John Keiser

Oh, so the idea is that if you ask for an integer, we treat -0 as INCORRECT_TYPE?

Yeah. I feel sort of positive about the principle that we refuse to parse to a type if it loses information that would be preserved if we parsed to a...

Yeah, this is for On Demand only. Even with document stream, you have the text of the document around while the document is being processed.

This is a definite optimization thing. I'll move to 1.0, but I don't think we'd block 1.0 for it either.

Yeah, the issue is much smaller here. In fact, the user can choose to get the raw bytes of the JSON string, if they want, by asking for the raw...

This is definitely a great approach for DOM. The big difficulty is On Demand: because the user drives the parsing (which is unlined), I can't see a way to "switch...

I no longer think it's a good idea to do this exactly as supported: we should absolutely have `.default()`, but we should rely on out of order field support to...

This is a lot less terrible now that object lookups are order-insensitive by default. However, having `default()` is still a good idea.

Very related: #670.

A couple of things here seem like we may want to change in terms of featureset: * We should definitely check if the document has extra crap after the end...