On Wed, Aug 14, 2013 at 6:07 AM, Carsten Bormann <cabo@xxxxxxx> wrote: > On Aug 14, 2013, at 13:40, Randy Bush <randy@xxxxxxx> wrote: > >>> YANG seems to be incompatible with CBOR. >> >> so what does that say about yang, yang's suitability for netconf, cbor, >> and cbor's suitability? > > Good question. I'm not sure the jury is even out on that yet. > > Yaron reminded me that netconf is using XPath for various purposes, so a CBOR rendition of netconf would need to do the same that you would need for doing netconf with JSON: either replace XPath with something else (indeed causing some incompatibility), or explain how to use XPath with the CBOR/JSON translation of the YANG model (which might be interesting). > > CBOR certainly has no trouble carrying around the data described by a YANG model. > (Indeed, CBOR's decimal fractions were inspired by YANG's.) > But then, I don't have an opinion whether adding a binary object representation to the netconf ecosystem is on the agenda. > I am not sure this is an actual problem. NETCONF uses full XPath only if the :xpath capability is supported. The "select" expression represents the data to be returned from the conceptual datastore and possibly the state/statistics within the device. The select expression is not related to the message on the wire. There is also an "error-path" field that is returned in errors that contains the absolute path of the message or data node that caused the error. Neither of these message fields should cause an encoding problem. They are just strings at the message layer. > Grüße, Carsten > Andy