Re: [Netconf] YangPush now

["Eric Voit (evoit)" <[e-mail]>]

> From: Kent Watsen, July 26, 2018 5:29 PM
<snip>
> That said:
> 
>   1) my preference is to have *no* netconf-notif or restconf-notif drafts
>      if possible and, if it is, then this reduced netconf-notif scope is
>      a moot point...as dynamic subscriptions just work, for both NC and
>      RC automatically.

I just took a cut at looking at what is removed for configured subscriptions from NETCONF-notif,  The result is at:
https://github.com/netconf-wg/notif-netconf/blob/master/draft-ietf-netconf-netconf-event-notifications-11-configured-removed.txt 

To me there are normative requirements specific to NETCONF where I believe we need to go with Option 2.


>   2) however, if netconf-notif MUST be defined, or else NC-based dynamic
>      subscriptions are undefined, then it follows that RC-based dynamic
>      subscriptions are also undefined, unless a restconf-notif is also
>      published.  And since the WG treats both protocols symmetrically,
>      I added RC-notif (for dynamic-only) as also being a needed.  Makes
>      sense and should be fairly easy, right?  [hint: it's just restconf]

I think this makes sense.   There will still be some decisions to make as Section 3.4 of current RESTONF-notif asserts HTTP2 as preferred for dynamic subscriptions (because it gets rid of SSE, and enables QoS features.)  Section 3.5 shows what could be done with existing RESTCONF over HTTP1.1 and SSE, which I believe is Andy's preferred method. 

Eric

> We should see if (1) is possible, and if not, go for (2).  This shouldn't
> require any more consensus than we have already.  That said, as always, if
> anyone objects, please speak now!
> 
> 
> Thanks,
> Kent
> 
> 
> 
>