Re: [core] YANG to CBOR mapping

[Carsten Bormann <[e-mail]>]

Ladislav Lhotka wrote:
>> On 19 Nov 2015, at 16:27, Juergen Schoenwaelder <[e-mail]> wrote:
>>
>> On Thu, Nov 19, 2015 at 03:54:57PM +0100, Carsten Bormann wrote:
>>> So I think that Peter's reminder to stay close to
>>> draft-ietf-netmod-yang-json -- unless we do need to deviate -- is quite
>>> germane.
>>>
>> I do not understand what "stay close to draft-ietf-netmod-yang-json"
>> means. I think this requires an explanation.

"-- unless we do need to deviate --"

>> To give you an example: draft-ietf-netmod-yang-json is following
>> I-JSON rules and one of them says that I-JSON numbers are limited in
>> precision. As a consequence, 64-bit numbers are I-JSON encoded as
>> strings for I-JSON compliance. I think CBOR does not require this
>> since it does not have this particular restriction. So do we now
>> continue to do this even though CBOR does not suffer from I-JSON
>> limitations?

No.  Here we clearly need to deviate to reduce unnecessary complexity.

>> I think my preference would be a clean mapping YANG to CBOR and not a
>> mapping YANG to I-JSON to CBOR that carries I-JSON restrictions over
>> to CBOR. The simple reason is that the longer the transformation
>> chain, the more arcane rules you accumulate.

I completely agree.  What I was trying to say was that the two mappings
should stay close, at least where the mappings can reasonably use the
common data model of JSON and CBOR.

The one place where we probably will have additional divergence is the
area of identifiers/selectors, where for COMI the CBOR mapping will need
to provide more flexibility.  Can we decouple this aspect from the other
elements of the YANG-CBOR mapping?  I hope so, but we haven't
demonstrated that yet.

> I agree, it isn't probably necessary to couple the CBOR mapping with JSON as defined in the yang-json draft. Some of the I-JSON restrictions are an ugly compromise, really, and you'd better avoid them. I believe a mapping between CBOR, JSON and XML encodings will be possible anyway.
> 
> Lada
> 
>> But then the CBOR spec has already JSON to CBOR translation rules so
>> YANG to I-JSON to CBOR is already defined somehow. 

... and the result of this two-step process might be useful as a
"default" mapping, which we use unless there is a reason to deviate.

>> Someone needs to
>> compile a list of cases where YANG to CBOR would be different from
>> YANG to JSON to CBOR before we can take an informed decision.

Indeed, and this is work that needs to be done very soon.

Do we need to build a formal design team for this work around the
extraction of the YANG-to-CBOR mapping from the COMI (and COOL) documents?

Grüße, Carsten