Re: [Json] draft-williams-json-text-sequence-00

[Nico Williams <[e-mail]>]

On Mon, Mar 17, 2014 at 12:47 AM, Manger, James
<[e-mail]> wrote:
>> Title:          JavaScript Object Notation (JSON) Text Sequences
>> http://tools.ietf.org/html/draft-williams-json-text-sequence-00
>
> Nico,
>
> Did you really mean to allow commas as separators (%2C in ABNF)?

No, I hadn't (I'd just argued in response to Phillip that it was a bad
idea).  However, I'm not as convinced of it as I'd like to be.  I
think there may indeed be some value in it.  Suppose you expect a JSON
text to be a very large (possibly never-ending) array... you might
parse it as a sequence by skipping the initial '[' and going from
there.  You'd get streaming parsing of the sequence without streaming
parsing of the individual JSON texts.  That seems worthwhile as a way
of dealing with any cases that might exist of very large arrays.

> I would be tempted to say:
>
>   A JSON sequence is a concatenation of any number of JSON values,
>   remembering that each value (matching <JSON-text> in [RFC7159])
>   can be preceded and followed by insignificant whitespace.
>   Whitespace MUST be present between two JSON values when both
>   values are true, false, null, or a number. Whitespace between
>   values is allowed, but not required, when either value is an
>   object, array, or string.

I like that, but it's important that the separator always _follow_
each text.  Whitespace may also precede a text, sure, but that's not
important here.

>   Concatenating JSON values, each terminated with a newline (U+000A),
>   is a simple way to match the syntax rules.

Yes.

>   An empty text and texts consisting only of whitespace are both valid
>   JSON sequences of zero JSON values.

s/empty text/empty sequence/, but yes, that's nice.

Nico
--