Re: [core] Comments on draft-ietf-core-yang-cbor-00

[Juergen Schoenwaelder <[e-mail]>]

On Sat, May 21, 2016 at 08:29:33AM -0700, Andy Bierman wrote:
> 
> This is a legal union type even though in XML type 'int32' would never be
> used
> because 'string' would match everything.
> 
>  leaf foo {
>    type union {
>       type string;
>       type int32;
>    }
>  }
> 
> This is a problem for config nodes:
>    JSON/CBOR client sets { "foo" : 42 }
>    XML client reads back <foo>42</foo>
> 
> JSON (and CBOR) will send { "foo":42 }, not { "foo":"42" } if they intend to
> send the number, but XML cannot do the same.  There is no problem
> unless XML is being used in addition to JSON or CBOR.
> 
> Some solution choices:
> 
>   1 - do nothing and let XML and JSON/CBOR be broken together
>      (solution is do not use XML and JSON/CBOR together in the same server)
>   2 - force receiver to convert JSON/CBOR to a string for union type
> matching
>   3 - tag the data somehow so the receiver knows which union member type is
> being sent
>   4 - write YANG guidelines on how to design union types that will produce
> the same
>      match with all encoding schemes

I think 2 is the right short-term approach. It does not mean an
implementation has always to convert to a string but an implementation
needs to behave as if it has converted to a string.

> IMO CBOR has to solve the enumeration and bits issues with (3) but
> for the problem in general, just do (4)

Option 3 may be a long-term solution but this is non-trivial to work
out and to introduce with a backwards compability perspective in mind.

/js

-- 
Juergen Schoenwaelder           Jacobs University Bremen gGmbH
Phone: [phone]         Campus Ring 1 | 28759 Bremen | Germany
Fax:   [phone]         <http://www.jacobs-university.de/>