Re: [Json] Unpaired surrogates in JSON strings

[Carsten Bormann <[e-mail]>]

On Jun 6, 2013, at 02:58, Paul Hoffman <[e-mail]> wrote:

> Code points

That is a confusing term here, and I have a hard time understanding what people are trying to say.

Code points can refer to those of the characters or those of the code units (byte for UTF-8, etc.).
Please be specific which of the ones you mean.

RFC 4627 normatively references RFC 4234, which makes very clear that it is about characters.
This is further reinforced by the grammar in RFC 4627, which addresses code points up to %x10FFFF -- clearly it is about the Unicode characters (an assessment that also agrees with the rest of the text).

1) Surrogates are not characters, so they can't appear in the source character set.

2) Within strings, escape sequences can be used to construct surrogates from ASCII characters.  This contradicts the definition of strings in section 1.  However, the usage of this escape sequences to represent arbitrary binary data appears to be common enough (JavaScript only recently grew support for binary data, and this support will not be retrofit into JSON) that a clarification in the specification is required.

Please separate these two items clearly.  Thanks.

Grüße, Carsten