[AFS3-std] Second Draft of Standardisation Document

[Jeffrey Hutzelman]

--On Wednesday, August 27, 2008 12:46:58 PM -0700 "Buhrmaster, Gary" 
<gtb@slac.stanford.edu> wrote:

>
>> Personally, I'd be comfortable with just flipping a coin.
>
> It is an accepted practice in many US election jurisdictions.
> It is likely to occur rare enough that anything more is
> going to be overkill.

A coin flip or similar random event is not a bad approach.  As you point 
out, this approach is used for breaking ties in public elections in many 
jurisdictions.  In fact, according to a Wikipedia article I recently read, 
use of a random event is prescribed in a number of jurisdictions that 
formerly solved the problem by giving a casting vote to the vote-taker.

> The MD5 hash of the entire email (including headers) of
> the last person to vote for the specific individuals?
> So if you want to plan to rig the tie breaker for a
> planned tied election, you have to do a lot of work
> (and other than myself, who would even think that devious?)

Which headers?  All the headers?  All headers except trace headers?
This matters, because when the process is running normally, there are 
multiple vote-takers each of whom receive a copy of the message, and they 
must be able to independently arrive at the same result.

I was actually thinking of resolving ties by dropping the last vote 
received, which would potentially have the effect of encouraging people to 
vote early (which is good because people who put it off may never get 
around to voting at all).  Unfortunately, that has a related problem which 
your proposal also shares; namely, last according to whom?

-- Jeff