Re: [Json] Using a non-whitespace separator (Re: Working Group Last Call on draft-ietf-json-text-sequence)

["Manger, James" <[e-mail]>]

> Both require escaping in strings.  The difference is that LF can appear as
> insignificant whitespace, whereas RS cannot.

Correct.

>  So to achieve the same thing
> as RS achieves (restartable/correctable writes), it's simply necessary never
> to spit out an insignificant LF.

Incorrect.
It is not *necessary* to omit insignificant LF.
It is just *easier* for parsers if every LF terminates a single JSON value (and is never other insignificant whitespace).

With LF as the terminator but also allowed as insignificant whitespace, <endchar> LF <startchar> unambiguously separates JSON values.
This allows you to restart from anywhere in the middle of a sequence.
This allows you to restart after one value is corrupted or truncated.
If you have had some corruption (so we are already in a niche case) and you really don't want to miss the very next value, you are still ok when values are objects or arrays (as log entries will invariably be). The end of the next (non-corrupted) value is found as normal (<endchar> LF <startchar>), then look back for the matching { or [ that started it.

In my judgement it is worth accepting a little extra complexity for JSON sequence parsers (eg having to look for <endchar> LF <startchar>) to make it much easier for JSON sequence producers (no awkward control characters, can be pretty-printed, can be cut-n-paste, can be manually edited, can cat multiple files together, can wrap long lines...).

--
James Manger