.html#B2">02</a></TD> <TD COLSTART=3D"2" ALIGN=3DLEFT= WIDTH=3D"550"> &nbs= p; <a href=3D"brnotes.html#B2">Beelzebub</a> arose,= </TD></TR> <TR><TD COLSTART=3D"1" ALIGN=3DRIGHT VALIGN=3DTOP= WIDTH=3D"25">03</TD> <TD COLSTART=3D"2" ALIGN=3DLEFT= WIDTH=3D"550"> &nbs= p; With care his sweet person adorning,</TD></TR> <TR><TD COLSTART=3D"1" ALIGN=3DRIGHT VALIGN=3DTOP= WIDTH=3D"25">04</TD> <TD COLSTART=3D"2" ALIGN=3DLEFT= WIDTH=3D"550"> &nbs= p; He put on his Sunday clothes.</TD></TR>
Now this complicated set of tags does not intelligibly describe the structure of Shelley's poem: rather, Reiman and Fraistat are trying to produce layout effects in a Netscape browser, using the HTML tags for tables. They are using these tags to work around the fact that the Netscape browser (or any HTML browser, for that matter) does not represent blank space well at all. So, in order to include properly-spaced line numbers with their text (and here "properly-spaced" means as we are used to seeing them in a book), and to indent the second and fourth lines of the stanza, they must create a double column table (one column for numbers and the other for text), and insert a large number of "dummy" entities ( , the SGML entity for an en space) in order to get lines of text to line up right. The end result is a fairly nice looking page, for an HTML representation of letterpress layout, but we all know that letterpress looks better. Moreover, Fraistat and Reiman's complex work-around, because it is browser-oriented, is not likely to stand up very well to changes in software or future extensions of the HTML DTD. The whole edition will likely have to be retagged all too soon.
Fraistat and Reiman, as they tell us, mean their on-line edition only as an experiment, and as a kind of foretaste of their letterpress edition. They also tell us that they are preparing a version employing a richer implementation of SGML than HTML 3.2 (I assume that their intention is to use one of the DTDs of the Text-Encoding Initiative). So my remarks here are not a specific criticism of their work, but a more general caution about the current rush to the web. I do not believe that computers can outdo books on their own territory, I am thoroughly unconvinced that the hypertext link is the most interesting thing a computer can do with texts, and I am certain that the ruckus about the virtues of nonlinearity, however culturally interesting it may be for the moment, will fade away and be quickly forgotten. But I am just as certain that the computer will thoroughly reshape our analytical procedures, once we begin asking computers to do things with texts and the books that contain them that otherwise cannot be done.
--BG, OCTOBER 1997