Many Outputs - Many Inputs: XML for Publishers and E-book Designers

1 Introduction

Newspapers and periodicals have been online for years. In e-learning the amount of net-material is growing. E-books are slowly spreading and the reading of e-books and other material from handheld devices has only just started. Publishers around the world are struggling to manage the transition from being bare paper book producers to be multiple content producers selling a variety of products in many different channels.

In this situation, where publishers are seeking access to the new network economy, many actors promote XML using the doctrine of "one input - many outputs". In the realm of publishing the doctrine claims that from one book, or one document, you can produce multiple formats and a variety of end-products.

In this essay I shall question the doctrine of one input - many outputs. Supported by theoretical insights of linguistics and semantics and the practical experiences of writers and publishers, I shall even claim that the doctrine is basically wrong. I will in the essay argue that the doctrine is a result of what could rhetorically be called a false analogy. This means that concepts from one context on erroneous grounds are used to argue on the state of affairs in a different context, obviously resulting in wrong conclusions.

The concepts of content and content structure in XML are not the same as the concepts of content and content structure in publishing. The separation of content from presentation in XML documents has no parallel in paper books and e-books, where obviously the presentation is inseparably a part of the content.

The content-presentation separation is the logical premise for the doctrine of one input - many outputs. To be moulded, shaped and presented at will, the content must in some way be pure and independent of presentation in the first place. So long as, in publications and publishing, this premise is not valid, neither is the conclusion.

Without arguing the powers of XML, I will in this essay warn against computer scientists and conversion house salesmen telling publishers that XML will bring them into a relatively uncomplicated era of one book - many formats. XML will not change the fact that new electronic media give rise to new genres, new presentational principles and new distribution channels. XML will rather add to the diversity of publishing and lead publishers into a challenging state of "many outputs - many inputs".

2 XML Doctrine of One Input - Many Outputs

As part of Open eBook Forum, the object of the OEB Publication Structure Working Group (OEBPSWG) is to develop platform-independent e-book standards.[1] OEBF has recommended two Document Type Definitions (DTDs), one describing the publication structure of e-books, the other the structuring of metadata and package information. The DTDs are based on XML and the current basic oeb-document DTD (OEBPS 1.0.1) can best be described as a simplification of XHTML, designed to describe a common presentation structure for electronic books.[2]

On the independent Web site "eBookWeb", Dorthea Salo, a member of the OEBPSWG, publishes articles trying to educate book publishers and typesetters in the use of OEBPS, the advantages of XML and the notion of structure.[3]

To avoid the word 'structure' getting "lost in the shuffle", Salo defines the concepts of structured data and structured markup.[4] These concepts are often used when experts discuss e-books and p-books and the use of XML in book production processes. To "break down a text into pieces, name the pieces and determine how they fit together" is, she says, what markup experts "think of as 'structuring' text".

After this initial definition Salo goes on describing how print production programs like Microsoft Word and Quark XPress structure text in a flat, two-level structure, whereas XML structures text in a logical and hierarchical way. The problem with this comparison and her conclusion that "typesetters should experiment with XML for themselves" is not the description per se but its implicit assumptions of what great achievements publishers will gain by turning to XML.

These implicit assumptions are explicitly expressed in another of Salo's articles in which she argues on the necessity of using XML for publishers who want to join the "era of same-book-different-formats".[5] "Know what the pieces of a text are and how they are put together, and then you may manipulate and design the text as you choose" Salo proclaims, clearly supporting the highly arguable XML doctrine of "one input - many outputs", which in the area of book production means that from one document or book publishers can produce many different outputs.[6]

The multiple uses claim of the one input - many outputs doctrine is widespread among experts and commentators on XML and SGML. XML is developed from SGML and (like SGML) is a platform-independent metalanguage describing markup languages. XML is stricter and more formal than SGML, but it is easier to learn. XML is specially designed to utilize the possibilities of the Web, improving on the shortcomings of HTML. Many of the arguments used in promoting SGML are now used in promoting XML.

In one fashion or another the doctrine of one input - many outputs is found in practically all books introducing XML. Pitts' book includes a chapter simply called "One Document, Many Output Options".[7] In an example, DuCharme manages to get 12 publications from the same set of data or elements: a complete reference work, a tutorial and a quick reference card, each of these in paper, compact disc, handheld computer and Web versions.[8] A similar example is put forward by Walsh and Muellner, whose introduction says an article, once marked up according to the DocBook DTD (which has both a SGML and an XML version), can be published on the Web, in different paper formats, in Braille and in audio.[9]

DuCharme's example is based on information on a car's motor engine used throughout organisations that produce, sell and repair cars. Walsh and Muellner use a white paper as an example. Obviously it is easy to find examples where the one input - many outputs doctrine makes sense, like in corporate and government information flow (DuCharme) and in repetitive publication of strictly structured material (Walsh and Muellner). In such cases the multiple uses claim is relevant and the use of XML correspondingly fruitful.

What XML experts seldom mention are the limitations of the doctrine, as in important branches of publishing. Modern full-scale publishing houses deal with a diversity of texts in all genres and formats, both fictional and non-fictional. In this area of publishing the range of the doctrine is not only limited, but very often the claim of the doctrine is basically wrong. This is rarely communicated, not even by Walsh and Muellner in their Definitive Guide. Even if the main focus of DocBook is documents made for software and hardware documentation, DocBook also deals with markup of articles and books and aims at making available a DTD that covers many genres. A moderating remark on the range of the example would not have been inappropriate.

Kasdorf, who in an article on SGML and PDF explicitly deals with book and journal publishing, similarly seems to adopt the doctrine of one input - many outputs.[10] Later this article was extended and turned into a white paper in which the same arguments Kasdorf had earlier used in regard to SGML are used in favour of XML.[11] The paper gives balanced and convincing arguments for needing both PDF- and XML-based work flows in modern (electronic) publishing, which of course strongly indicates that XML does not solve all problems. Even if the paper emphasizes the complexity publishers meet as they move into electronic publishing, it seems to slightly overestimate the advantages of XML, stating: "When publishers need to publish in more than one medium - typically print, CD-ROM, and Web - XML almost always dramatically reduces the cost of producing subsequent versions, whether they are simply converted, or modified and augmented to take advantage of the power of the various media". In a summarized form it says: "The richness and power - and complexity - of XML comes from its ability to produce print and electronic products from a single coded file". These citations say almost the same as the doctrine of one input - many outputs.

In a seminar on e-books aimed at publishers in Norway one of the key speakers, Asplund, did not have the same balanced view on XML as Kasdorf, neither in the presentation nor in the printed documentation.[12] In the seminar publishers were presented the model in Figure 1 describing an XML workflow in a publishing organisation (here translated into English).

As presented, the interpretation of the model is both simple and appealing: the content of different text formats can be converted into XML, and from the XML file(s) content can be reformatted, encrypted and made into a variety of new formats and publications, i.e. one input - many outputs.[13]

The greatest weakness of this model is that it only tells part of the story. In real life the situation is far more complicated. The model suppresses the fact that among the huge variety of texts dealt with in a publishing house, most texts are made for a special use in a specific context for a limited group of readers. This means that there is little use of those texts in other contexts.

The model also suppresses the fact that most publications are made in close collaboration with designers of the output, whether they be graphic designers or Web designers. These collaborations have important aesthetic and communicative consequenses which affect both the structuring of the texts and the numerous finishing details of the texts. The specific design principles of different media often give characteristic texts, which mean one cannot use these texts in just any environment without changing them.

A slightly extended interpretation of the model would presume that the converted content is stored in an XML repository or database. With the use of formatting and transformational applications (like XSL and XSLT) bits and pieces of texts can, according to this interpretation, be shaped and transformed into new texts and formats and used in all kinds of media. Or as Salo says: "Know what the pieces of a text are and how they are put together, and then you may manipulate and design the text as you choose".[14]

However appealing it might be - and despite the fact that it describes some cases - this latter understanding of the model is even more remote from the realities of publishing. Exactly the same arguments of simplification and media-specific texts can be used against this interpretation, even if you consider the transformational capabilities.

The extended model presupposes a clear distinction between content and presentation, an assumption you have to make to be able to format and transform the content at your choosing. The concept of such a distinction is, on the other hand, contradicted by both the insight of semiotics and the experience of writers and editors. There simply does not exist any sign-system with a clear-cut distinction between content and presentation.

The extended interpretation of the model also overlooks the fact that a piece of text is always part of a greater textual and communicative context. Both the creation and interpretation of a piece of text is heavily dependent on what kind of text or genre it is part of, how it is presented and what it is used for. Give a piece of text a new wrapping and put it in a new context and that text no longer has the same meaning.

It is notable how many XML experts uncritically relate to the doctrine of one input - many outputs. It is especially suprising to see that members of the Open eBook Forum seem to do so. In what follows I shall deal with these matters in closer detail, starting with Salo's definition of structure and assumption that one can manipulate a text as one chooses.

3 Many Outputs - Many Inputs

The way Salo defines structure, as a system of elements and their relations, is hardly controversial in itself. To regard structuring as a process of breaking texts down into pieces is common, but from a producer's or an author's perspective structure can also be seen the other way around: to structure a text is for an author to put pieces together and build up a text, that is, to write and arrange a text. The way an author writes and arranges - or structures - a text depends on the subject, the readers (age, gender, professional level, etc.) and on the communicative goals of the text. Further, the writing and organising of the text is done according to more or less predefined conventional patterns or genre norms which, based on centuries of experience, have been developed to fulfil specific communicative objectives.[15]

An intimate knowledge of genre norms is a central part of the know-how of editors and writers. When authors write and editors edit they do so with certain patterns, or genre norms, in their heads. These genre norms define and heavily influence the way texts are structured at all levels, both at macro and micro levels of the text. At macro levels genre norms define the overall order and organisation of themes and issues into chapters and paragraphs, the use of illustrations, frames, notes and indices. At middle levels genre norms prescribe ways to dramatize, describe, make an argument or tell a story, and at micro levels authors, depending on genre, use different words, metaphors, expressions and technical terms creating all kinds of language styles.[16]

In other words, a text is a whole where all the pieces fit together. This does not mean that you cannot "break down a text into pieces and determine how they fit together", as Salo puts it, or, for that matter, develop word processing programs that markup texts in logical and hierarchical ways for later formatting. What it does mean is that these text pieces or elements cannot be rearranged and reused in an automatic creation of new texts belonging to other genres. Such reuse contradicts all linguistic knowledge.

First of all, texts belonging to different genres not only have different structures, but also qualitatively different parts, or element types, which means that you cannot use the same element wherever you like. Different genres also use different language styles, and mingling of styles is seldom successful. A specific text is further a texture of syntactic constructs and linguistic tokens used to make the text coherent and to make the transition between sentences or paragraphs intelligible and smooth.[17] Trying to rearrange and reuse the elements will not only distort the logical order of themes and issues, it will also leave meaningless clues and linguistic signs pointing in all directions. Such reuse would be like taking a pair of scissors, cutting up a tapestry weaving, rearranging it, and hoping to create a nice new weaving where all the threads are still connected.

In the context of rich book production this means that there is no easy way to "manipulate and design the text as you choose", putting parts and pieces together, creating new texts to be used in other media than they were created for. Publishers cannot take a chapter from a printed textbook and use it as an article in an online encyclopaedia without reediting and rewriting it. And neither the textbook part nor the encyclopaedia entry can be used in a WAP environment. To be used in an e-learning curriculum the original text must also be adjusted to the overall pedagogical idea of the online course. And all of this is exactly the same the other way around: publishers cannot take the documents of a highly developed hypertext or an e-book, send it to a printer and say: "Make a paper copy!"

Among photographers it is obvious that you cannot use the same pictures or digital picture files to produce posters, printed book illustrations, pictures to be watched on PC monitors or illustrations meant for Pocket-PC displays. But in the same way as photographers and graphic artists must carefully prepare resolution, colours and motif for the specific uses of pictures, writers and editors must carefully prepare the wording, linguistic style and overall structure to comply with the communicative objectives and specific media uses of texts.

So, with the possible exceptions of some typographically uncomplicated genres, for publishers there is no such thing as "one input - many outputs" or "same book - different formats". On the contrary, as writers and publishers move into the network economy, with its diversity of publication forms and distribution channels, they will face far more complicated production processes, where texts and books in most cases will have to be produced in quite a lot of editions.[18] Both writers, editors and publishers will in the future inevitably have to submit to the fact that many outputs require many inputs.

4 Illusion of a Content Structure and Format Separation

The doctrine of "one input - many outputs" is built on a widespread conception among XML-experts, the belief that in texts you can separate content structure from format, or, when talking of books, the assumption that you can separate structure of content from typography.

This understanding is probably tied up with statements like the one by Pardi saying "the power of XML is its ability to separate the user interface from the data".[19] Or with Harold's claim "XML markup describes a document's structure and meaning. It does not describe the formatting of the elements of the page", which in the short version reads: "XML describes structure and semantics, not formatting".[20] Kasdorf even says about SGML, the basis of XML: "SGML is all about structure and meaning, and has little or nothing to do with appearance."[21]

Similar statements are frequently made in DocBook: The Definitive Guide (Walsh and Muellner), and even in OEBPS 1.0.1 there are remarks on an intended separation of content from presentation. These statements may make good sense in some circumstances, such as in pure XML environments. In the context of books and e-books they are misleading, to say the least.

XML-experts define content narrowly as character data contained in an element (or as character data and (or) child elements contained in an element). Uses and combinations of child elements and character data within an element, as defined in DTDs, is called the element type declaration's content model or the element's content structure.

In semantic markup or content-oriented markup, elements describe, often using humanly understandable element type names, what the content of the elements are, rather then what the content should look like. The latter, called presentational markup, is frequently used in HTML and OEB. Structural markup, as XML-experts define it, is used to build logical, hierarchical structures in XML documents. The totality of elements in an XML document is often called its content structure or simply its content.

These are specialised ways of using the terms semantics, structure, content and content structure. When these mechanical and strictly logical meanings of the concepts are used in more general accounts of books and e-books and mixed up with humanist and publicist meanings of the same terminology, naturally a lot of misunderstandings and wrong conclusions are unavoidable consequences.

Things get especially confused when the content-presentation separation of XML is assumed to apply to books or other publications. On such an assumption it is easy to mix up things in saying that markup of books should only consist of content-oriented and structural markup and no presentational markup. This way of arguing misses the point that one of the essential features of books and e-books is that content structure and presentation are fundamentally interrelated. In the context of publications there really is no way of separating content structures and presentation, or for that matter; content and format.

Salo is the first to admit this: "Some XML experts will consider this heresy, but I speak as I find" she says. Under the heading "Structure Breaking Down" she claims that "hierarchical structures as expressed in XML do not always represent text perfectly" and that "not all visual distinctions in publishing are easily described in terms of a hierarchical structure". Salo even says that "some clear and obvious structures are difficult to articulate in XML".[22]

As a "canonical example" of the latter, Salo points to tables, which is an illustrative example. Using columns and rows, the cells of tables can at the same time show data values of two different variables. This information is literarily "shown" in a visual, two-dimensional way and the strength of table layout is that readers can easily compare values within different cells. Meanings in tables are expressed by the combination of data (words and figures) and visual layout.

There is no problem in marking up tables in XML, but the whole meaning conveyed by a table will not reveal itself before the elements are formatted and presented in a visual, two-dimensional way, either on paper or on screen. This shows that the meanings of the tables, their contents, bear on visual representation and cannot fully be captured by the structural logic of XML, irrespective of its hierarchical character. Which again proves Salo right, that some obvious structures and visual distinctions are hard, not to say impossible, to express in XML. The example of tables is a clear illustration of the fact that in visual-based publications there is no such thing as a sharp distinction between content and formatting.

Salo also points to other examples, like markup of italic and small caps, illustrating the same point, but in totally different ways. Whereas XML fails to articulate the semantic of tables, the use of XML in the case of small caps and italic letters forces designers to markup parts of text with little (or none whatsoever) relation to the content structure of the text. Often, inline formatting - like italic, bold, semi bold, small caps and lowercase figures - are used for pure aesthetic or typographic reasons or to improve the legibility of texts. These pieces of text still have to be marked up and made into elements and part of an overall structure (if you are an XML purist you must pretend these elements are parts of some kind of content structure).

This does not mean that marking up these parts of texts is meaningless. On the contrary, the ability to manipulate the formatting of smaller, inline parts of a text using style sheets is a useful XML property. But the example again shows that XML markup is not only about hierarchical content structures, it is also about typographical formatting structures. Often there is no way of separating the two.

5 What This Thing Called Structure is Not

The example of italic (and bold and other font features) is complex. In many cases the use of italic (or underlining or font colour) is not primarily aesthetic or typographical, but essentially semantic. Italic is used to emphasise names, qualities or concepts, to space out quotes, figure texts or certain statements, or to stress important passages in the text. In these examples it is perfectly reasonable to say that the marked up elements are parts of the content structure or semantic structure of a text. As opposed to the semantically close-to-empty <i> or <em> tags of HTML and OEB, XML lets you distinguish between as many semantic or content categories as you like at this level. It is still, though, the visual appearance, the italicness, that makes the reader aware of the extraordinary semantic value of the specifically formatted piece of text.

Salo treats the examples of tables and italics as exceptions, showing that the distinction between text structure and format is not always that clear. What she passes over in treating these examples as pure exceptions is that the examples also illustrate the very important general point that the distinction between structure and format is never that clear.

Before she treats the exceptions, Salo uses a simple text to illustrate the way MS Word and Quark XPress structure text in a two-level structure, whereas HTML and especially XML structure the same text in a hierarchical way - a way, she says, that is much more applicable and in tune with the demands of digital publishing.

Salo's example, which is composed of a title, a subtitle and some paragraphs in Latin, is, without Salo mentioning it, a very good illustration of how typography more often than semantics is the dominant feature when we determine how to markup text. And this is irrespective of whether we use a two-level or a hierarchical element structure. Here we shall use another example to illustrate exactly the same point as Salo passes over.

Except for the visual (!) signal or typographical peculiarity of a missing single stop, that might indicate a title, there is nothing in the following string of characters and words that indicate where one text element ends and the other starts, not even for readers familiar with the language used, which is Norwegian:

\0xC5 seile inn i fremtiden Livet i en seilb\0xE5t eller rob\0xE5t gir folk anledning til \0xE5 gjenerobre den sakte tiden, den som i v\0xE5re dager i v\0xE5re dager er i ferd med \0xE5 bli en mangelvare. Forbundet KYSTEN har formulert som vesentlige m\0xE5lsetninger \0xE5 gi vern til kystkulturen, ta vare p\0xE5 det som var i ferd med \0xE5 g\0xE5 tapt, i tillegg til \0xE5 styrke v\0xE5r identitet som kystfolk. Denne fortidsorienteringen har sine kritikere. B\0xE5de blant ekstrem-urbanistene som Erling Fossen og blant samfunnsforskere har man sett tradisjonsorienteringen som nostalgiske klynk etter en svunnen tid.

Even for Norwegians there are no semantic clues as to what is what in the text above; for the reader it could all be part of a much longer paragraph. But put some typographic information into it and everybody will know what parts of the text are what:

You don't have to understand Norwegian to see what is the title in this text and what seems to be some kind of resume or introduction. Without understanding any of the semantics, people around the world familiar with HTML or XML can easily pick out the elements and markup the text using <h1>, <p> and <em> tags of HTML and OEB or XML tags like <title>, <resume> and <paragraph>. We all know that titles are larger and usually placed above the rest of the text, that the first paragraph of the ordinary text in a chapter has no indentation, and that the text above it, in italic, accordingly, must be some kind of introduction. In this example the markup of the text can be based solely on visual appearance, that is, on typographic formatting or data interface. The example shows that structure and display are closely interwoven and in many cases indistinguishable.

What is even more important in this example is that typography affects the way we perceive and comprehend a text. For Norwegian readers the marking-up of the text above adds extra meaning to the original string of words. When a part of the text is revealed as a resume readers will interpret the piece of text accordingly, first because a cultural understanding of the meaning of a resume is brought upon the specific text string, and second because the piece of text, as a result, will relate to the rest of the text in a new and different way. A text is more that the sum of its parts; it is a coherent whole where the meanings of the parts are dependent of their position relative to the whole text. A part of a text can therefore be fully understood only when placed in a greater context, and typography, as it is, helps us do that.

Nor for writers or editors is typography merely a question of aesthetics or readability. For the writer typography is first of all a way of organising a material. For the author, content structure and text appearance are mutually dependent qualities. Knowing the genre and its visual presentations, the writer organises the subject matter accordingly, into an abstract, titles, an introduction, chapters and paragraphs, all of which at the same time are semantic and typographic structures.

XML markup experts are of course also part of a long typographical tradition. An expert rarely defines element types and marks up books trying to distinguish parts of narrative progress or the different rhetorical phases of an argument. Even when defining a logical content structure, the markup expert uses the same typographical elements that have always been used.

As Kasdorf points out, "print pages have been communicating structure for centuries - visually".[23] Markup of manuscripts being prepared for print has been an essential part of the activity of publishers and typographers for hundreds of years.[24] Most text elements used today (and markup used in desktop publishing programs or HTML and XML editors) are historically and culturally generated, and many are of approximately the same age!

Some typographical and semantic elements, like chapter and paragraph (originally an index sign: \0xB6 ) date back to the medieval times of handwritten manuscripts. The term 'block', as in text blocks and block elements (like in the DTDs of HTML and OEBPS), dates back to the earliest print era when text and illustrations were carved out of wooden blocks and printed. In addition to these old instances, a surprisingly high number of familiar typographic elements date back to the time immediately after Gutenberg's invention of movable letter types. Regular provision of titles and title pages, use of colophons, tables, indices and the use of section breaks, all became widespread and common in printed books in the first four or five decades after Gutenberg's invention. Innovations like graduated types, running heads, footnotes and (Arabic number) pagination were also made in this very early era of print, which make all these typographic features about 500 years old.[25]

Since then innumerable genres have developed combining text elements in many ways, using all kinds of language styles and text types (narratives, arguments and descriptions). But in spite of this variation, the basic text divisions are mostly the same as five centuries ago, and these early typographical inventions have made enduring impressions on the way we work and think.[26] Typographical structures have particularly affected the way we organise subject matters and structure their mediation, both in fictional and non-fictional genres and in print and electronic media.

In other words, there is no way the content structure of texts escapes the imprint of typograhy. The example above is also meant to show this: in publications there is no clear distinction between content and appearance.

Consequently, the narrowly-defined XML concepts of content and a content-format separation should not be mixed up with the (apparently) parallel concepts in the area of books and texts and e-books (where these concepts obviously mean different things). Such a mix-up easily leads to the idea of text content as a presentation-independent, pure and malleable substance that can be shaped at wish in some "one input - many outputs" kind of production. This is not the case in a print environment, as shown above, and it is not the case in an electronic environment, as a closer examination of XML will confirm.

6 Element Structure and Formatting

The example above, the historical account, the discussion of italic and tables and the reference to linguistic and semantic knowledge, all clearly show that typography has semantic significance. Typography organises and structures content in such a fundamental way that one cannot differentiate between content structure and appearance in a text or in a publication, especially not when marking up the text.

From the point of view of XML, there really is no reason why we should differentiate these features! XML is an extensible and generic markup language. This means that its elements can be marked up based on content, on appearance, on structure, or on all three! In the context of books and e-books the markup of text is obviously based on all three, and often at the same time.

This means that marking up a piece of text as a paragraph, for example, is sometimes based on semantic criteria, as when a new theme occurs or a new line of argument starts. At other times legibility or visual criteria will dominate, as when a new paragraph gives the reader a rest or makes the page look nice. Usually, as repeatedly pointed out, the markup is done with both semantic and typographical criteria in mind. The interrelation between content and display is also to some degree present in most of the important and seemingly pure structural markup; the use of container and group element types like <chapter>, <section>, <text> and <list>, the objective of which is to create a logical and hierarchical structure in the XML document. Finally, as shown in the discussion of italic and small caps, some markup is for purely semantic reasons or for purely aesthetic reasons.

XML really doesn't care what reasons are used in text markup, whether it be structural, semantic or presentational criteria, separate or in combination, two or three! In XHTML 1.1[27] and OEBPS 1.0.1[28], which both are XML-compliant; the markup is based on structural, semantic as well as presentational criteria. As long as the syntactic rules and naming rules of XML are applied, the XML specification does not lay any constraints on criteria used in markup or in defining or naming of elements types. On the contrary, one of the main goals of XML is to be able to markup as many kinds of information as possible.

What you later want to do with the marked-up pieces of information is of no concern to XML or the XML processor. If you want to build and use an application or a program that treats some of the markup as presentational instructions, that is all up to you. Web browsers and e-book readers do this today. But both W3C and OEBF want to move in a direction where treating markup as presentational instructions can be minimized. The preferred W3C and OEBF way of describing presentational information is by the use of style sheets like XSL[29] or CSS.[30]

As pointed to earlier, what applies in an XML environment does not necessarily apply in the area of book and e-book publications. This important point is naturally also valid the other way around: whereas there is no way of separating content and presentation in a publication, such a distinction is essential in pure XML environments. It is obviously correct, as the XML experts point to in the above citations, that XML separates content (in the narrowly-defined XML way) and presentation. Information conveyed in XML documents is usually displayed using separate style sheets, which give programs instructions on how to display (or print) the information. OEBF wants as much as possible - and ideally all - of the formatting information to be separated from the element structure of e-books. In the OEBPS 1.0.1 it is said that future versions of the specification will continue in the direction of a "more rigorous separation of content and presentation", hoping that the use of style sheets will eventually prevail.

7 XML - the real difference?

XML documents have both a logical and physical structure, as described in the XML specification.[31] The logical structure is composed of declarations, comments, character references, processing instructions and elements, all of which are indicated by explicit markup. The physical structure is composed of storage units called entities, which all convey content. The content of the entities can be internal character data, as we have seen, or it can be can be external texts, figures, pictures, videos or all kinds of contents that are included in the document by reference.

The external contents included in XML documents are often files or parts of files residing on either the host system or on some other remote system. Surrounding XML, the linking and pointing capabilities of XLink and XPointer increase the possible ways of pointing to resources and representing their content in multiple ways.[32]

One main objective of XML is to extend the possibilities of the Internet both in terms of advanced publications and extensive trade. In the area of publishing the XML vision can be illustrated by a geography student browsing and reading a publication made by a textbook publishers, looking for all kinds of information about the country she is studying. In the near future, the vision says, the student can crawl into the sofa, turn on a wirelessly-connected handheld computer and open an electronic geography textbook.

If it is Italy she wants to study, she can read about Rome, watch videos from St Peter's Cathedral or watch a digitally animated reconstruction of Forum Romanum. Pointing to links, in pop-up windows she can view the latest economic statistics from Italy or perhaps a list of the current top news issues, which of course she can go on reading if she wants to. In a secondary window she may view this week's Italian top 50 list and check off three or four melodies to be played in her headset, while she continues to read about the sinking of Venice, all the time making annotations and highlights for later use in her monthly geography (electronic) paper. Now and then, with a tick of her finger in the lower right corner of the display, she can check the price of her reading, deciding to insert a bookmark and close the book as she passes 1.25 dollars.

Much of this is possible today using cabled computers, HTML and Web browsers. The rest can be done tomorrow with the use of XML, or so the story goes. When mobile connectivity, handheld computers and computer screens improve, the sofa scene with the student might even come true.

How long this will take and how many XML features will actually be implemented in future e-books, is too early to predict, but digital publications using many advanced features will obviously be produced.[33] For publishers accustomed to compact disc productions and multimedia Web productions, the inclusion of video and sound in publications are not new. But in one single publication, to link and combine fresh and updated information of all sorts from many different sources is a new way of publishing, quite different from today's paper book productions. So different actually, that much of the content made to be used in electronic publishing will be useless in a paper environment, and vice versa.

8 New Media - New Genres

By digging into the philosophy and work of W3C and the distinct features of XML we not only get a deeper understanding of electronic publications, but in addition a grasp of the diversity of future publishing. Such an examination shows that old-fashioned paper books and future e-books, as we can predict them, are really two very different media.

Paper books and other paper publications are combined presentation and storage media where the display of information is altogether visual and the content is physically tied to the paper and the pages of the publications. Books are thus static, solid and enduring. Books and other paper publications have developed a high level of legibility, sophisticated layout and high quality reproduction of pictures.

In e-books and electronic publications, storage and presentation are separate. The digital nature of the storage units makes e-publications highly mobile, across the Internet and other networks. The on-screen display of information is instant, and combines content data (in the narrow sense) and presentational data. The information of e-publications can be presented on smaller or bigger screens (and to certain degrees by loudspeakers) and they can include all kinds of media (text, video, sound). The included storage units do not necessarily have to reside on the system that is displaying the information. This makes electronic publications fragile, less enduring, but easy to update and extremely flexible and accessible.

Both paper and electronic media will have advantages and disadvantages, but in the longer run they will specialize and develop their stronger features. Paper-based publishing and electronic publishing will coexist, capturing different market segments and fulfilling different communicative goals. What parts of traditional markets e-publishing will take over and what new areas of application electronic publications and e-books will develop, is an interesting question, to which we have no answer yet. What seems quite sure is that a whole range of new genres and communication formats will develop, along with a number of new distribution channels and business models. Interestingly, this division of markets will have the natural consequence that multiple uses and re-use of content will not be at all as extensive as many claim today. In many cases, as the division develops, the contents used in electronic productions and paper productions will not even overlap. In such cases the doctrine of one input - many outputs will certainly be misplaced.

In other cases the doctrine will not be entirely misplaced, but clearly an exaggeration. In the narrow area of writing and pure text production, the features of XML and e-books will also cause new genres and language styles to develop. XML and related technologies will increase the diversity of books, creating specialized e-books and specialized p-books.

9 E-books and P-books

Engebretsen shows, by example, how the use of hypertext and multimedia is likely to change the online-news genres. Some of the changes are due to the fact that new (hyper)textual structures need new linguistic mechanisms to make the overall news items in some way coherent.[34]

Most XML based e-book formats today (Microsoft, Cytale, MobyPocket and HieBook) do not support "living" media like videos and sound. Neither do they import external content to be included as part of the publication. In current e-books, text files and graphic documents are, along with metadata files, packed into one rather closed publication. Except for the use of dictionaries, the hypertext and link functions do not apply outside the publication. But even this rather limited use of the XML will, as it evolves, change the structure of texts at all levels (as in Engebretsen's news), even more so as linking and navigation within e-books are improved.

When writing a p-book, be it a textbook or a scientific report, most writers have felt the frustrations the linear characteristics of paper-bound written accounts lay upon the presentation. The argument you want to go through in chapter two in many ways presupposes an understanding of the concepts and theories that are explained in chapter four. But of course, the concepts and theories in chapter four can only be adequately explained after the completion of the arguments in chapters two and three.

To get out of this trap the writer, in chapter two, has to present a preliminary definition of the concepts and write an extra paragraph or two explaining some theoretical premises, before going on with the argument. Some of these inaccurate preliminary accounts will have to be repeated in chapter three before the concepts are finally explained in chapter four. If I were allowed to speculate along the theoretical lines of Engebretsen, we could try to imagine how e-books will change this.

In an elaborate e-book, many paper-imposed tendencies towards a redundant (and loquacious) style can certainly be avoided using linking and hypertext facilities. Instead of, in chapter two and three, repeatedly and varyingly inaccurate accounts of a concept, the term can be linked to an exact definition that is presented in a pop-up window for the readers to activate. In the same way, redundant explanations can be hidden and only activated by those who need them. The hidden pop-up explanations can be standalone pieces of text or they can be illuminating parts of the explanations in chapter four. Indexed words or places in the text may also, through pop-ups, point to other places in the book where the same concept or theme is treated. The reader can jump to one of these places or just activate one or more of the spots in secondary windows. This way the writer can in a straightforward way carry on the argument, without having to make detours defining and explaining the concepts. Undisturbed, the writer can follow the line of thought, right on target.[25]

Even if we don't know exactly how, cross-referencing and linking to other parts of the publication, enabling definitions and explanations to be brought in all around the book, will obviously influence ways of writing, and the changes will be found at all levels of the text, both micro and macro. Result: new genres.

10 E-genres

New formats and new genres are an inevitable consequence of new media technologies, as the history of communication has shown. But in a period of transition, well-known formats will usually be transferred, more or less unchanged, from the old media to the new. This was the case when oral narrative genres for a long time dominated the manuscript tradition. As time passes, though, the characteristics of the new media will transform the old formats and influence the shape of a variety of new genres,[36] as did print technology more than 500 years ago.

In the general area of electronic publishing we are in an early phase of transition. Even though hypertext and multimedia are used in a number of publications, most of what is called electronic publications are basically print formats and print genres presented on computer screens (and substantially adding to the piles of loose paper sheets that grow in most offices and homes).

In a dissertation, Engebretsen shows that a transitional phase still dominates online newspapers, where news from the paper version is often used more or less unchanged in the online editions. Not surprisingly he predicts that many online newspapers will in time mature, make more exstensive use of hypertext and multimedia possibilities and eventually develop into genuine online news media.[37]

In the narrower field of e-books the development of a new medium has been the focus from the beginning. E-book technology combines hardware and software efforts trying to transform enduring screen reading into a pleasant experience. Light handheld devices running reading software that exploits the possibilities of LCD displays, both in enhanced font rendering and in flexible design, so far seems the most promising path to follow in creating a reading medium independent of paper.[28]

In spite of this history, in the area of e-books there will also be a period of transition in which print formats and paper genres will be tried out in e-book environments. As technology improves and formats evolve, including more XML properties, unique e-book genres will evolve, especially in the area of textbooks. Both writers and publishers will get used to the new XML formats. Soon e-books that are composed to utilise the possibilities of XML will be impossible to print, and after some time it will be the same the other way around: textbooks written for print will not be attractive e-books. Even if you add e-values to p-books, their language styles, their descriptions, explanations and their overall text progression will not be adapted to the expectations of advanced e-book readers.

Collaborating with writers and publishers, during the ongoing transition period markup experts will create a whole range of new element types, in addition to those already created. More than 500 years ago typographers created new ways to visually present written content in printed books within in a few decades. These typographical features are even today the core elements in the markup of books and e-books. In the next decades a similar transformation is likely to occur in electronic publishing. Hypertext and linking facilities are new ways to present written (and other) content, adding new element types (and attributes) to the markup of texts. As more advanced linking and navigation properties are implemented in e-book formats, creative writers and markup experts will invent many new element types to make the most of the new technology. Familiar element types, like definitions, tables, notes and indices will all change their meaning as they become important nodal points in the navigational and linking structures of e-books. Exactly as with typographical elements, the names and markup criteria of the new elements will be based on their presentational features and functions, more than on semantic content properties.

When hypertext and linking facilities are incorporated into e-book genres, the overall structures of texts will change, as will the logical (and hierarchical) element structures of texts. In XML terms this means that the logical structure of an e-book will be very different form that of a p-book, making Salo's "era of same book-different formats" a broken dream, even more so, of course, as the content of the "common" elements (of chapters and paragraphs) will differ as language styles, arguments and descriptions adapt to the communicative goals of the new e-genres.

The whole idea of one book - many formats will be even more remote when e-book formats start to implement videos and sound and in addition include content from remote computers into the e-books. Then, in XML documents, not only the logical structures but also the physical structures of e-books and p-books will be totally different. In such cases you must be a magician to carry out the doctrine of one input - many outputs.

Consequently, there really is no point in XML experts promoting XML saying it will bring publishers into a relatively uncomplicated era of one book - many formats or one input - many outputs. It is better to realise that e-books and p-books are in fact two different media and that XML is one of the technologies that make the difference. As publishers move into electronic publishing, XML will not remove the complexity created by new media, new genres and by XML itself. On the contrary, XML will add to this complexity.

11 XML in Action

So, where does this leave XML?

This leaves XML in the middle of the action, at least in e-publishing. If publishers markup their texts in the richest possible way using both semantic and typographic coding criteria and a logical and hierarchical structure, the publishers have a good starting point for creative design, extensive use of content and a widespread distribution of electronic publications.

Using XML gives designers a flexible way of changing styles, but not, as argued in this essay, without giving the text a slightly different semantic flavour.

This feature is a great property of XML. When you have marked up the element structure of a text, it is easy to manipulate the appearance of different parts of the text (Table 2) using external style sheets. Designers do not have to go through the whole text to make the changes, and they can do the same changes in many texts at the same time.

Even more important is that the standardisation work of OEBF will make it easier to work with different electronic formats and stylistic philosophies. When designers have determined the element structure of an e-book and marked it up according to the specifications of OEBPS, they can quite easily make up e-books, using style subsets and formatting specifications of lots of different formats, like for instance those of MS Reader, Gemstar E-books, Cybook or Hiebook, as long as these formats follow the OEB specifications. When compatibility and interoperability count, XML and OEBPS, because of their platform-independent and open-source character, will be the choice for most publishers who want to sell their e-stuff in as many formats and channels as possible.

XML will also be important, in spite of all warnings in this essay, in the area of uncritical multiple uses and reuse. Vast quantities of paper content can and will be reused electronically without any reediting or rewriting at all. By reformatting text material and other contents for new purposes, publishers in growing numbers will exploit the possibilities of network-based distribution channels. Since utilizations of this kind neither takes into account the limits nor makes use of the possibilities of electronic publishing, these more or less automatic conversions will inevitably bring many low quality publications on to the markets, especially in the areas of e-textbooks and e-learning.

In other areas of publishing, automatic conversions will be far less problematic. For many typographically simple genres, like most present fiction, reuse has already proved to be relatively easy. In these cases poorer quality is due to bad computer displays, immature reading programs and ignorant designers, more than to unedited reuse of original material. In the future, XML-based workflows will make re-use of many fiction genres even easier, as these visually and navigationally uncomplicated texts can be made into a variety of paper and electronic editions from the same XML document by use of style sheets (until fiction writers develop genres that use the possibilities of e-books, making conversions into paper books literally impossible).

As electronic publishing matures, reuse or multiple uses of texts will certainly develop and increase, with pieces of text crossing boundaries of genres and media. Parts of a textbook chapter might for example be used in an online encyclopaedia article, provided relevant hyperlinks are added and a new beginning and ending is written. The explanations of a linked e-book definition list may be presented as margin or frame texts in a print-on-demand version of the same book, and the first, forty-fifth and ninetieth picture from a vector based e-book animation might be used in printed pamphlets promoting the e-book.

In this situation of "many inputs - many outputs" it will be extremely important to keep track of all the pieces and parts of texts (and pictures and other media) that are scattered around in different editions of a publication (and sometimes will have to be updated). Only the hierarchical and logical structure and the pointing and linking properties of XML make such complicated logistic tasks possible. XML will be at the core of most Digital Asset Management systems.

As the sales of digital content increases and multiple distribution channels and DRM systems come into use, it is correspondingly important to be able to identify parts of a publication, like a chapter or a section of a book (or a melody, a picture or video), so that these parts can be sold separately or in combination with other parts of publications. Whatever identification system follows ISBN and ISSN (DOI being the likeliest successor[39]) and whatever sophisticated DRM systems dominate, publishers will have to markup their content in a logical way to be able to utilise the possibilities of these systems. XML is an obvious candidate for this task.

Thus, there are many reasons why XML will be part of the e-action, but as has been indicated, the most important reason is that XML has the power to make electronic publications into completely new media succeeding traditional print media. Salo summarises the new properties of e-books as being: text reflow, accessibility for the visually impaired (e.g. text-to-speech), linking (both within e-books and from an e-book to other resources), easy navigation through an e-book, searching, rights-related behaviours, multimedia display, interactivity, annotation and bookmarking. As Salo points out, most of these properties rely on a logical and hierarchical identification and markup system like that of XML.

12 Conclusion

This article questions the XML doctrine of one input - many outputs, especially the way it is used in publishing. Supported by linguistic and semantic theories and practical examples from writing and editing, I have argued that the doctrine is basically wrong.

A text almost always belongs to a media-specific genre. Every genre has rules or norms telling the author how to organise subject matter, how to design an argument (or a narrative plot) and how to use words and a vocabulary in shaping the genre's common language style. This makes it a difficult task to take a piece of text and use it inside a text of another genre. Or, for that matter, to take a text from one medium and use it in another medium. It is not easy to use print genres in electronic environments automatically, and especially not electronic genres in print environments.

On these grounds I warn against computer scientists and conversion house spokespeople telling publishers that XML will lead them into a relatively uncomplicated era of one book - many formats. This article shows the opposite to be the case. Electronic media give rise to new presentational principles and new genres, and XML is one of the technologies that will add to the diversity of publishing, forcing publishers to develop even more complex production and distribution systems.

This does not mean that I underestimate the powers of XML. On the contrary, XML and related technologies, like handheld computers and wireless connectivity, will add valuable features to e-books and other electronic publications. Electronic publishing will be transformed into a qualitatively new medium, having less and less in common with the older print medium.

New properties of electronic publications, like linking and navigation, inclusion of multimedia and combination of content from many sources, will make e-publications impossible to print. The written parts of e-publications will in addition adapt to the new medium making e-texts and p-texts two different, often incompatible, things.

This all indicates that the doctrine of one input - many outputs is misplaced. Publishers that learn the possibilities of XML and move into the new medium of electronic publishing will, contrary to what the doctrine claims, learn to master many new skills and to utilise a new, challenging and complex state of many outputs - many inputs.