On 06/12/2013 12:13 AM, Craig Ringer wrote:
> On 06/12/2013 08:42 AM, Andrew Dunstan wrote:
>> If we work by analogy to Postgres' own handling of Unicode escapes,
>> we'll raise an error on any Unicode escape beyond ASCII (not on input
>> for legacy reasons, but on trying to process such datums). I gather that
>> would meet your objection.
> I could live with that if eager validation on input was the default, but
> could be disabled by setting (say) compat_lazy_json_validation = on .
> I don't like the idea of leaving us saddled with weak validation just
> that's what we've got. It's been an ongoing source of pain as UTF-8
> support has improved and I'd really like a way to avoid semi-valid JSON
> making it into the DB and causing similar problems.
>
I think it's rather too late in the cycle to be proposing new GUCs. We
can revisit this for 9.4 perhaps.
cheers
andrew