Quote from: CIYAM on May 18, 2015, 01:27:52 PM

Quote from: CryptKeeper on May 18, 2015, 01:23:04 PM

OK, I understand that if Alice's AT - which goes first - has a small timeout of 24 hrs and Bob's AT has a much greater timeout, say 72 hrs, not much can go wrong.

Actually you haven't quite got it - Alice's timeout needs to be the longer one (as it is Alice that will be sending the "key" to Bob's AT first).

I always fail when the chances are 50:50 ... Grin

Quote from: CIYAM on May 18, 2015, 01:27:52 PM

Quote from: CryptKeeper on May 18, 2015, 01:23:04 PM

Could the timeouts somehow be part of the secret, so that none of the participants can be tamper with it?

I'm not sure how that could be achieved - the best approach IMO is that the hash of the ACCT AT code could be published and compared to (to prove that it is the code that you think it is without having to actually understand AT machine code) and that you check the timeout value before proceeding.

Assuming there is enough of a time gap then there is not much that can go wrong (unless Bob's internet dies after Alice sent the secret and he can't get it working until the refund expiry time for Alice happens).

Couldn't you hard code reasonable timeout values into the AT code and publish the code's hash? So if everyone agrees on using the same AT code for ACCT, the timeout values would be common knowledge.