Re: [manet-dlep-rg] Session iniation and discovery

[Martin Duke <[e-mail]>]

I'm late but I endorse what Rick and Stan came up with.


On Fri, Nov 22, 2013 at 10:28 AM, Teco Boot <[e-mail]> wrote:

> +1 judgement as Stan.
>
> You know my preference: make radio the TCP server.
>
> Op 22 nov. 2013, om 17:58 heeft Rick Taylor <
> [e-mail]> het volgende geschreven:
>
> > Extra note:
> >
> > d) The IANA assigned port for DLEP is used for both multicast UDP and
> TCP.  It is the same port number.
> >
> > ________________________________________
> > From: manet-dlep-rg [[e-mail]] on behalf of Rick
> Taylor [[e-mail]]
> > Sent: 22 November 2013 16:57
> > To: [e-mail] Group, ([e-mail])
> > Subject: [manet-dlep-rg] Session iniation and discovery
> >
> > Gents,
> >
> > Am I right in thinking we are achieving some kind of consensus on
> session initiation and discovery?  My understanding of the proposal is:
> >
> > Discovery:
> >
> > 1) Router SHOULD send 'Advertise' UDP messages on IANA assigned
> link-local multicast address:port at periodic interval ANNOUNCE_INTERVAL.
> > 'Advertise' messages MUST contain the DLEP version TLV.
> > 'Advertise' messages MAY contain alternate TCP address TLV.  (E.g. ipv4
> fallback)
> > 'Advertise' messages MAY contain alternate TCP port TLV.
> > 'Advertise' messages MAY contain secondary reliable transport protocol
> endpoint address TLV.  (E.g. SCTP address)
> >
> > Initiation:
> >
> > 0) Router listens on a TCP port, the port SHOULD be the IANA assigned
> DLEP port.
> >
> > 1) Modem connects to a TCP endpoint either discovered from 'Advertise'
> messages, or from alternate discovery mechanism (e.g. mDNS), or a-priori
> configuration.
> >
> > 2) Modem MUST send 'Initialize' message.
> > 'Initialize' message MUST contain DLEP version TLV.
> > 'Initialize' message MUST contain Identification TLV.
> > 'Initialize' message MUST contain for all mandatory DLEP metric TLVs
> with values.
> >
> > 2) Router MUST reply with 'Accept' message, or shut down the connection.
> > 'Accept' message MUST contain DLEP version TLV.
> > 'Accept' message MUST contain Identification TLV.
> > 'Accept' message MUST contain Status TLV.  This will indicate: Success,
> Reject, etc... (TBD)
> > 'Accept' message MAY contain secondary reliable transport protocol
> endpoint address TLV.  (E.g. SCTP address)
>
> Isn't this in the transport layer?
>
>
> >
> > 2a) If Status TLV is a failure: Router MUST and Modem SHOULD close the
> TCP connection.
>
> I think it is better to specify which node MUST close the connection.
> That's the one that sends the Reject. The FIN follows. Why SHOULD the other
> start closing the connection? It can wait on FIN, it MUST come. Or wait for
> heartbeat dead time, as safeguard.
>
>
> > 2b) If Status TLV is a success: Session is established.
> >
> > 3) Router MAY stop any active discover process.
>
> Not sure what you mean, if TCP server is on modem.
> Router SHOULD continue to listen to discovery packets.
> FYI: new BGAN terminals support some kind of bonding. Some deploy an RF
> link with Tx&Rx-only modems.
>
> >
> > 4) Router MAY stop listening for more connections on the TCP port.
> >
> > Notes:
> >
> > a) Stan probably has different names for the messages.
> > b) I have made the Router the TCP server in this example, because it
> forces the modem to announce it's capabilities/TLVs in the 'Initialize'
> message, and I believe it's the way Stan is leaning.
> > c) If the roles are reversed then the 'Accept' message must carry the
> mandatory TLVs, and an extra 'Unacceptable' message is required from the
> Router to Modem explaining why the Router closed the connection if it
> doesn't like the TLVs from the Modem.
>
> So this is the three-way handshake.
> I didn't understand why the Router didn't send the full set of supported
> TLVs. OK, Ik know this is more relevant for link metrics from modem to
> router. But for protocol specification, I haven't a clear picture what we
> designed.
>
> Teco
>
>
> >
> > Does this sound right?  Or am I way off?
> >
> > I'd just like to get this part fixed as I have more topics up for
> debate, and I reckon we could announce this as progress on the WG list.
> >
> > Rick
> > _______________________________________________
> > manet-dlep-rg mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/manet-dlep-rg
> > _______________________________________________
> > manet-dlep-rg mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/manet-dlep-rg
>
> _______________________________________________
> manet-dlep-rg mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/manet-dlep-rg
>