Re: [rtcweb] Update of future interim meeting locations

["James M. Polk" <[e-mail]>]

At 04:28 PM 4/17/2012, Eric Rescorla wrote:
>On Tue, Apr 17, 2012 at 5:15 PM, James M. Polk <[e-mail]> wrote:
> > At 04:09 PM 4/17/2012, Eric Rescorla wrote:
> >>
> >> On Tue, Apr 17, 2012 at 5:06 PM, Marshall Eubanks
> >> <[e-mail]> wrote:
> >> > On Tue, Apr 17, 2012 at 4:57 PM, Eric Rescorla <[e-mail]> wrote:
> >> >> On Mon, Apr 16, 2012 at 3:45 PM, Cullen Jennings <[e-mail]>
> >> >> wrote:
> >> >>> First of all, the chairs would are going to declare that there is WG
> >> >>> consensus for a 1:1:1 meeting rotation between west coast of north
> >> >>> america, europe, and east coast of north america and plan to run
> >> >>> approximately equal number of meetings in these locations.  In
> >> >>> counting the number of past meetings in a given region, we will
> >> >>> include all the face to face RTCWeb and WebRTC meetings as this work
> >> >>> is closely joined and a large number of the participants travel to
> >> >>> both sets of meetings. The face to face meetings that happen at the
> >> >>> main IETF or W3C meetings are included in this count. Collocated
> >> >>> meetings will be counted just once as they only require one set of
> >> >>> travel to that location.
> >> >>
> >> >> Cullen,
> >> >>
> >> >> Thanks for this clarification. Just to be totally sure I understand,
> >> >> do you mean the following algorithm?
> >> >>
> >> >>   For an interim meeting at time X, sum up all the prior meetings
> >> >>   for each of the regions. The region with the fewest prior
> >> >>   meetings is then selected for the next interim. [0]
> >> >>
> >> >
> >> > Which implies that  Paris doesn't count.
> >>
> >> That's certainly not what I meant. Perhaps it would help to rewrite
> >> "next interim" as "interim X". Concretely, to select an interim meeting
> >> in mid-September, I would expect to count meetings up to and
> >> including Vancouver.
> >
> >
> > why don't Paris and Boston (I think that was the last interim, though that
> > might have been CLUE) count here?
>
>They would. That's why I said "Up to and including". Paris was before
>September 2012.
>
>Obviously I'm being unclear.

No, I think I knew what you wanted to get to the first time, but what 
you have below articulates it quite nicely though.

And, as you pointed out, we need a means for solving for ties - 
mostly because the goal of this is to have a tied once a year between 
3 (if we keep the 1:1:1 format), then once a year between 2, right?

James

>Here's a slightly more formal description.
>
>Each scheduled or past meeting is represented as a tuple [region, date].
>E.g., Paris would be something like [Europe, 20120325]
>
>To select an interim at date X, take the subset of all meetings with date < X
>and count the number in each region. To just make up some data, this
>would look like:
>
>{
>   'Europe':1,
>   'West Coast':2:,
>   'East Coast':0
>}
>
>Then, select the region with the smallest number of meetings for the
>interim at date X.
>So, in this case, the interim would be held on the East Coast.
>
>For additional concretenss, iff we were executing this algorithm to 
>select this
>interim, we would be building a table of all meetings up through Paris. I.e.,
>the last two would be IETF 83 and the Jan/Feb Interim in Mountain View.
>[I'm assuming for the purposes of this discussion that we are not counting
>CLUE interims, since the chairs referred to "RTCWeb and WebRTC".
>
>One detail: you need to select the interims in chronological order because the
>previous interim selections obviously affect the next ones.
>
>Best,
>-Ekr