Date: Tue, 2 Jun 1998 16:25:45 -0700
From: Dan Wing <dwing@Cisco.COM>
To: kre@MUNNARI.OZ.AU
CC: DRUMS@cs.utk.edu
Message-Id: <980602162545.2289f68b@Cisco.COM>
Subject: Re: Moving ABNF to Draft Status 

>I used it in a draft which had case dependant text - ie: if you really
>want "help" and not "HELP" you can't use an ABNF string, as they're
>case independant (which is generally useful) so my draft had stacks of
>stuff like %x68.65.6c.70 scattered through it (actually, when I look, I
>see I actually used %h... which looks to have been some kind of delusion).
>Anyway - the WG decided to switch to case independant commands, so all of
>that has gone (that wasn't a unanimous decision, it isn't impossible it
>might switch back - but I wouldn't be holding my breath waiting for it).
>
>There's a dilemna here - there are a few things in ABNF (like this one)
>which are clearly going to be useful over time, just not in every doc
>that uses ABNF.   If we wait a while, the uses required by the process
>will occur, and all would be OK.   However, we don't want to wait, as
>that would cause anything that uses ABNF to stall in the standards process,
>and if that starts to happen, other groups will define their own grammars,
>just to avoid being stalled, and the chances of actually seeing uses of
>all of ABNF's features drop.
>
>I have no idea what to do about this...

It is silly for a new grammer to be invented for the sole
purpose of case sensitivity when the ABNF draft could work
if its case-insensitivity could be (in essence) used or
not used.


The specific problem of case sensitivity can be easily avoided by 
having the ABNF core rules be something like:

  ALPHA_LOWER = %x61-7A           ; a-z
  ALPHA_UPPER = %x41-5A           ; A-Z
  ALPHA       = ALPHA_LOWER / ALPHA_UPPER

and
  
  HEXDIG_LOWER = DIGIT / "a" / "b" / "c" / "d" / "e" / "f"
  HEXDIG_UPPER = DIGIT / "A" / "B" / "C" / "D" / "E" / "F" 
  HEXDIG       = HEXDIG_LOWER / HEXDIG_UPPER

and removing the Note (and supporting text) in section 2.3.

which allows referencing the uppercase, lowercase, or case-insensitive
defintions easily. 
 

Perhaps adding a section like this in the ABNF document would suffice

  "Any document which references this ABNF document must
   indicate if its ABNF strings are case-sensitive or
   case-insensitive.  In the absence of an indication,
   ABNF strings are assumed to be case-insensitive."

-Dan Wing