Email list hosting service & mailing list manager

The meaning of braces in various Schemes John Cowan (05 Sep 2012 06:24 UTC)
Re: The meaning of braces in various Schemes David A. Wheeler (05 Sep 2012 11:50 UTC)
Re: The meaning of braces in various Schemes John Cowan (05 Sep 2012 17:20 UTC)
Do we NEED a marker at all? David A. Wheeler (05 Sep 2012 13:25 UTC)
Re: Do we NEED a marker at all? Jens Axel Søgaard (05 Sep 2012 20:42 UTC)
Re: Do we NEED a marker at all? Shiro Kawai (06 Sep 2012 04:27 UTC)
Re: Do we NEED a marker at all? Alan Manuel Gloria (06 Sep 2012 12:36 UTC)
Re: Do we NEED a marker at all? David A. Wheeler (06 Sep 2012 13:07 UTC)
Re: Do we NEED a marker at all? John Cowan (06 Sep 2012 17:09 UTC)
Re: The meaning of braces in various Schemes Donovan Kolbly (07 Sep 2012 02:12 UTC)
Re: The meaning of braces in various Schemes David A. Wheeler (07 Sep 2012 03:41 UTC)

Do we NEED a marker at all? David A. Wheeler 05 Sep 2012 13:25 UTC

John Cowan:
> http://trac.sacrideo.us/wg/wiki/BracketsBraces shows the meaning
> of square brackets and curly braces in my Scheme test suite, which
> currently includes 45 implementations.  Only Chibi and RScheme currently
> do something special with braces; the other implementations treat them
> as either synonyms for parentheses, lexical syntax errors, or identifier
> characters.

On guile's development list, svanleent asked if we could (also?) define a recommended filename extension for curly-infix.  Obviously we could.  But this *rarity* of conflict in existing implementations (thanks for the analysis!) suggests something else.

Maybe we should just drop the marker completely (e.g., #!srfi-105).  Not needing a marker or any other indicator would simplify things.  In particular, it would mean that we don't need an ugly marker in every file, and implementations would no longer need to look for it (and maybe handle it specially).  Chibi's {} can already be enabled or disabled by a compile-time flag; perhaps the implementer would be willing to add another flag for an alternative semantic using {}.  Besides, if an implementation doesn't support curly-infix, having a marker won't help :-).

Thoughts?

--- David A. Wheeler