Re: [Netconf] netconf-binary-encoding comments

[Andy Bierman <[e-mail]>]

On Fri, Jul 6, 2018 at 1:39 PM, Kent Watsen <[e-mail]> wrote:

>
>
> > Working on binary (gNMI).
>
> > Very interested in CBOR+SID as a better solution.
>
> > The NETCONF WG is done wrt/ RESTCONF.
>
>
>
> Collections are still pending and, if RESTCONF is ever to supersede
> NETCONF, there would need to be something like a commit-confirmed, or
> perhaps this is already taken care of via /restconf/operations/ietf-
> netconf:commit?
>
>
>
> BTW, I notice that nowhere in the nmda-restconf draft does it mention that
> the auto-commit feature of the "unified" datastore is no active and that
> clients must(?) now use /restconf/operations/* to do things.  Is this an
> oversight?
>
>
>
>
>


Why would the unified API be considered harmful to the Internet if NMDA
datastores are also present?
I see no reason to remove existing APIs just because new ones are added.

With NMDA, a client has to edit candidate and then use POST
/restconf/operations/commit to finish the edit,
which requires 2 steps (or 6 if locking is used). With unified (RFC 8040) 1
step is needed and no locking
is required.

Andy


> CORE WG will standardize YANG to CBOR and SID.
>
> > Seems like any use of CBOR in NETCONF, YANG Push, or RESTCONF
>
> > is gated on the completion of these RFCs.
>
>
>
> Regarding YANG-push, perhaps the NETCONF WG should consider a "coap-notif"
> transport layer draft and, by extension, the CORE WG should consider
> defining a "coap-client-server" draft.
>
>
>
>
>
> > Andy
>
>
>
> Kent // contributor
>
>
>
>
>
>
>