Re: [core] CBOR Encoding of Data Modeled with YANG

[Ladislav Lhotka <[e-mail]>]

> On 11 Dec 2015, at 16:22, Michel Veillette <[e-mail]> wrote:
> 
> I plan to propose an update to section 3.1.7 based on the text of draft-ietf-netmod-yang-json-06 section 6.8.
> I can either use the module name or the module ID as namespace identifier.
> For compactness, I intent to initially propose the module ID.
> 
> For example, if the module ID registered for iana-if-type is 72, we will have "72:ethernetCsmacd".

Sure, but my impression from sec. 3.1.7 is that the encoding attempts to alleviate the need for determining the chain of identity derivations from a data model and simply encode the chain in the identityref value. But maybe I am just confused.

Cheers, Lada 

> 
> Is this make sense?
> 
> Regards,
> Michel
> 
> -----Original Message-----
> From: Ladislav Lhotka [mailto:[e-mail]] 
> Sent: December-11-15 10:05 AM
> To: Michel Veillette <[e-mail]>
> Cc: Carsten Bormann <[e-mail]>; Core <[e-mail]>
> Subject: Re: [core] CBOR Encoding of Data Modeled with YANG
> 
> 
>> On 11 Dec 2015, at 15:50, Michel Veillette <[e-mail]> wrote:
>> 
>> Hi Ladislav
>> 
>> About tagging and tag 4 specifically, the overhead require by them will need to be justified.
>> We need to identify which specific applications require schemaless processing of unknown data nodes.
> 
> Fair enough.
> 
>> 
>> About identitities
>> This issue have already been mentioned by Juergen and the current 
>> intent is to adopt the same representation as 
>> draft-ietf-netmod-yang-json-06
> 
> Juergen questioned the need for a new namespace notation but if I understand correctly the example in sec. 3.1.7, in
> 
>  iana-interface-type.ethernetCsmacd
> 
> there is no namespace at all: iana-interface-type is an identity, and so is ethernetCsmacd.
> 
> Lada
> 
>> 
>> Regards,
>> Michel
>> 
>> -----Original Message-----
>> From: Carsten Bormann [mailto:[e-mail]]
>> Sent: December-11-15 4:55 AM
>> To: Ladislav Lhotka <[e-mail]>
>> Cc: Michel Veillette <[e-mail]>; Core 
>> <[e-mail]>
>> Subject: Re: [core] CBOR Encoding of Data Modeled with YANG
>> 
>> my 2 µJ for now...
>> 
>>> 1. Would it make sense to use semantic tagging (major type 6) to
>>>  distinguish numbers or strings with special semantics? Possible
>>>  candidates that come to my mind are instance-identifier and
>>>  identityref values.
>> 
>> In general, the assumption has been that semantic information would come out of the YANG schema.  Is there a case where that is not so, and we need to represent it at the representation level?
>> 
>>> 2. Why is the "decimal64" type encoded as an integer and not as "decimal
>>>  fraction" (major type 6, tag 4)? Whilst the decimal value can be
>>>  obtained by using the leaf's definition,
>> 
>> Because we can get the exponent value out of the schema, we don't need to transmit it.
>> 
>>> perhaps it may simetimes be
>>>  useful to be able to correctly interpret the values without using a
>>>  schema.
>> 
>> Right.  We probably have to make up our mind how useful YANG-based information should be in a schemaless environment.  If we want to make it useful, that requires quite some different thinking than what we have now.
>> 
>>> 3. I think it is necessary to include namespaces/module names in the
>>>  encoding of identitities (sec 3.1.7). It is probably not very likely
>>>  in that particular example, but somebody could define another
>>>  "iana-interface-type" identity in another module. Identities were
>>>  designed to be extensible in a decentralised fashion, so the CBOR
>>>  encoding should IMO also represent the namespace.
>> 
>> We should take note of this problem and put it as one more objective for the identifier discussion (the hashes vs. structured IDs vs. ... one).
>> 
>> Grüße, Carsten
> 
> --
> Ladislav Lhotka, CZ.NIC Labs
> PGP Key ID: E74E8C0C
> 
> 
> 
> 

--
Ladislav Lhotka, CZ.NIC Labs
PGP Key ID: E74E8C0C