Re: [core] Comments on draft-ietf-core-yang-cbor-06

[Robert Wilton <[e-mail]>]

On 09/07/2018 17:20, Carsten Bormann wrote:
> On Jul 9, 2018, at 18:09, Robert Wilton <[e-mail]> wrote:
>> ambiguity as to whether the top level nodes are using absolute or delta sids
> We don’t need to be able to include “absolute SIDs”(1) into the delta positions in the encoding.
> The deltas are always deltas.  They just happen to be relative to a context SID at the top of the tree; mostly, that is 0 (but that’s what we are also discussing here).
Apologies if my terminology was confusing.

I agree, that we are discussing whether the parent of the top level 
nodes should always have a SID of 0, or otherwise if it should be 
explicitly specified.

For me, one of the big advantages of CBOR over schema based encodings 
like protobuf, is that the structure of the encoded data is self 
describing.  I think that it is worth spending a few extra bytes in the 
encoding to get this property.  So, if I get a bad CBOR message that I 
can't interpret then I can easily convert it into a human readable 
structure to debug the problem (or know the CBOR itself is 
invalid/corrupt).  I see that requiring external knowledge for the 
parent SID of the top level nodes seems to take a step away from that.

E.g. if the device accidentally gives a response for the wrong node in 
the tree, and if the response relies on the SID in the request to be 
fully decodeable then the response will somewhat look like garbage and 
it will be harder to debug.  Is this worth saving a couple of bytes for?

Thanks,
Rob


>
> Grüße, Carsten
>
> (1) there is no need for this term; SIDs are SIDs and SID deltas are SID deltas.  The map key positions in YANG-CBOR are always SID deltas.
>
>
> .
>