next up previous
Next: References Up: THE COMMON BUSINESS Previous: CBCL and natural language

1998: Advice for XML, W3 and ICE


This summarizes and extends remarks made in the 1998 footnotes.

  1. It is important to keep the language incrementally extendable. Many extensions will add detail to messages so that less human intervention will be required.
  2. Lisp notation is better. Oh, well, the committees have decided otherwise. Anyway XML is isomorphic to the subset of Lisp data where the first item in a list is required to be atomic.
  3. Lisp data provides the additional generality that the first item of the list may itself be compound and have to be evaluated to determine how the rest of the list is to be interpreted. I don't know whether this would get much use in CBCL or XML.
  4. Use modifiers like ADJECTIVE. Thus (ADJECTIVE FOO YELLOW) means a yellow FOO. However, a program not yet equipped to understand YELLOW but which can understand FOO may be able to do something useful with the information, given the convention that (ADJECTIVE x y) is a kind of y. Many English adjectives are not used this way, and we propose only to use such proper adjectives.
  5. All expressions that may be taken apart should have the standard syntax at least as an alternative to special string syntaxes. The Lisp systems don't do this properly, e.g. with time strings, and I notice that the W3 draft doesn't either. In Lisp a time has the string format exemplified by Wed Nov 11 12:58:28 1998. It should also allow the format (TIME <weekday> <month> <hour> <minute> <second>). The operators on such string should have the names they would have in the list or XML format.

John McCarthy
Fri May 7 13:56:17 PDT 1999