Issue CHARACTER-VS-CHAR Writeup

Issue:        CHARACTER-VS-CHAR

Forum:	      Editorial

References:   Character Proposal 2.3.1,  MAKE-DISPATCH-MACRO-CHARACTER (p363),

	      GET-MACRO-CHARACTER (p362), SET-MACRO-CHARACTER (p362)

	      GET-DISPATCH-MACRO-CHARACTER (p364), 

	      SET-DISPATCH-MACRO-CHARACTER (p364)

Category:     CHANGE

Edit history: 12-Nov-90, Version 1 by Pitman

	      14-Nov-90, Version 2 by Pitman (minor mods to Rationale, Discussion)

	      15-Mar-91, Version 3 by Pitman (one new proposal, some endorsements)

Status:	      For Internal Discussion


Problem Description:


 Character proposal 2.3.1 adds the names BASE-CHARACTER and

 EXTENDED-CHARACTER.  These are naming-wise inconsistent with

 STANDARD-CHAR.


Proposal (CHAR-VS-CHARACTER:INCONSISTENTLY-STATUS-QUO)


 Leave things as they are.


Proposal (CHAR-VS-CHARACTER:LESS-INCONSISTENT-SHORT)


 Rename BASE-CHARACTER to BASE-CHAR. 

 Rename EXTENDED-CHARACTER to EXTENDED-CHAR. 


Proposal (CHAR-VS-CHARACTER:LESS-INCONSISTENT-LONG)


 Rename STANDARD-CHAR to STANDARD-CHARACTER.


Proposal (CHAR-VS-CHARACTER:CONSISTENTLY-CHAR)


 Rename BASE-CHARACTER to BASE-CHAR.

 Rename EXTENDED-CHARACTER to EXTENDED-CHAR.


 Rename MAKE-DISPATCH-MACRO-CHARACTER to MAKE-DISPATCH-MACRO-CHAR.

 Rename SET-MACRO-CHARACTER to SET-MACRO-CHAR. 

 Rename GET-DISPATCH-MACRO-CHARACTER to GET-DISPATCH-MACRO-CHAR.

 Rename SET-DISPATCH-MACRO-CHARACTER to SET-DISPATCH-MACRO-CHAR.

 Rename GET-MACRO-CHARACTER to GET-MACRO-CHAR.


Proposal (CHAR-VS-CHARACTER:REALLY-FIX-THINGS-UP)


 Rename BASE-CHARACTER to BASE-CHAR.

 Rename EXTENDED-CHARACTER to EXTENDED-CHAR.


 Rename MAKE-DISPATCH-MACRO-CHARACTER to MAKE-DISPATCH-MACRO-CHAR.


 Remove the functions GET-MACRO-CHARACTER, SET-MACRO-CHARACTER,

  GET-DISPATCH-MACRO-CHARACTER, and SET-DISPATCH-MACRO-CHARACTER. 


 Create setfable functions MACRO-CHAR and DISPATCH-MACRO-CHAR, 

 such that:


  - (MACRO-CHAR c [r]) means what (GET-MACRO-CHARACTER x r) 

    used to mean.

    

    (SETF (MACRO-CHAR c [r]) (VALUES fn [nt])) means what

    (SET-MACRO-CHARACTER c fn nt r) used to mean.


  - (DISPATCH-MACRO-CHAR d s [r]) means what 

    (GET-DISPATCH-MACRO-CHAR d s [r]) used to mean.


  - (SETF (DISPATCH-MACRO-CHAR d s [r]) fn) means what

    (SET-DISPATCH-MACRO-CHAR d s fn [r]) used to mean.


Rationale:


 We've enumerated the positions that people were likely to have

 so that they could refer succinctly to their position by name.


  INCONSISTENTLY-STATUS-QUO:  Leaves the language unchanged.


  LESS-INCONSISTENT-SHORT:  Changes only that part of the language

    which is not in CLtL, and which users probably aren't using

    much yet anyway.  At least makes all the type names use

    "-CHAR" consistently.  (Stops short of addressing why there

    is a name SET-SYNTAX-FROM-CHAR but another name 

    GET-MACRO-CHARACTER only to preserve the status quo.)


  LESS-INCONSISTENT-LONG:  Changes only one name  At least makes

    all the type names use "CHARACTER" consistently.  (Stops

    short of addressing why there is a name SET-SYNTAX-FROM-CHAR

    but another name GET-MACRO-CHARACTER only to preserve the 

    status quo.)


  CONSISTENTLY-CHAR:  Changes even some things which were in

    CLtL, but does so in a manner that is predictably consistent:

    The word "CHARACTER" is shortened to "CHAR" whenever it appears

    hyphenated together with other words.  This would be harder

    on existing programs, but easier for new users.


  REALLY-FIX-THINGS-UP:  Like CONSISTENTLY-CHAR, but also takes 

    advantage of a last opportunity to make the language use SETF more

    consistently.  Probably the reason this didn't originally use SETF

    was the lack of clarity in the managing of multiple values.

    Also, SET-MACRO-CHARACTER and SET-DISPATCH-MACRO-CHARACTER are the

    only two symbols in CL which begin with the substring "SET-" but

    are not set-manipulation primitives.


Current Practice:


 No one does this.


Cost to Implementors:


 Small, in all cases.


Cost to Users:


 INCONSISTENTLY-STATUS-QUO:  None.


 LESS-INCONSISTENT-SHORT:  Very small.


 LESS-INCONSISTENT-LONG:  Small.


 CONSISTENTLY-CHAR and REALLY-FIX-THINGS-UP: It's very easy to

   write the requisite compatibility functions in order to support the

   old names during a transition period.  With all the other things

   people are having to do to convert, this is relatively little extra

   burden.


Cost of Non-Adoption:


 See aesthetics.


Benefits:


 Aesthetic and Mnemonic.


Aesthetics:


 If this is adopted...      The affected parts of the language will be...


 INCONSISTENTLY-STATUS-QUO  glaringly unaesthetic

 LESS-INCONSISTENT-SHORT    tolerable, but not pretty

 LESS-INCONSISTENT-LONG	    tolerable, but not pretty

 CONSISTENTLY-CHAR          reasonable

 REALLY-FIX-THINGS-UP       very pretty


Discussion:


 KMP prefers (in decreasing order of preference) REALLY-FIX-THINGS-UP,

 then CONSISTENTLY-CHAR, then LESS-INCONSISTENT-SHORT or 

 LESS-INCONSISTENT-LONG, then INCONSISTENTLY-STATUS-QUO.

 Scott Cyphers (Symbolics) concurs with this ordering, except his first 

 choice would be for the [non-existent] option REALLY-FIX-THINGS-UP-CHARACTER.

 Scott McKay (Symbolics) likes CONSISTENTLY-CHAR but would thinks 

 REALLY-FIX-THINGS-UP would be ok.


 Bob Cassels at Symbolics asked why there isn't a CONSISTENTLY-CHARACTER

 option.  KMP replies:

  In general, we have tried to spell out words in the language. There

  are, of course, some glaring exceptions.  However, I feel that

  certain English words (like "paren", "char", "info", and "demo") 

  have acquired such high frequency usage in the Computer Science 

  arena that we are basically just waiting for the dictionary-makers

  to catch up with the natural-language phenomenon of shortening of

  higher frequency words by its speakers.  As such, I feel the issue

  is not one of either "char" or "character" being more correct, but

  rather one of choosing the names in such a way that users can 

  reliably remember which name is used in which situation.


 Masinter supports REALLY-FIX-THINGS-UP.


 Moon and Sandra prefer the long name CHARACTER to the short name CHAR.

 (Moon voiced implicit support for LESS-INCONSISTENT-LONG, even though

 the option didn't exist yet.)


 JonL says he and others at Lucid think that changing a lot of names at

 this point is a bad idea.  [It is for this reason that option

 LESS-INCONSISTENT-LONG was added in v3 of this proposal--to offer a 

 minimalist fix that might satisfy them.]