Re: [Coin] Minutes of the interim meeting and some words from thecochairs

[Toerless Eckert <[e-mail]>]

Thanks, Sharon


Terminology and reference architecture would be good goals to work towards, but 
how would we validate what is proposed in such work ? I think for that it would
be very helpful to start from some significant enough, but not "ocean boiling"
set of well described use-case scenarios

It sounds to me as if the aspects you mention could apply all to edge-mobility
such as car-2-curb. Would that be a good set of use-case/scenarios ?

Setting context with some tutorial writeup of such use-cases and a survey of
existing/proposed mechanisms for the aspects you mention might be a good starting
point to establish a common base to work from towards those goals you mention.
And a survey/tutorial from a broader set of authors might also be a much easier
way to get a community working on the problem started.

Of course i am specifically interested in the subset of the work areas you mention
that do impact not only the compute overlay, but also the routing underlay. And
that impact goes IMHO both ways.

For example, if application developers including those working on such
"compute overlays" would better understand what could be done with the new
evolving stateless multicast technologies such as BIER, then it might help
to better solve such application challenges. But alas, we are also somewhat
slow so far to produce good material to educate about such new opportunities
in the routing underlay (its on my bucket list though). 

On the other hand, working in the routig underlay, i am of course even more
interested in what challenges such overlays face that have not yet good
solutions from the routing underlay. But which maybe could have better
solutions if we worked on more/better semantic routing solutions ;-)

Cheers
    Toerless

On Thu, Mar 10, 2022 at 09:17:52PM +0200, Sharon Barkai wrote:
> 
> Thanks Toerless, good observations.
> 
> The well mapping is not just of “location“ subscription-replication to network topology, solve big Joins, theres also:
> 
> o network addressing supporting privacy while publishing-subscribing to states
> o dns-less seamless context switching while physically moving between states of interest 
> o high availability and dynamic allocation of addressable states per car activity/facilities
> o addressable stat-mux coalescing low latency low bw access queues to metroE queues
> o etc. 
> 
> But unless somebody writes and COIN adopts some agreed terminology and reference architecture for Compute Overlay in (routed) Network its really hard to discuss research or generalize these topics. Iv seen a few proposals fly by the list by Luigi and others. Maybe if one of these gets a green light the discussion can move forward. Perhaps its premature but the interest seems real.
> 
> --szb
> Cell: +972.53.2470068
> WhatsApp: [phone]
> 
> > On Mar 9, 2022, at 19:41, Toerless Eckert <[e-mail]> wrote:
> > 
> > When subscription patterns
> > do for example map well with the network topology (vehicles subscribe to interest
> > channels specific to their vicinity and network topology is vicinity based), then
> > a lot of optimizations can be done.

-- 
---
[e-mail]