Here are suggested fixes to the current ABNF draft.
Carl-Uno
>Date: Mon, 14 Apr 1997 16:46:30 -0700 (PDT)
>From: Chris Newman <Chris.Newman at innosoft.com>
>Subject: ABNF near finished?
>To: Detailed Revision/Update of Message Standards <drums at cs.utk.edu>
>>At the IETF meeting, we spent a fair amount of time discussing some
>possible problems with ABNF. The conclusions were as follows (using my
>own interpretations):
>>(1) It's too hard to get the set notation {} both clear and useful, so
>we'll punt.
>>(2) Range notation needs to be simplified to avoid confusions like
>"a".."Z", which is ambiguous since each end is case-insensitive. Also
>decided to use "-" instead of "..".
>>(3) The "#" notation is used incompatibly in many specs (e.g. 822 & 2060).
>Decided to drop it from notation since it's usually #thing or 1#thing both
>of which are fairly simple to write without that notation. In addition,
>current definition implies linear-white-space which is troublesome.
>>(4) Case-sensitive strings can be written with literals, and shouldn't be
>used often anyway. No special notation needed.
>>Given these decisions it appears the ABNF document has no remaining
>substantive open issues. If you think this is wrong, please speak up now.
>>[As an implementation test, I updated my working copy of the ACAP
> specification to use the new ABNF plus these decisions. Removing "#"
> wasn't bad, and the literal numbers + ranges allowed me to get rid of
> all <description> rules.]
>>>>