I believe that this goal can be better achieved by exposing data rather than procedures. Data are much more stable and although the design depends on what you o

I believe that this goal can be better achieved by exposing data rather than procedures. Data are much more stable and although the design depends on what you originally plan to do with them, this dependency is not rigid and there's usually many different usage scenarios. Also, data are much more comprehensible and malleable than procedures so quite a few compatibility issues can be solved automatically. E.g. the primary goal of DTD was to validate XML against a simple grammar, but the same DTD can be used for auto-completion when editing the XML. Same data, different usage. Also, if we want to shift to a more powerful validation engine, e.g. XSD schema, we can parse DTDs and mechanically convert them into equivalent XSDs.