Re: New JSON Parser

I think it would be worth saying a bit about what kinds of errors are
forgiven, especially given the results of Jon's testing.

Procedures that parse JSON documents generally return a meaningful
reference to the "root" node when the JSON is valid. That reference is
generally required in order to retrieve any meaningful content from the
document. However, parsers may return a null pointer or meaningless
reference in the event that the JSON contains "unexpected" characters.

My inclination is to always return a meaningful reference regardless of
unexpected characters found in the stream, so that any valid content can be
extracted or reported. That's my idea of error forgiveness. On the other
hand, I think it might be worthwhile to report the position where the first
unexpected character was encountered in the stream.

In the case of Jon's example:

[ "name" : X , "another" : 12 ]

The colon character encountered in position 9 would not be expected.