Re: Please find evidence of >80bit election

From: David Mertz <voting-project_at_gnosis_dot_cx>
Date: Wed May 05 2004 - 12:15:36 CDT

> Before we decide that 2-D barcodes are needed, let's find SOME
> specific evidence for the need. My tool let's you determine precisely
> and quantitatively how much actual information any election contains.
> If it's not more than 80 bits, it's not an argument for 2-D. Show me
> the data!

I constructed a hypothetical election for Cook County, IL. Based on
Doug Jones' recollection, I guessed at 40 judicial retention issues.
And to fill it out, I left on the first few single-selection
multi-candidate contests that occurred on the demo. Here's what I get:

   $ ./election-entropy.py < cook-hyphothetical.data
   Election summary for OVC demo ballot (write-ins count as candidate)

   15756334434937779726096 distinct votes are possible
   Optimal encoding is approximately: 74 bits
   Contests at bit-boundaries, approx: 92 bits
   Self-delimited (16 char symbology): 184 bits

Well, this continues to make self-delimited codes look obese; but
that's a side issue.

One thing this points to in my mind is that we may not want to respect
bit-boundaries for a large collection of similar up/down issues. That
is, for a Judge, you can vote Retain, Exclude, or No Preference.
Up/down still has three options to store, or 1.58 bits of information.
If we use a whole 2 bits to store that, we are a bit inefficient, each
contest discards almost a half-bit of information (for those
unfamiliar, a half-bit of info might sound odd, I know, but it makes
sense).

What you can do instead is store the whole Judicial Retention section
as one big number in base-3 (one digit per judge), then convert that to
another big binary number before putting it on the ballot. Once you
have some CRCs there, the data integrity is fine.

Btw. Possibly slightly contra Karl, I don't see a need for a whole lot
of error correction in codes. A CRC can detect accidental corruption.
But if a barcode goes bad, for whatever reason, the simple fallback is
to require a manual examination of the printed content of the ballot.
We don't -need- to be able to reconstruct the barcode for every
possible error on it--just know when to treat it as non-readable.
==================================================================
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
==================================================================
Received on Mon May 31 23:17:15 2004

This archive was generated by hypermail 2.1.8 : Mon May 31 2004 - 23:18:15 CDT