Re: [RE] Minutes of the subscription adhoc
For unknown reason I was not connected to the teleconference group. My guess
is that the same thing happened to Dirceu and Geoff. I wonder what happened.
Best regards,
Alexei Beliaev
Gibson Labs
408-313-2665
----- Original Message -----
From: "David V James" <dvj@ALUM.MIT.EDU>
To: <STDS-802-3-RE@listserv.ieee.org>
Sent: Thursday, June 02, 2005 8:05 PM
Subject: [RE] Minutes of the subscription adhoc
> Minutes by Felix Feng/David V James
> Telerconference of
> Subscription Adhoc of
> IEEE 802.3 Residential Ethernet Study Group
> 2005June01 14:00-16:00 PST
>
> Meeting ID: 802373
> Toll Free (US Only): + 1 877-827-6232
> International: + 1 949-926-5900
>
> Attending:
> George Claseman Micrel
> Felix Feng* Samsung
> Geoffrey Garner* Samsung
> David James JGG
> Tom Mathey Northern Data Systems
> Eric Ryu* Samsung
> * via teleconference
>
> Note:
> If you were attending and have not been listed,
> or observe error in the preceding listing, please
> send correction instructions to feng.fei@samsung.com
>
>
> 1) Background
> The Tuesday 2005May31 teleconference identified the need to
> further discuss the topic of subscription, with the
> desire for convergence. With this in mind, the Wednesday
> 2005Jun01 meeting time was allocated for this discussion.
>
> 2) Rough agenda:
> a) Felix's slides on SRP alternative design
> (20050526_Felix_GSRP_SRP.ppt)
> b) DVJ's white paper review (if time allows)
> c) General discussion
>
> 3) Slides on SRP alternative design:
> Felix presented slides on an alternative design of SRP, which
> is an elaboration of "the second example" in his presentation
> at the 2005May Austin Residential Ethernet Study Group
> interim meeting.
>
> We discussed the deploying environment of SRP. In some scenarios
> such as 802.17 LAN, the point-to-point assumption might not be true.
> In those cases a full GARP participant state machine could be required.
>
> A key distinction from the description of GARP is that the
> registration protocol is asymmetric, as appropriate for
> the 1-to-N talker-to-listener scenario. This might be doable,
> although not currently described in GARP.
>
> We discussed the possibility of merging part of GARP processes
> with RESV signaling process. For example, we may use the
> periodically RESV signaling as the LeaveAll messages of GARP.
> We conclude this might be possible; it will be further studied.
>
> General comments on this proposal are:
>
> -It is similar to the proposal that was described in DVJ's white paper.
> The listeners announce themselves, and the talker responds to them.
> This new approach integrated the response of talker with the
> reservation
> function. A resource locking difference is:
> a) The DVJ white paper locks resources from listener-to-talker.
> b) The Felix presentation locks resources from talker-to-listener.
>
> -All preferred using distinct talker-to-listener control frames
> (as opposed to data frame transmissions) as the talker-side
> handshake indication:
> a) This works even when data flows are interrupted
> - possibly more compatible with Congestion Management,
> or other protocols that reserve bandwidth for
> variable-rate (possibly bursting) transfers.
> - bursting and bunching no longer affect timeouts
> b) Status/commands can be returned in the talker-to-listener
> direction.
> - this allows error status to flow towards the listener(s)
> - BW can be committed in the talker-to-listener direction.
> (This seems to be the preferred direction.)
>
> -Many(talker)-to-many(listener) stream scenarios are possible,
> although not necessarily needed.
>
> Attendees provided various editorial comments, which were
> acknowledged by the presenter and will be incorporated into
> next edition of this presentation.
>
> Tom Mathey was interested in whether a listener can listen
> to multiple streams, possibly initiate by multiple talkers.
> We clarified this was possible and necessary for RE.
>
>
> 4) DVJ's White paper review:
> We noted the distinct protocols appear to be converging.
> The lack of talker-to-listener messages was always a
> concern in the DVJ proposal, and will now be assumed.
> (No time left for further discussion.)
>
> 5) Next Subscription Adhoc meeting:
> Wednesday 2005June08 14:00-16:00PST
> a) Review of DVJ-edited working paper
> (as revised, based on today's discussions)
> b) Discussion of parameters within the messages
>
> 6) Reminder of next RESG conference call:
> Tuesday 2005June07 15:00-16:00 PST