Standard Generalized Markup Language (SGML) offers at least three possibilities for representation of textual information: generic identifiers (GIs), attributes, and PCDATA element content. This paper discusses the question of when to use elements, attributes, or data content to represent information in SGML or Extensible Markup Language (XML) documents.

The paper describes the general text encoding initiative (TEI) [1] approach for ensuring the flexibility of a set of document type definitions (DTDs) designed for multiple purposes, and addresses the TEI approach for the specific case of textual critical editions. The paper identifies problems stemming from the TEI DTDs concerning what may be represented in critical editions, and also how that representation may be processed with SGML tools. The paper goes on to present three strategies for preparing TEI-conformant textual critical editions, which overcome those problems. These three strategies allow much greater structural control than the standard TEI DTDs, rely entirely on SGML for all validation, and create a final document that is TEI-conformant.

The paper is well organized, with examples that help the reader to understand its content. In addition, the author presents an appropriate level of detail. The reader, however, should be familiar with SGML. The paper is particularly recommended to those interested in the use of TEI DTDs.