Re: [Json] JSON Log file encoding JSON-L

[Nico Williams <[e-mail]>]

On Tue, May 6, 2014 at 9:05 PM, Manger, James
<[e-mail]> wrote:
>> Thank you Carsten.  I think I'll do the following: a) add RS to the
> whitespace list for JSON text sequence separators, b) RECOMMEND
> surrounding JSON texts with RS in log-like sequences.  Then JSON
> sequences will be meet the needs of logs.
>
> Don’t do that. Surely no one uses RS anymore; and certainly no one who uses JSON. It doesn't make much sense to use SEPARATOR TWO without using ONE, THREE and FOUR.

RS would be used to help distinguish incomplete (because interrupted) entries.

> We now have Nico’s JSON Text Sequences [draft-williams-json-text-sequence],

It's actually a WG work item now, since the charter now includes this work.

> and Phillip’s JSON Log Format (JSON-L) [draft-hallambaker-jsonl].
>
> They are basically the same. Can we just keep one? Probably the more generic JSON Text Sequences.

I agree.

However, PHB has a point about recovery from corruption.  My proposed
JSON Sequence work does nothing about that.  An OPTIONAL RS would.

I.e., loggers could (but wouldn't have to) emit RS JSON_TEXT RS NL to
make resynchronization possible.  Other use cases could just emit
JSON_TEXT NL (or any other whitespace.

Nico
--