Re: [manet-dlep-rg] DLEP session establishment

["Stan Ratliff (sratliff)" <[e-mail]>]

On Nov 12, 2013, at 10:35 AM, Henning Rogge <[e-mail]>
 wrote:

> I must admit I like our original solution more, the one where the
> radio initiates the contact.
> 
> I don't buy in the "complexity of tcp socket" thing, because most DLEP
> radios/modems will only allow a single router to connect to them. This
> should keep the complexity in check.

The additional complexity comes with issuing the listen() call, handling receipt of the listen (which in over 95% of the cases I've seen, requires non-blocking socket mode and use of select() or poll()), and getting the accept() done…. 
I'll try to contact the radio vendors I was talking to a couple of years ago and see how they feel about the radio taking on the role of TCP "server". I don't have an ETA for a response.

Regards,
Stan


> 
> We also talked about leaving things out of DLEP that belong to SNMP
> (or maybe web interfaces)... if the radio can handle SNMP, it can also
> handle a single TCP session.
> 
> Henning Rogge
> 
> On Tue, Nov 12, 2013 at 4:16 PM, Stan Ratliff (sratliff)
> <[e-mail]> wrote:
>> OK, looks like my opinion is "in the rough" on this one…
>> 
>> But I'm still concerned about putting the TCP "server" code into the (probably more) constrained modem device. So, what if we make the *router* issue the Peer Discovery (call it a Peer Advertisement) to an MCAST address/well-known port? That Peer Discovery/Peer Advertisement would contain the unicast IP address/port for the TCP listener. The modem would respond with the TCP SYN…
>> 
>> The router would continue to periodically "strobe" the Peer Discovery/Peer Advertisement ad-nauseum (as much as I hate that approach).
>> 
>> I'm really concerned about modem-side complexity, based on discussions I had a couple of years ago with some radio vendors who were looking at implementing.
>> 
>> Regards,
>> Stan
>> 
>> On Nov 12, 2013, at 7:47 AM, "Taylor, Rick" <[e-mail]> wrote:
>> 
>>> Sorry for the delay, I was out of office yesterday.
>>> 
>>> My thoughts:
>>> 
>>>   Router                                        Modem
>>>   ===================================================
>>> 
>>> 1)                                  TCP socket listen()
>>> 
>>> 2)  <--------------------------- Peer Discovery Message
>>>                             UDP unicast or broadcast?
>>>                                      + Session Cookie
>>>                                    + TCP address/port
>>>          + Alternate reliable protocol endpoint infos
>>> 
>>> 3)  TCP connect()
>>> 
>>> 4)  Initialize (was Peer Offer) ---------------------->
>>>   + Session Cookie
>>> 
>>> 5)  <----------------------------------- Initialize ACK
>>>                                      + Session Cookie
>>>                               + Supported metric TLVs
>>> 
>>> 
>>> My reasoning, often agreeing with others:
>>> 
>>> The Modem 'advertises' its DLEP support, and therefore should
>>> be the one that listens for the TCP connect.
>>> 
>>> A cookie passed between the UDP discovery message and the TCP
>>> connection adds a little security (is this the modem I think I am
>>> connecting to?)  This could be extended to a full signature TLV
>>> in a later RFC.
>>> 
>>> The Peer Discovery message could carry additional reliable
>>> protocol endpoint information for non-TCP transports.
>>> 
>>> The Initialize ACK is the correct place to put the 'default'
>>> metric TLVs, and is sent by the modem.
>>> 
>>> A 3-way handshake seems safer to me.
>>> 
>>> I have a question over whether the Peer Discovery message should
>>> be unicast to a configured destination, or broadcast to all
>>> connected peers on a TBD port.  I prefer broadcast as it is more
>>> ZeroConf, but I can see use-cases for unicast to a configured
>>> destination for more complex topologies.
>>> 
>>> Cheers,
>>> 
>>> Rick Taylor
>>> 
>>>> -----Original Message-----
>>>> From: [e-mail] [mailto:manet-dlep-rg-
>>>> [e-mail]] On Behalf Of Teco Boot
>>>> Sent: 11 November 2013 16:29
>>>> To: Stan Ratliff
>>>> Cc: DLEP Research Group ([e-mail])
>>>> Subject: [manet-dlep-rg] DLEP session establishment
>>>> 
>>>> 
>>>> Op 11 nov. 2013, om 01:55 heeft Stan Ratliff (sratliff)
>>>> <[e-mail]> het volgende geschreven:
>>>> 
>>>>> Also, as to the Discovery: Here's what I'm writing up as we speak:
>>>>> 
>>>>> 
>>>>> Router
>>>> Modem
>>>>> ========================================
>>>>>                                   <------------------------ Peer
>>>> Discovery Message
>>>>> Peer Offer with           ------------------------->
>>>>> unicast IP addr/
>>>>> port for TCP connect
>>>> 
>>>> So this is multicast reply, telling modem to connect?
>>>> Why not TCP connect from router to modem? Makes more sense to me, the
>>>> modem is the peer offering a service.
>>>> Maybe add a TcpPort TLV in the Peer Discovery, this allows other than IANA
>>>> assigned ports.
>>>> 
>>>> 
>>>>> 
>>>>> *Connect on TCP socket. Router has issued TCP "listen",
>>>>>  modem issues TCP "connect" (e.g. Modem is the TCP "client",
>>>>>  router is the TCP "server")
>>>>> 
>>>>> Now, The modem's UDP socket can be closed.
>>>> 
>>>> Please don't.
>>>> 
>>>> 
>>>>> Over the TCP
>>>>> Socket,
>>>>>                                  <------------------------- Peer
>>>> Initialization containing
>>>>> 
>>>> TLVs/default values for ALL
>>>>> 
>>>> supported metric values - all
>>>>> 
>>>> meaning the MANDATORY ones,
>>>>> 
>>>> plus any optional metrics (right now,
>>>>> 
>>>> just Resources) that are supported.
>>>> 
>>>> Yes, here the full set TLV exchange takes place. This should not be in the
>>>> Peer Discovery. That's why I suggested to put this in the Peer Offer.
>>>> 
>>>> 
>>>>> 
>>>>> Peer Initialization ACK ---------------------->
>>>>> MAY contain optional
>>>>> Layer 3 (address) TLVs
>>>> 
>>>> I'm fine with three way handshake, not with this two way. Or use TCP
>>>> disconnect when modem modem is not willing to accept first message from
>>>> router.
>>>> 
>>>> Teco
>>>> 
>>>>> 
>>>>> .... And, everything from there is basically the same as before.
>>>> 
>>>> _______________________________________________
>>>> manet-dlep-rg mailing list
>>>> [e-mail]
>>>> https://www.ietf.org/mailman/listinfo/manet-dlep-rg
>>> The information contained within this e-mail and any files attached to this e-mail is private and in addition may include commercially sensitive information. The contents of this e-mail are for the intended recipient only and therefore if you wish to disclose the information contained within this e-mail or attached files, please contact the sender prior to any such disclosure. If you are not the intended recipient, any disclosure, copying or distribution is prohibited. Please also contact the sender and inform them of the error and delete the e-mail, including any attached files from your system. Cassidian Limited, Registered Office : Quadrant House, Celtic Springs, Coedkernew, Newport, NP10 8FZ Company No: 04191036 http://www.cassidian.com
>> 
>> _______________________________________________
>> manet-dlep-rg mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/manet-dlep-rg
> 
> 
> 
> -- 
> We began as wanderers, and we are wanderers still. We have lingered
> long enough on the shores of the cosmic ocean. We are ready at last to
> set sail for the stars - Carl Sagan
> _______________________________________________
> manet-dlep-rg mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/manet-dlep-rg