Re: [Json] String comparisons -- LAST CHANCE ON PROPOSALS

[Carsten Bormann <[e-mail]>]

On Jun 11, 2013, at 19:44, Paul Hoffman <[e-mail]> wrote:

> For purpose of establishing key equality, comparisons MUST be conducted, after all unescaping is done, by comparing numeric character code points. There MUST NOT be any modification of any kind to the characters in keys, including change of case or change between precomposed and decomposed forms.
> 
> For example, the following four keys MUST be considered equivalent:
> 
> * "\u002F"
> * "\u002f"
> * "\/"
> * "/"

-- This uses equal and equivalent as if they are equivalent.
   Note that 2.2 uses "unique", which doesn't quite connect to this text.

-> I would prefer to have a complete proposal that covers the 
   sentence in 2.2 as well and makes sure they connect.

-- Talking about "modification" confuses me.  
   Of course, there is no modification; 
   it is not OK to turn {"A": 1} into {"a": 1} in the parser, 
   or {"\u212b": 1}*) into {"Å": 1}.
   Is this still about the equality check?  Not at all clear to me.

-> Make it clear that this is just reinforcing the numeric comparison of code points.

-- Are "ab" and "ba" equivalent?  They use the same numeric character code points.

-> Say something about comparing the sequences in order.

And, of course, s/key/name/ (equality of names in objects, or really, uniqueness of names in objects).

Grüße, Carsten

*) (had to write the escape: my mail client normalizes the Angstrom sign under my fingers, WTF).