Re: [Json] fun with streaming, was The names within an object SHOULD be unique.

[Tim Bray <[e-mail]>]

On Tue, Jul 30, 2013 at 9:07 AM, John Levine <[e-mail]> wrote:

> >> I've never seen an implementation ...  Better, can you give an example
> of such an
> >> implementation?
> >
> >I haven't either, but I could see the uses of an implementation


I don’t think “I could see the uses of” is a good reason for writing an
Internet Standard.  Once again as with naked surrogates, there’s an
invisible class of alleged implementors of JSON sending/receiving code for
whom dupe-checking is intractable; except for we haven’t heard from anyone
whose application actually depends on this behavior.

Has anyone any personal knowledge of an app whose correct function depends
on the use of duplicate keys?  -T




> that
> >streamed arrays but not objects (of course, it could hardly stream an
> >array that was within an object).
>
> It has been my impression that the reasons that people use streaming
> implementations are either that it's a tiny system that doesn't have
> room to buffer full arrays and full objects, or else it's a realtime
> application that needs to see the first item in the array or object
> before the rest of it arrives.
>
> I understand that there are plenty of implementations that do not have
> those constraints, but since there are apparently also plenty of
> implementations that do have one or both, Paul's short list of
> possible treatments of duplicate names is the best we can do.
>
> _______________________________________________
> json mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/json
>