Re: [Json] Generators

[Nico Williams <[e-mail]>]

On Mon, Jun 10, 2013 at 12:35 PM, Paul Hoffman <[e-mail]> wrote:
> <no hat>
> On Jun 10, 2013, at 12:38 AM, Stefan Drees <[e-mail]> wrote:
>>   Generators MUST NOT duplicate names in objects if they can avoid or
>>   detect such duplication.
>
> +1 to the others who have said "yuck" to the poor use of 2119 MUST here. Further, this is a significant change from both RFC 4726 and ECMAScript. Proposed new wording:

Agreed.  If we want a MUST then we need to distinguish types of
encoders/parsers ("non-streaming ... MUST ...; streaming ...
MAY/SHOULD ...").  That, I think, is why it might be valuable to
distinguish types of encoders/parsers: so we can apply requirements to
one common implementation type that we cannot to the other.

The RFC2119 approach to this is to say SHOULD/SHOULD NOT and explain
what exceptions might be made.

Six of one, half a dozen of the other... I'd prefer a construction
that ends up having a MUST because I suspect it will be understood
more readily by future implementors, but either construction will do
in the end.

Although, to be extra careful here, since this sub-thread is about
encoders, I'm not sure that a non-streaming encoder can be expected to
detect duplicates in a has table (which would normally be an error at
insertion time, but let's say it's happened).  And so perhaps we can't
really distinguish *encoder* types, only parser types.

Nico
--