recursion

People often suggest that electronic voting machines should give people receipts. Wired magazine recently ran a "Found" photo about the idea, which Boing Boing just linked to, saying "Wow".

Voting receipts are a really, really bad idea, and here's why:

When you vote, your vote is secret. The secret vote is essential to protect your freedom from coercion and protect the system from vote buying.

Someone can threaten you to vote a certain way, or else. But as long as your vote is secret, you can tell them "Yes, sir" and then vote however you choose. If they can't look at your ballot, they can't find out how you voted.

Electronic voting machines change this picture because they're impossible to validate without a voter verified paper trail. So people suggest, "We should have it print out a receipt. You can take it with you and know how you voted." Wired's illustration takes this to the extreme, with online vote verification, win/loss record, and tracking numbers.

But once you've got receipts, you've opened the system to coercion and vote buying, because it's possible to check up on people. Bad, bad idea. Your vote must be kept secret.

How should electronic voting machines work, then? There must be a paper trail, and the voter must see the paper version of their vote and sign off on it. But then the vote should be placed in a secure, secret ballot box in case of a recount. There should be no identifying information on the paper version of the ballot. Essentially, what we have now is a printed optical scan ballot.

Some also suggest doing spot recounts of paper ballots to ensure that the machines are counting accurately.

Here's some notes about Lucene Desktop, Kevin Burton's latest crazy concoction.

PDFbox library for PDF extractions (anything for doing thumbnails?) BSD license.

TextMining.org for Word documents. Apache License.

OpenOffice.org also has an API.

Java 1.4 for creating image thumbnails? GIF, PNG, JPEG, BMP supported...

POPsearch, a similar idea, already implemented in C. Lots of features.

Index should have basic file metadata structures for all files, plus additional fields for each file type (images would have width and height, for example).

Field names should be lower case and human type-able for advanced queries.

Easy install should be a goal.

Windows has Google Desktop -- who would switch from that? Mac OS X Tiger will have Searchlight -- who would switch from that? But there is a niche on pre-Tiger Macs and Linux.

How do you index email?

Jetty would make a good embedded webserver and servlet engine as Tomcat blows and is way too big.

Index home directory only? What about privacy? Probably needs to be configurable.

All config should be possible through the web interface.

File formats: HTML, text, Word, Open Office, PDF, MP3, images (GIF, BMP, JPEG, PNG -- create thumbnails in the cache directory)...

Meta-file formats (must index inside the file): mbox, Microsoft mailbox formats, maildir (sorta).

Need some spiffy 16x16 icons to represent file types for above.