Occurrences

Paper

../papers/03-01-02/03-01-02.html

Date of Presentation

Wednesday, 22 May

Time of Presentation

09.45

Presentation Level

In-The-Middle

Abstract

Many commercial publishers were looking to XML Schema technologies to replace DTDs and to solve some of the quality and data typing problems inherent in large volumes of marked-up data. But do XML Schemas actually provide a better way to define full-text documents This presentation outlines the motivation behind the decision by Cambridge University Press to pilot development of an XSD Schema for use in its book and journal production. The advantages of Schemas over DTDs were readily apparent, especially their more meaningful data typing and better XML Namespace support. Data typing would allow quality assurance checks to be embedded in the Schema, rather than devolved to external programs. Namespace support would allow vocabularies to mingle freely together without some of the hacks necessary with DTDs. The main findings of the pilot can be summarized thus:Initial development was hampered by the (understandable) difficulty of finding tools whose behavior conformed properly to the Recommendation, though many vendors are promising Schema support ‘real soon now’. Tools that did manage to parse the Schema were many times slower than those which validated against an equivalent DTD. Moreover, the multi-entity disposition of a large Schema means access via HTTP over a slow network is tortuous. XSLT-based transformation of XML data is obviously a common activity, yet with XML instances valid to a Schema, this becomes a much more involved process. In-depth understanding of XML among suppliers to the publishing industry is still variable, and the introduction of an XSD Schema with its attendant rigors would be likely to cause expensive confusion. The data typing available through XSD Schemas falls well short of that necessary for use in full text documents being unable, for example, to allow a processor to determine something as fundamental as whether a table can be rendered. There are, however, some positive suggestion that emerge through Schema development, particularly concerning how a DTD can be split for modularisation.