Why SGML is Prescriptive and Interpretive

I have argued elsewhere that, like any encoding syntax, SGML - although it does not impose any particular kind of interpretation on texts encoded - favours some types of interpretation over others. For example, SGML invites us to organize information hierarchically and makes us particularily consious of overlapping between such hierarchies, a concern not at all obvious to anyone working with a syntax which does not encourage hierarchical structuring. This feature of SGML (and the critisisms of it) has lead to valuable theoretical insights, but also to serious practical problems for text encoders.

I have also argued that the TEI encoding scheme, while excellent for a large number of projects, is particularily unsuited for projects aiming at what Michael Sperberg-McQueen has termed declarative presentational markup of certain kinds of common text types important to humanities scholars. While I have been and hope to continue to be involved in attemptes at accomodating TEI for such purposes, I am of the strong opinion that this adaptation would have been a lot easier if the TEI had not used SGML but rather some other syntax.

This does not mean that TEI should not have used SGML. Given scarcitiy of resources in humanities and the overwhelming success of SGML in other areas, SGML was if not the happiest, then still the only viable choice. Despite my criticisms I always advice colleagues new to text encoding that the first thing they should do in planning a new project is to look carefully at the TEI Guidelines and only contemplate deviating from it if they are able to give clear and specific reasons, and know exactly which price they will have to pay.

We can limit the magnitude of the price if SGML more truly reflects the concerns and needs in the scholarly text communities. Thus, I will argue that SGML should be changed in a way that reflects the concerns and needs of humanists. Who knows textual traditions better, and why should society at large neglect this valuable source of experience and leave it to the computer industry to design text representation tools of the future?

We will conclude the paper with a discussion of non-interpretational facts about texts, and its relationahip to markup. Even if we argue that texts are not objectively existing abstract entities (like, e.g. Michael Sperberg- McQueen holds) of which we can make true or false descriptions, this does not mean we cannot distinguish between correct and incorrect representations of texts. We can make this distinction perfectly well within any given framework of discourse. And there are some interpretational judgements on which in practice all participants in any discourse will agree. We have also established, during a long tradition, certain agreed ways of deciding in situations of disagreement.