Balisage Paper: How to Play XML: Markup Technologies as Nomic Game

Rules and rule systems

While it is very difficult to generalize meaningfully about the context and nature of such agreements, if we are to try to extend an argument like mine – markup technologies are games or are like games, or at least provide occasion for games – it is necessary to do so at least in a rough way. Along these lines, I would like to propose several categories or terms of art, which we might use to distinguish the outlines of any agreement to play.

Figure 1: Rule spaces and regulation spaces

Figure 2: Lawful and unlawful states in Scrabble™

Although all of these represent states on the terrain, only one of them represents a lawful game state according to conventional rules. (According to the Official Tournament Rules, another one is legal if no player chooses to challenge it before another move is made.)

How are markup technologies like games?

Markup technologies, considered generally (that is, including HTML/CSS and wiki markup as well as XML/XSLT) have a number of game-like characteristics:

Notably, competition is excluded from this list. While competition is clearly part of activities related to markup technologies, it is not obviously necessary to them: win/lose outcomes are not inevitable at least within the straightforward practice of markup. While one might allow that successfully validating or publishing a markup instance is winning in some sense, such a victory would not ordinarily require anyone else to lose.

Figure 3: Markup technologies as game

Just as games come in more or less strict and regulated forms, so do markup language applications.

Yet beyond those just listed, there is a more striking similarity. Markup technologies show the same systematic elaborations as games do between games-as-rules (rule spaces) and games-as-regulated-rules (regulation spaces). Just as we have board games (games played, usually in turns, on a board at a table), Scrabble (a particular such game, whose terrain is a board with a grid and tiles carrying letters with point values), and then NSA Official Tournament Scrabble™, we have text-based markup, then XML, then an application of XML such as TEI or Docbook, then an application profile of one of these (or of some other common tag set, or a private tag set) in use by a particular publishing system or organization, and so on.

Moreover, the regulatory systems that are deployed to manage and support markup applications, including schemas, documentation, stylesheets and so on, are themselves subject to a kind of game play. Certainly, proponents of TEI, Docbook and other widely-used tag sets have been known to feel themselves in competition, as if to be widely used is to win. Browsers and commercial toolkits compete with one another over who gets to define the rules of how markup is to be handled, as a guarantor, in their case, of commercial viability; this is a reflection and an outgrowth of the fact that even proprietary applications may be used as regulators (sometimes the only regulator) to manage the rule space of even a supposedly non-proprietary, standards-based technology such as HTML or CSS. At a higher level still, standards organizations and industry consortia are enlisted as regulators of this competition; then standards bodies themselves compete over markets. Evidently there is plenty of zero-sum (or at least the players behave this way) and plenty of competition.

However, the fact that markup technologies are not inevitably or inherently competitive is only one instance of a more general distinction, namely that markup technologies as a whole, for all their rules and rigidities in some respects, do not make for the same kind of stable bounded system that a game is, even when this is true of particular markup-based applications. Markup technologies are more complex than games, even complex ones, and more specifically situated within (less isolable from) particular economic and cultural projects or endeavors.

Also, participants in markup technologies usually have less clearly defined tasks or roles than players of games do; indeed our work commonly mixes together the roles of player (someone who acts within game play to achieve aims within the game) and regulator (with responsibilities to monitor and maintain the game play as such). Nor is this because the game is simple enough not to require external regulation; on the contrary, markup technologies are also characterized by the opportunity to make and change the rules of play. All this makes them less like games, and more like nomic games.

Nomic games

In some ways, nomic games are the most gamelike of games, in that they afford the greatest opportunity for self-structured play: they are agreements to play in their most essential form. Yet they are also un-games, in the sense that they cannot be defined in terms of particular sets of rules. That is their point. To the extent that a game most closely approaches the mathematical ideal of a closed system in which moves, tactics, payoffs and their interactions are fully defined (and hence, at least in principle, fully intelligible), it is not a nomic game – even as the outcome of those interactions becomes too complex to be predictable. A nomic game is a system that is gamelike in one crucial respect – there are rules governing play – but critically ungamelike in another: those rules are not fixed, but instead are themselves in play.

Figure 4: Games and nomic games.

The term nomic game was coined by Peter Suber in 1982, in reference to legal and constitutional systems, to designate precisely the sort of game or gamelike system in which the players have scope to change the rules.^[5] Beyond this, it is necessarily somewhat difficult to generalize how nomic games work (as indeed this is up to the players) without resorting to group psychology or social dynamics. Although accounts of nomic games are not hard to locate on the Internet, they are nevertheless difficult to research, since as actually played they typically become highly introspective, self-involved, encumbered or elaborated with peculiar terminology (as nomic games are inevitably language and logic games, as well as opportunities for the display of linguistic prowess), and apparently all-consuming for their players (since they lack firm boundaries between game play and activities not within the game). Consequently, actual nomic games, even when their proceedings are archived in public, have a way of becoming insular and opaque.^[6]

But many activities or endeavors we do not ordinarily consider to be games are nomic in the sense Suber describes (which of course is what motivates his proposal). Indeed, the capability of defining their own boundaries by their rule-making activities might be an essential characteristic of cultural activities in general, considered as such, whether they be political and economic, artistic or creative, or technological. Herein lies much of the fascination of nomic games. Public institutions such as the law or commerce, which proceed in and by agreements among their participants, public and private institutions such as governments, clubs, corporations, and universities, are all characterized by the way their participants and stakeholders make and remake the rules of play, and sometimes challenge or avoid them. Art and poetry can be similarly identified with rules-making, to rhetorical and aesthetic ends. And so can technology, where the ends are practical and task- or (and) market-oriented. These various activities can be distinguished from one another by their various terrains – their material, social and economic contexts – as well as by the rules they follow and regulatory frameworks they enact. When we distinguish them in this way, we might almost be talking about games – a fact that has not escaped the notice of structuralist anthropologists.

Figure 5: Human culture and nomic games.

Many of the activities most distinctive of human culture and civilization, including law, technology, the arts, and various economic activities, may be thought of as nomic games. This etching by Blake both illustrates and exemplifies the principle of rule-making.

Yet even casual observation suggests some properties of such activities that distinguish them from more conventional games. Beyond the fact that we do not call them games or identify them as such except metaphorically (which we may do often), nomic games differ from conventional games in several respects, all of which have to do with how their boundaries are defined and maintained. For example, while ordinarily games are temporally ordered – either they proceed until a given conclusive state of play has been reached, at which point they end, or play continues for a definite and determined period until a score determines the outcome – nomic games, precisely since any rules governing time or sequence can be changed, tend to be temporally open-ended (even while there may be rules stipulating terms or rounds), and either unsustainable and short-lived, or interminable, sometimes making no provision at all for their own cessation. Also, conventional games are allogenic in the sense that they are motivated and staged by something outside them, whereas nomic games will be autogenic and even autopoietic (the term Maturana and Varela use to distinguish living systems as such), which accounts for their self-obsessed, self-consuming nature: nomic games frequently descend into rules lawyering. Perhaps most intriguingly, nomic games also tend to spin off or spawn other nomic games: they are replicative of themselves as well as mutable.

Finally, just as it is difficult (for reasons discussed earlier) to apply the idea of cheating in game theory, the notion of cheating only applies with difficulty to a nomic game, but for the opposite reason: the rules of a nomic game are mutable and always at issue, so nothing is by definition out of bounds. A nomic game may, at one moment, define a rule space (in the sense of a set of lawful states), and at the next, redefine it. It may enter into a paradoxical state in which the nominal rule space has been violated, but the agreement to play has not been suspended, so the game continues. And it may retroactively change the rules to make earlier moves (transitions) or conditions (states) legal or illegal, in order to invoke or to avoid invoking regulatory action as part of the game play.

To the extent that any game allows this (like professional bicycle racing, with its doping scandals) it is revealed as more nomic and less gamelike: players (a category which for these purposes will include the game's regulators, for example race officials and organizers) can take the rules into their own hands, and alter or break them, without ending the game. As long as as the agreement to play is sustained, even while the integrity of the game as a game (a transparent and fair competition) is compromised, as a nomic game (a self-sustaining but only relatively rules-bound activity), it may be reinforced. Thus, nomic games alter and stretch the notion of a game away from general principles of formal definition and into more socially located, problematic and provisional contexts. This helps account for why both games and nomic games, as actually played, are interesting, as each is always in the midst of becoming the other, in a Yin/Yang dynamic relation, in which cheating delineates precisely the in-between space where a game goes nomic (at the risk of collapse, if the agreement to play should become unsustainable), and where, conversely, an accusation of cheating may be a ploy that seeks to circumscribe a game-going-nomic back within the parameters of the game it is assumed or claimed to be – to regulate it.

Indeed, to recognize this is to come some distance to understanding of why thinking about games is useful for understanding phenomena and problems in the real world, a place where mathematical regularity is as often the exception as it is the rule. We might say that the reason we can so easily compare real-world activities (such as real estate speculation or international relations) to games (such as Monopoly or Diplomacy) is that the purpose of games is precisely to offer and encourage such analogies. As rules-bound and regulated systems, games are like the real world, only more so, with the special advantage of being more clearly bounded, delineated and self-contained, and hence more observable and intelligible – which makes them suitable occasions for practicing skills and abilities that may be useful outside them. Yet paradoxically, part of their resemblance to real-world activities is in their capacity to fly off, to become something other than themselves. And it is when games become more serious, when the stakes go up, that they have this tendency most strongly.

Yet just as games become more like nomic games as they become more serious and their players seek to take control of the rule space and shift it to their advantage, real-world nomic activities such as legal and political systems become more gamelike, more regular and rules-bound, as their participants find common interests in preventing ad hoc nomic redefinition on the part of other players, and in thwarting and punishing those who engage in it. That is, regulation spaces, both in games and in ostensibly more serious activities, are attempts to manage and control their nomic tendencies, to regularize and reduce complexity and seeming arbitrariness and instead offer systematic control or some sense of it. Intriguingly, such rules and regulations are what make game theory useful and relevant – at any rate, the kind of game theory we compulsively engage in as social animals, which is to say our more or less systematic reasoning about motives, options, objectives, tradeoffs, strategies and tactics. Within such systems (again, think of constitutional governance), although they remain nomic at their core, not just anyone is permitted to change the rules at any time, and there are complex rules, built-in checks, and regulation of regulations for determining who may engage in rules making, by what mechanisms, with whose participation or acquiescence, and within how wide a scope.

Again, for our purposes (and at the risk of anticlimax), it hardly needs to be stressed how markup technologies fit into this description of things, as both a technological and a cultural and economic phenomenon. A successful markup practitioner will soon have the opportunity to participate in rules-making activities, whether that be defining a schema, implementing a production or processing framework (where we frequently discover a fine line between interpreting presumed rules and making new ones), or merely developing ad-hoc conventions to systematize identifiers and taxonomies. Even writing stylesheets for presentation (to say nothing of code that converts from one markup format to another) will entail making new rules and, sometimes, ignoring or breaking old ones, if only to handle anomalies and edge cases. Successful schema designers and software developers will be pulled into standards work, to help define standards by which systems may be specified and which conformant applications and implementations of a technology must reference. And so on.

Figure 6: Entailment between states and rule spaces in nomic games

In a nomic game, every game state also implies a rule space, since no game state can be fully specified without also specifying the rules at that point in the game.

Valid, but not useful

Here is a valid fragment from a Balisage paper, valid being defined as conformant to the rules expressed in the Balisage DTD:

<personblurb>
  <para></para>
</personblurb>

A DTD can validate that an element is present, but not that it contains actual content. Even a schema language that can constrain the content to prohibit empty string values, such as XSD, cannot solve the general problem of ensuring that such content is correct or useful: XXX will usually suffice. In other words, schema validation is rather severely restricted in what kind of validation it can support.

This much is evident to any XML veteran – so evident that we may have stopped noticing. And the boundary between the rules enforced by a schema, and the rules that remain the concern of authors and editors (the human kind), is one of the first things newcomers to XML must get used to. Even if it never becomes entirely clear exactly where this boundary is placed – and it does vary from one application and one validation technology to another (which makes things worse for the learner) – the simple fact that the boundary is there (and indeed may move) is an important one. The XMLer who never catches on that the machine can check some things, but not others, will always remain at a loss.

What is at issue here is the authority and reliability of a schema as a regulator of rules. As such, practice shows that for a system to be sustainable, schemas – in addition to being well-fitted to the application(s) of the data, which should go without saying – must be consistent and as transparent as possible; hence the importance of documentation. Where they do not, in fact, enforce any meaningful constraints (such as the constraint that an author's biography should give some account of who the author is), users who are not experts in schema validation need to be helped to understand this. The differences between rules that are being validated, rules that might be validated but are not, and rules that cannot be validated by automated means, are often subtle.

One of the most difficult myths to dispel among newcomers to markup has to do with this authority, and to what extent a schema can or should be relied on as a warrant for correctness or fitness. The notion that a schema, as a regulator, performs some sort of punitive function – a validation error being construed to be some kind of warning that a document is wrong, which it is only in a fairly narrow and entirely convenient sense (inasmuch as the rules enforced by a schema may be motivated only by the need for the system to ensure a certain level of predictability in the data) – is probably less accurate than an alternative idea that a schema is just a piece of equipment that helps make game play feasible, by supporting (encoding) the specifics of the agreement to play. In the game of tennis, the rules require a racket to be used, and kicking or throwing the ball is not allowed; this is not because there is anything wrong with kicking, but because this game (on this court, with this ball) is more fun for players and spectators when played with rackets. Maybe we should be teaching that schemas in XML are like tennis rackets for users, helping them get the ball over the net, rather than bludgeons used by developers to keep users in line.

So, what is a document creator saying when he creates a personblurb element containing a para element with no content? Leaving aside the possibility that an empty para, particularly inside a personblurb, should be taken positively and frankly to assert there is nothing to say about this person, it is hard not to see this as a play of resistance (if that possibility does not in itself represent such a play). The schema says there must be an element here, and implies by the element's stated semantics (implicit in its name and perhaps explicit in documentation, as at http://www.balisage.net/DocumentModels/BalisageTL/index.html) that its data content should constitute some meaningful account of the document's author. But the document's actual tagging says, in effect, I see your personblurb element and choose to ignore it. The document is formally valid and enters the system. A warrant is made implicitly that information is given, which is missing. If, following the formulation of Sperberg-McQueen, Huitfeld and Renear, an element's semantics constitute the set of inferences licensed by the markup Sperberg-McQueen, et al. 2001, in the case of an element that is present with no meaningful content, an incorrect inference (there is a bio here) is being licensed; moreover, the more general capability to make such an inference when it needs to be made is confounded. And indeed, this much might be said of any case of tag abuse.

The system maintainers then have a choice: either they redefine processing logic to allow for this information to be missing (by changing the schema, and perhaps introducing defensive processing logic that tests whether personblurb has any content before doing anything with it), which is as much to concede tag abuse as normal. Or they can validate the document further through some means other than the schema, such as a Schematron or an old-fashioned editorial process, and use this as the basis of a counter-move, returning the document to its originator for correction. Or they can do neither, and allow the system subtly to degrade.

Sometimes system maintainers or process designers have no control over the data, and the third option is the only viable one. (One only needs to think of the history of HTML.) In this case, there may be system-wide changes in markup semantics as players (including document creators, publishers, systems developers, and tools vendors) learn how much (or rather, how little) they can trust, and make moves in response to one another.

Look ma, no hands

This problem of the role of the schema is also apparent, in a different way, here:

<!-- <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
                           "http://www.w3.org/TR/html4/loose.dtd"> -->
<HTML>
  <HEAD>
    <TITLE></TITLE>
  </HEAD>
  <BODY></BODY>
</HTML>

Is this document valid? According to both SGML and XML, validity requires an associated document type declaration, and this document has none.

Yet frequently, especially when working with other schema technologies, we work with documents that are not formally valid, but that would be when associated with their schema and validated, like the example above. Perhaps we might, in order to be clear, call such a document validable, to stress that it follows all the rules stipulated by its (presumed, but unassociated) schema, but actually has no formal association given, and can therefore not be valid in the strict sense.

At this point, readers who are experts in XML-based systems probably want to introduce qualifications, arguments and counter-arguments. For example, the schema language RelaxNG stipulates no formal mechanism by which an association such as that not made by this fragment should be made. Does this mean that documents cannot actually be valid to RelaxNG schemas? Presumably, in the context of RelaxNG, the definition of valid must be taken to be more like validable in the sense just given. (W3C XML Schema Definition Language takes something of a middle course: it offers a mechanism for documents to invoke schemas but does not require that it be respected, saying it might be just a hint.)

A similar problem arises when using XML catalogs, which can be used to intercept references to normative copies of schemas or DTDs and replace them at validation time with local copies. This is convenient because it reduces network traffic, but it introduces a source of potential error, in that if a local copy goes out of sync with a normative schema, local instances valid to it are not only not valid to the normative version (that is, to the schema actually named), they may not even be validable.

This much would be merely an academic argument, were it not common practice for XML developers to remove or comment out references to schemas as in the case above (where in fact, the schema named in the comment is an SGML DTD). This might indeed be considered a hazardous practice, like riding a bicycle without a helmet, except that it appears to be mandated by XML itself, which does not require a DTD or schema for processing.

At issue, again, is the question of who gets to make the rules – although this time, what we see is not the kind of power struggle within the game that is represented by tag abuse, but rather, a more fundamental difference in the way game regulators (in this case, schemas) are deployed on the game's terrain (the processing platform, which may be the entire network). The shift from SGML to XML, in which validation need not be performed, represents a highly significant shift in responsibility, from schema and system designers to local users and developers, who now get to determine whether and when to be valid at all, and (or) when to continue processing without schema validation (or any validation at all over and above syntactic well-formedness). This shift was possible for a number of reasons (including a number of syntactic restrictions and adjustments in XML to ensure that a non-validating parse would always be possible) and may have been intended primarily not as a way to increase options for users so much as for tool and application developers (since XML processing can, without a schema, be so much more lightweight). Yet it is highly consequential.

Generally, any time code in the document is commented out, the intended audience for the document is bifurcated between those who can see and make us of the commented information (whoever works directly with the raw code, including the developer who makes the comment) and those who cannot (any downstream users who cannot see the comments, but more importantly, the processing system itself). Again, this is a play of evasion, although different from the case of an empty personblurb or title element, in that it is not the spirit of the rules that is being evaded or faked out, but the processing system itself.

In this particular case (as long as the example is taken at face value), it is document validation that is being suspended. This may be a good or a bad idea; to determine which, we would have to know more. Either way, to allow (or require) documents themselves to carry the information that permits or implicitly requires them to be validated is itself a design decision on a meta-architectural level (a fact that RelaxNG and W3C XML Schema both recognize). Given XML's design, there is nothing here that actually prevents system designers from performing validation in any case, to any schema they choose. This single difference between SGML, where to comment out a DOCTYPE declaration would have been a futile gesture (preventing a fully conformant system from parsing the document), and XML, where it may (or may not) be a useful gambit, is one of the most important single reasons why XML can actually work on the web, a much wider, more open-ended, less well controlled and more resource-rich terrain than the systems for which SGML processing was developed.

And what do we have here

A final example is not an instance of game play in quite the same way as the previous two, but rather points to an issue in game design. Here's some code, valid to the NLM/NCBI schema (the text is from Bill Watterson's Calvin and Hobbes of May 27, 1990):

<boxed-text>
  <speech><speaker>Calvin</speaker><p>I don't have to sing the song!
    I was in the "No Song" Zone!</p></speech>
  <speech><speaker>Hobbes</speaker><p>No you weren't. I touched the
    "Opposite Pole", so the "No Song Zone" is now a "Song Zone".</p></speech>
  <speech><speaker>Calvin</speaker><p><bold>I</bold> didn't see you 
    touch the Opposite Pole! You have to declare it!</p></speech>
</boxed-text>

The interesting thing here (at least as far as the markup is concerned) is the boxed-text element. The semantics of this element are arguably underspecified. The text named by the element is presumably its data content, as is generally the case with descriptive markup. But this appears to be less descriptive, in the sense that it identifies the kind or type of text it marks, than presentational, asserting how it would appear in rendition. Yet it is also not clear whether it should be taken to be retrospective (the text appeared in a box in an earlier version) or prospective (the text should be placed in a box in a rendition generated from this markup). For the moment, we leave aside the possibility that boxed should mean something other than appears displayed in a box.

Yet perhaps not. The formal element description given in the NLM/NCBI version 3.0 Tag Library is somewhat more specific but also problematic. It offers examples of what kinds of thing might be tagged as boxed-text, but directly answers none of the questions just posed:

Interestingly, this definition rejects the single thing we thought we knew about a boxed text, that it should appear in a box. Apparently the box is not a renditional box but some kind of abstract rhetorical artifact: we know only that this text is outside the flow of the narrative.

What does this kind of markup say about itself; what licenses does it inference? Perhaps not much. Yet this is nevertheless more, I submit, than certain other escape hatch kinds of encoding, such as a TEI ab (abstract block), or an HTML blockquote. About these one cannot even say whether they are inside or outside the narrative flow. (In other respects these are of course not perfect analogues to boxed-text, for which either TEI or HTML might give better alternatives, but the point stands.)

I would describe semantics such as NLM/NCBI boxed-text as soft, as distinct from (say) XSL-FO block[@border='thin sold black'][@padding='2em'], which is hard by virtue of being bound to certain application semantics. And yet the element semantics of boxed-text are not as soft as those of ab or blockquote (which indeed might have harder semantics if it had not been subject to widespread tag abuse: the more blockquote has been used for things other than block quotes, the less useful the element has become for marking block quotes as opposed to other things that look like them). Maybe the meaning of boxed-text ought to be considered pliable: hard enough to mean something in application, but soft enough to be flexible in use and indeed capable, in consonance with its application semantics, of expressing more than one meaning.

How well this works is a matter of whether it sustains its semantics (such as they are) within its application domain. And since the purpose of the NLM/NCBI tag set, primarily, is to support the normalized aggregation into a single repository of data already marked up (typically by journal publishers), many or most of whose markup schemes have harder semantics of their own (usually but not necessarily display semantics), it has to establish a balance. When aggregating disparate data, to try to capture every detail of source encoding means risking both over-elaboration, and a failure in its goal of providing unified access and processing (a goal that is frustrated whenever there are many ways to do any given thing). But to erase the details of source encoding means, potentially, to lose information. The balance is in erasing incidental, insignificant differences and capturing meaningful ones. But ordinarily, the design of a tag set to do this has to be established up front, which is to say before the proper unification of data from disparate sources can be established (when this is even possible, which it may not be when source document types are themselves mutable and open-ended). The solution to this problem, as to many others related to it, appears to be in deploying a mix of element types, some of which are hard in the sense just described, and some of which are pliable or even soft, to serve as a kind of spackle or carpenter's putty (or the blank tiles in Scrabble or wild cards in poker). Indeed this is the case of all the tag sets mentioned here. HTML in particular is noteworthy for having taken on, in its history, both elements with harder semantics, such as b or i or the elements out of which browser forms are built, and softer ones such as div and span.

That is, rules are given, but latitude is also given to the players of the game to interpret how to apply the rules within their own situations. If this balance is found, then play can continue, since players can adapt. If this balance is missed, then either an overly rigid scheme is the result, in which individual publishers and projects are unable to achieve local goals (and eventually have to shift or adapt their schema, contrive workarounds, or stop playing entirely), or we have chaos, a markup free-for-all in which people pretend to be working together but meaningful aggregation or interchange is impossible.

The ethics of play

All of the examples explored above serve as examples of a more general point, having to do with the way we make judgments, as we approach both large-scale and minute problems working with markup technologies – that is, as we face both strategic and tactical decisions. What is useful and interesting about a game is that it is both isolable and insulated from the world, and yet at the same time, a mirror or model of it. This makes a game a kind of a microcosm within which moves and manners have their own significance in relation to one another, making for a kind of ethical laboratory. To consider right and wrong in a game is not to consider right and wrong in the abstract: a game defines its own ethical universe, wherein (as long as the game itself is respected) virtue, if it is to be virtuous, is always aligned with some sort of victory – if only of the moral sort. (The very notion of a moral victory points to the paradox that even in defeat can be a kind of success.) In other words, we are interested in games because they are arenas for the display and recognition not of abstract goodness, but of virtue in something more like the Latin sense than the modern one – that is, of particular virtues, of strengths such as persistence, fidelity, ingenuity, generosity, prudence, integrity, imagination, versatility, elegance or what have you. Some of these virtues, of course, may contradict others, and on occasion they may even stand in the way of victory – which is exactly why they become interesting when demonstrated in game play. The way games serve not only as training grounds but as arenas for the demonstration of such virtues goes a long way to explaining why we find them so compelling. Nor is this unimportant. Different games will foster different kinds of behavior in their players, depending on what sorts of strategies and tactics lead to better outcomes within the game.

Interestingly, to consider markup technologies in this light allows us to set aside from our consideration of markup-based systems, which as technologies, are social and economic and even political, rather than simply (like spectator sports) entertaining, the reflexive ethical judgements that so often seem to accompany them. For example, to prefer openly specified, standards-based technologies to proprietary alternatives, or to choose one tag set over another, is often if not usually characterized as a kind of ethical imperative. To use standard X or Y, in a standard way – or to be valid to a particular schema, or to use a particular schema language, or to prefer a certain processing platform, or a certain company's product – is considered good, and not to do so is condemned, stigmatized or derided as bad, without much reflection on why that should be. To see such a choice as a strategic or tactical decision within game play is both to relieve, and to resituate this ethical component. At one level, to prefer one schema (or schema language or naming convention or coding style or software product)`over another often seems to reduce to a kind of tribalism, the way the fans of one team or club must always show contempt for the fans of another; as such, a preference of this sort requires no explanation or apology beyond the play itself. Yet at another level, we can see that all such alignments have consequences in terms of strategy and the goals of a strategy, and as strategic choices, can make success in the game easier or harder to achieve – and not only because to make such a choice is to adopt the colors of one team and set oneself in opposition to others. One schema (to pursue that example) may actually be better suited than another to the task at hand. And yet, even when it is not, to forge a serviceable alliance with others who might help fit a particular schema to your purposes, may be to turn a tactical misstep into a strategic advantage. To understand such maneuvers as game play, rather than simply as conflicts of interest (although they are that), is to see how we can (and must) make sense of such complex considerations. In short, considering markup technologies as games can be a way of reframing our assessments and choices – this editing tool, that schema, even one pattern of tag usage versus another – within the assumptions, predispositions and prior commitments that otherwise drive them, to see how they might be more carefully considered and deliberate, less reflexive and driven by the demands of constituencies, audiences or authorities, even while those commitments remain real.

This is important because it allows us to discriminate real ethical imperatives – here, real means outside the game, and has to do with conduct in the world – from fictive and factional ones that have only to do with the play. These are easily fused, and their fusion is easily recognized by anyone who does not have a stake in the game, but the very essence of play – to play with commitment, with heart – is to fail to recognize this distinction. Perhaps the best players make this commitment, identifying the good of their cause with the Good in the abstract; this helps motivate them. But the very best players might ultimately be those who are able to step away from it again, having won or lost, motivated only by love of the game itself, and having done what they could to make the game itself worth the playing, whatever the outcome.

Lessons for practitioners

Another reason the analogy is important is that it enables us, as practitioners of markup, to look to game design and game play for ideas about how to go about our work. There are three general characteristics of games, in particular, which might be considered by users and especially designers of technology. All of them, I submit, have been at the root of the success of the SGML family of markup technologies (including HTML and all XML-related technologies) over the last twenty years, and in particular of the most successful application built on this platform, namely the web itself.

Figure 7: Markup technologies on the web

A successful nomic game is one that both grows and spins off variations.

The big picture

At Balisage 2008, Eduardo Gutentag drew attention to a paradox at the heart of XML. Its proponents have proposed as its raison d'être the principle that the encoding of an electronic document, inasmuch as it secures the document's intellectual content, must be accessible to the document's creator and indeed owned by the document's owner. Standards – which is to say, externally specified, published sets of rules that provide the basis for commodity markets in technologies and toolsets, so that the rules of the game are not absolutely controlled by proprietary interests, and anyone, at least in principle, can play – are critical to this accessibility. Yet the networked media that have been built on the foundation of open text encoding (that is, on XML and its sister, HTML, along with their associated specifications), allow and in fact foster a radically open-ended exchange of data that undermines the very premise that information may be owned. To be accessible to anyone is, in a certain respect, to be owned by no one, at least insofar as gaining access means gaining control, just as a ball, once the game has begun, belongs to anyone holding it. Given this tension, what will be the rules of this new information economy? Gutentag was not able to offer any definite prognostications.

If space, time and materials are no longer factors (once the platform is in place, the cost of copying a document declines to nil), and technical barriers (such as copy-protection schemes or opaque, binary formats) are not interposed between users of information and the technical and material infrastructure within which that information is created, maintained, transmitted and manipulated, then seemingly only social and legal barriers remain to prevent me from appropriating your data however I see fit. And even if such social and legal means are introduced, they may not suffice, at least for the purposes of those who have depended in the past on maintaining some measure of exclusive control over information they have managed. More deeply and more problematically, however, if the rules of the game are part of the play, the game is a nomic one, and runs the risk of all nomic games, that in the process of self-reinvention it becomes unsustainable, collapsing in on itself. Then there can be no winners, at least until a new game emerges with new rules. These and related anxieties are at the heart of why, for example, Open Access policies, such as those now being developed at major universities, are so controversial.^[8] On the one hand, producers of knowledge, such as researchers and scholars offered the opportunity (and perhaps the requirement) to make their work available through open access channels, have new and more powerful means of fulfilling their purpose of making their work available. On the other, to the extent that old institutional arrangements are imperiled (including implicit or explicit exchanges of value), so also are the securities offered by those arrangements.

As legitimate as these concerns are, however, there is a deeper issue that may prove to be a critical determinant. The basis of the new information economy, if it proves sustainable (and it seems fair to expect that some sort of sustainable model will emerge), is and will be largely in non-proprietary technologies; sooner or later, this would seem to be necessary for a system that is truly global. Yet if the technologies grow so complex that to play XML, or to work with markup, means you have to devote a career to it despite its openness in principle, then the removal of merely legal encumbrances counts for less, and it is also up to us technocrats: our loyalties will determine what masters XML will serve. In that case, even in an open-access world, where theoretically anyone can play the game of publishing, not everyone will be able to play the game of technology that underlies and supports it (to say nothing of kindred activities such as data interchange, curation and archiving, aggregation and analysis), and a corporate oligarchy based on property rights will at best be replaced by a technological meritocracy based on access, education and knowhow. More likely, an uneasy hybrid between these two alternatives will emerge, at least until a new set of standards emerges to attract new players with new rules, more open to the self-taught. In any case, much depends on how markup technologies are promulgated: who teaches them, and who learns them, and for what purposes.

Figure 8: Markup technologies: stages of play.

Win a round, and you may be asked to help make the rules for the next one.

Since this is a game, in short, that is played for keeps, within a larger world with larger stakes, XML, the design of markup technologies and applications, and the promotion of digital literacy in general, will all matter, at every level. Whether you are designing a document, a tag set, a transformation, a publishing system, a schema language, a processing language, a standard, a standards organization or a corporate strategy, keep this in mind. It is up to you and how you play that will determine the character, if not always the outcome, of the game.