Re: [lisp] Martin Duke's Discuss on draft-ietf-lisp-rfc6833bis-27: (with DISCUSS and COMMENT)

[Albert Cabellos <[e-mail]>]

Hi Martin

I´ve just posted clarifying this, find below the relevant paragraph updated
(new text in *bold*)

https://datatracker.ietf.org/doc/draft-ietf-lisp-rfc6833bis//

After sending a Map-Register, if a Map-Notify is not received after 1
second the transmitter MUST re-transmit the original Map-Register with an
exponential backoff (base of 2, that is, the next backoff timeout interval
is doubled), the maximum backoff is 1 minute.
*Map-Notify messages are only transmitted upon the reception of a
Map-Register with the M-bit set, Map-Notify messages are not retransmitted.
The only exception to this is for unsolicited Map-Notify messages, see
below.*
A Map-Server sends an unsolicited Map-Notify message (one that is not used
as an acknowledgment to a Map-Register message) only in conformance with
the Congestion Control And Relability Guideline sections of <xref
target="RFC8085"/>. A Map-Notify is retransmitted until a Map-Notify-Ack is
received by the Map-Server with the same nonce used in the Map-Notify
message. An implementation SHOULD retransmit up to 3 times at 3 second
retransmission intervals, after which time the retransmission interval is
exponentially backed-off (base of 2, that is, the next backoff timeout
interval is doubled) for another 3 retransmission attempts.
*Map-Notify-Ack messages are only transmitted upon the reception of an
unsolicited Map-Notify, Map-Notify-Ack messages are not retransmitted.*


Please let me know if this clears your DISCUSS

Thanks!

Albert


On Wed, Oct 28, 2020 at 9:35 PM Dino Farinacci <[e-mail]> wrote:

> Thanks for confirming Martin.
>
> Dino
>
> > On Oct 28, 2020, at 1:34 PM, Martin Duke <[e-mail]>
> wrote:
> >
> > Yes, the design you describe here makes perfect sense. I did not get
> this distinction from the current text at all. So yes, please reword it.
> The framework you presented in this email is much clearer and may serve as
> a good basis.
> >
> > Thanks
> > Martin
> >
> > On Wed, Oct 28, 2020 at 1:30 PM Dino Farinacci <[e-mail]>
> wrote:
> > > If I parse your answer correctly, the answer to my question is 'no'.
> So in the scenario where the Map-Notify is lost, both the Map-Register and
> the Map-Notify are on retransmission timers. The most straightforward
> reading of the text is that
> > > - I respond to every Map-Register with a Map-Notify (if it requests it)
> > > - For every Map-Notify I send, I start a retransmission timer.
> >
> > Let me be a bit more clear about how retransmissions of Map-Registers
> and Map-Notifies work.
> >
> > There are two broad cases,
> >
> > (1) Map-Notify messages as an ack to Map-Registers.
> > (2) Map-Notify messages that are unsolicited from Map-Servers.
> >
> > In the first case:
> >
> > (1) When Map-Registers are sent with the bit set to request
> acknowledgment for Map-Registers received by Map-Servers, a retransmission
> timer is set by the xTR for Map-Register retransmissions (which is more
> often than the periodic Map-Register timer).
> >
> > (2) The Map-Server sends a Map-Notify for each received Map-Register.
> There is NO retransmission timer for the Map-Notify.
> >
> > In the second case:
> >
> > (1) A Map-Server detects a RLOC-set change and wants to Map-Notify the
> xTRs in the old and new RLOC-set by sending a Map-Notify message. These
> messages are acknowledged by the xTR by Map-Notify-Ack messages. In this
> case the Map-Server DOES have a retransmission timer for the Map-Notify for
> each xTR.
> >
> > (2) The Map-Notify-Ack DOES NOT have a retransmission timer and simply
> is sent by an xTR when it receives a Map-Notify.
> >
> > So having said that, you probably still want some better rewording.
> Please confirm?
> >
> > Dino
> >
>
>