Re: [Netconf] netconf call home connection type

[Martin Bjorklund <[e-mail]>]

Kent Watsen <[e-mail]> wrote:
> Hi Martin,
> 
> Catching up on mail from last week.
> 
> 
> >> Also, it seems that, if we want YP+SN to be able to use call-home 
> >> for configured subscriptions, then we need to update the SN draft
> >> to A) define the "activate-configured-subscription" RPC
> >
> > I started to write that I don't think this belongs to the SN draft,
> > since it is NETCONF-specific.  But after thinking some more about it I
> > think it makes sense to define this operation in the SN draft.
> >
> > For RESTCONF (not a new HTTP protocol) this rpc could be used
> > by augmenting a 'location' leaf to its output.
> >
> > It is probably useful for any protocol that does RPCs.  But for a
> > notif-specific transport like a UDP-based protocol, that doesn't do
> > normal RPCs at all, an "activate-configured-subscription" RPC doesn't
> > make sense.  But that is ok.
> 
> Right, but there is an issue of timing.  There is a question if this
> is needed now or if it can be added in a SN-bis.  Adding the RPC would
> easy, but adding the identity (rfc8071bis?) would be more involved.
> 
> Note that said RPC would be used for call-home connections (i.e., where
> the roles are reversed), but unlikely to be used for regular connections
> (i.e., when the publisher is the NC/RC-client).  The RPC would need to
> be documented as only being relevant to call-home connections.

Yes.

> >> and B)
> >> define an identity off of the base "call-home-reason" identity
> >> defined in the aforementioned I-D.  This is needed before any
> >> notif draft can define support for "configured" subscriptions,
> >> right?
> >
> > Yes, but if we define this identity now in SN it would delay SN.
> > Better to define it later either in SN-bis or directly in a separate
> > document.
> 
> So, configured subscriptions in SN is known to be broken for call-home
> connections?  Should the SN draft identify this as a known limitation?

If we have the rpc, it is less broken.  I would prefer to define the
rpc in SN now.  Since we decided to work out the details for
configured subscriptions now in SN, I don't think we can ignore this
issue.

The identity is less important imo.  And of course, the identity is of
no use until the "call-home-reason" data model has been defined...


/martin