Re: [netconf] HTTPS notifications (draft-ietf-netconf-https-notif)

[Martin Bjorklund <[e-mail]>]

"Eric Voit (evoit)" <[e-mail]> wrote:
> > From: Martin Bjorklund, September 24, 2019 9:56 AM
> > 
> > "Eric Voit (evoit)" <[e-mail]> wrote:
> > >
> > >
> > > > > From the email I sent Martin on Sep 9th, each POST MAY contain
> > > > > more than one notification:
> > > >
> > > > .... to which I replied:
> > > >
> > > >   I'm not so sure about the streaming though.
> > > >   Perhaps pipelining is the right mechanism.  For "bulk" sending, the
> > > >   "bundled-message" defined in
> draft-ietf-netconf-notification-messages
> > > >   seems right.
> > > >
> > > > To clarify: if the client just sends a stream of notifs it becomes a
> > > variant of
> > > > SSE.  The server doesn't know when the stream will end, and thus
> > > > cannot simply close the session.  You probably want to indicate
> > > > end-of-message somehow (like in SSE).  And the content type in the
> > > > example below cannot be "application/yang-data+xml", since it is not
> > > > a valid XML instance document; you'd have to invent a new media type
> > to indicate the streaming.
> > > >
> > > > I think we should stick to simple HTTP where each notif is POSTed,
> > > > as in
> > > your
> > > > diagaram above.  With HTTP pipelining you can do:
> > > >
> > > > ------> establish TCP
> > > > ------> establish TLS
> > > > ------> Send HTTPS POST message with YANG defined notification 1
> > > > ------> Send HTTPS POST message with YANG defined notification 2
> > > > <-----Send 204 (No Content) for 1
> > > > <-----Send 204 (No Content) for 2
> > > >
> > > > ------> Send HTTPS POST message with YANG defined notification 3
> > > > <-----Send 204 (No Content) for 3
> > > >
> > > >
> > > > If the server wants to send multiple notifs at once, it can use
> > > > "bundled- message".
> > >
> > > This seems a reasonable approach.
> > > draft-ietf-netconf-notification-messages
> > > has several advantages:
> > > (1) can push multiple YANG notifications at once
> > > (2) can include the subscription-id in notifications when subscribing
> > > to a stream.  (Right now including an explicit subscription-id is only
> > > available when subscribing to a datastore.)
> > > (3) includes methods to discover lost/dropped notifications
> > >
> > > Two things which would need to be worked:
> > > (1) discovering receiver support for bundled notifications.  (As some
> > > form of understanding/verifying configured receiver support over HTTP
> > > is already needed, this is a topic which perhaps can be merged into
> > > that.)
> > 
> > This can be configured or auto-detected.  For auto-detection, perhaps we
> > can use HTTP OPTIONS with the server returning a special body or header to
> > indicate capabilities such as this one?
> 
> The method previously proposed for HTTP configured subscriptions was an "OK"
> to the POST of a <subscription-started>.  The "OK" had to be received prior
> to sending YANG notifications.  If the authors want to build upon
> this

This is now a WG document, so it should be discussed in the WG.

> , I
> could see HTTP OPTIONS as a way to provide receiver capabilities.

Or perhaps simply do a GET to the same URI as you POST to, and in
return get a sx:structure with the capabilities of the server.

> Some
> discussion about what capabilities would be advertisable as options would be
> worthwhile.

Yes.

> Also worth discussion would be any intersections with
> NETCONF/RESTCONF capabilities discovery/exchange.

Sure.  But this is a new protocol, with a new (limited, hopefully) set
of capabilities.


/martin


> 
> Eric
>  
> > > (2) Completion of draft-ietf-netmod-yang-data-ext
> > 
> > Yes!  IMO it is ready for WGLC; I'll ping the chairs again.
> > 
> > 
> > /martin
> > 
> > 
> > 
> > >
> > > Eric
> > >
> > > > /martin
> > > >
> > > >
> > > > >
> > > > >     POST /some/path HTTP/1.1
> > > > >     Host: my-receiver.my-domain.com <http://my-receiver.my-
> > > > domain.com/>
> > > > >     Content-Type: application/yang-data+xml
> > > > >
> > > > >     <notification
> > > > >       xmlns="urn:ietf:params:xml:ns:netconf:notification:1.0">
> > > > >       <eventTime>2019-03-22T12:35:00Z</eventTime>
> > > > >       <foo xmlns="https://example.com
> > > > >       <https://example.com/>/my-foobar-module">
> > > > >         ...
> > > > >       </foo>
> > > > >     </notification>
> > > > >
> > > > >     <notification
> > > > >       xmlns="urn:ietf:params:xml:ns:netconf:notification:1.0">
> > > > >       <eventTime>2019-03-22T12:35:00Z</eventTime>
> > > > >       <bar xmlns="https://example.com
> > > > >       <https://example.com/>/my-foobar-module">
> > > > >         ...
> > > > >       </bar>
> > > > >     </notification>
> > > > >
> > > > >     <notification
> > > > >       xmlns="urn:ietf:params:xml:ns:netconf:notification:1.0">
> > > > >       <eventTime>2019-03-22T12:35:00Z</eventTime>
> > > > >       <baz xmlns="https://example.com
> > > > >       <https://example.com/>/my-foobar-module">
> > > > >         ...
> > > > >       </baz>
> > > > >     </notification>
> > > > >
> > > > >
> > > > > With response:
> > > > >
> > > > >       HTTP/1.1 204 No Content
> > > > >       Date: Thu, 26 Jan 2017 20:56:30 GMT
> > > > >       Server: my-receiver.my-domain.com
> > > > > <http://my-receiver.my-domain.com/>
> > > > >
> > > > >
> > > > > Kent // co-author
> > > > >
> > > > >
> > > >
> > > > _______________________________________________
> > > > netconf mailing list
> > > > [e-mail]
> > > > https://www.ietf.org/mailman/listinfo/netconf