Issue COLON-NUMBER Writeup

Issue:        COLON-NUMBER

References:   Parsing of Numbers and Symbols (p339-341, 343-344)

Category:     CLARIFICATION

Edit history: 22-Oct-87, Version 1 by Pitman

Status:	      For Internal Discussion


Problem Description:


  CLtL is unclear about whether a colon followed by a potential number

  is a potential number. There are passages which seem to address this

  issue unambiguously but fail.


Proposal (COLON-NUMBER:UNDEFINED):


  Clarify that syntax involving a leading colon followed by a potential

  number is not well-defined. That is, use of notation such as :1, :1/2,

  and :2^3 in a position where an expression appropriate for READ is

  expected is an error.


Rationale:


  This makes the status quo apparent.


Current Practice:


  Some implementations, such as Symbolics Lisp, claim the right to 

  interpret this as an ``is an error'' situation where their

  upward-compatible extension is to parse ``:1'' as the number 1

  (incidentally, but uninterestingly, expressed in the KEYWORD package).


  Other implementations tokenize ``:1'' as a single token, identify it

  as a symbol, and then parse it as :\1 would be parsed.


Adoption Cost:


  None. This clarification forces no implementations to change.


Benefits:


  Programmer expectations that any useful behavior can be portably relied

  upon in this pathological case should be soundly trounced.


Conversion Cost:


  Slight. Some users may have mistakenly believed that this was an aspect

  of the language that was guaranteed and may have written programs that

  exploited that belief. Such incidences are probably very rare, however.

  Also, even in the few cases where such code fortuitously worked,

  implementations are likely to continue to support it so such code will

  probably continue to fortuitously work in many of those rare situations.


Aesthetics:


  Arguably a slight improvement in visual aesthetics. What was already 

  a pretty marginal syntax is discouraged.


Discussion:


  Pitman supports this clarification. He thinks this issue is not a big

  deal, but it does crop up often enough to be a nuisance. It would be

  nice to have everyone acknowledge a unified position, even if that only

  meant agreeing to disagree.