Re: [Json] Unpaired surrogates in JSON strings

[Tatu Saloranta <[e-mail]>]

On Fri, Jun 7, 2013 at 11:09 AM, Nico Williams <[e-mail]>wrote:

> On Fri, Jun 7, 2013 at 5:42 AM, Bjoern Hoehrmann <[e-mail]>
> wrote:
> > Actually there are many good reasons for having unpaired surrogates in
> > JSON documents. A simple example would be a test suite for string APIs.
>
> Or what heck, if ECMAScript allows any 16-bit values, 0x0000..0xFFFF
> to be used (escaped as \uXXXX if necessary) then one very useful use
> of that is encoding binary data: when parsing you know if you have
> binary data when you see any 16-bit code units that don't make any
> sense in Unicode text.  Not that I'm advocating this... but if we did
> allow this then it wouldn't preclude us from saying that a string of
> text must not include unpaired surrogates.


Not really. Since most content is exchanged as UTF-8, it gives you 3-to-2
conversion, which is no better than Base64 (4-to-3), nor necessarily any
faster. It would only help when using UTF-16.
APIs would expose this as text, requiring conversion or coercion (depending
on platform), so it would not even be a short-cut.

-+ Tatu +-