On 01/26/2010 08:12 AM, Avi Kivity wrote:
On 01/26/2010 04:05 PM, Anthony Liguori wrote:
On 01/26/2010 07:55 AM, Avi Kivity wrote:
The risk is that if we support a private extension (like '') and
then json is officially extended to support a conflicting or similar
syntax with a different meaning, then we cannot advance to the next
revision of json without breaking compatibility.
The paragraph I quoted from the RFC seems to suggest that the authors
of JSON boxed themselves in with respect to extending JSON. The
reason being that a conforming implementation is given free reign to
extend with "non-JSON forms or extensions". That would seem to
prevent any extension.
A json generator is required to generate conforming text. So there
are three choices:
- reject 's
- unofficially accept 's, nonconforming generators break if json
changes, nor our problem
- officially accept 's, look stupid when json changes
Keep in mind, JSON is a proper subset of ECMAScript which means the
likelihood of extension going outside of ECMAScript would be
extremely unlikely. I don't expect JSON is ever going to change.
Who knows? Let's not take unnecessary risks.
Keep in mind, I've already agreed to not allow '' strings for external
JSON. The only reason the thread's still alive is because we like to
argue apparently :-) Single quoted strings are not sufficiently useful
to warrant taking any risks here.
Regards,
Anthony Liguori