Quote

Sorry, you're describing desired behavior here, not an actual mechanism

Sorry, I'm very aware of this. The point of this post is to simply describe an end-goal: transactions that are required to be signed on multiple devices before they're relayed, also allowing users to see attempts to spend in the process.

Why is this worth talking about? Because multi-sig is still too obscure for the average user to create, theft is rampant within the bitcoin community, and it's difficult to tell when part of your security is compromised before it's too late. Hopefully the idea sparks enough interest and discussion so that it can lead to better security on the basic client. If you're going to press me for exact mechanisms, then I'm afraid we won't be speaking the same language. There may be more than one way to accomplish this type of behavior and I was hoping to introduce the general concept first so that multiple different strategies could be theorized to adopt it. If you think multi-sig is the best way to go, great. I still have reservations that it may not bring all the desired functionality to the table.

Quote

What prevents one from completing the transaction entirely without the aid of the other?

Upon setting up the blind wallet for the very first time, the user is prompted to enter two passwords. Each will encrypt the private keys in the wallet with rounds of AES-256-CBC, for example. The blind wallet will never again ask for the second password.

This means that upon entering password 1 with your blind wallet, you form a partially decrypted transaction that still requires password 2 to fully decrypt. Why are we doing this? So that the user never has to enter their full decryption information on a single device again. They may even set up x-of-y passwords. Assuming a thief obtains all the information they possibly can from one device, they'll still end up with a partially encrypted wallet.

Each wallet by itself is still incapable of forming completed transactions.
The blind wallet isn't asking for the 2nd password
The spend wallet doesn't even have the private keys to initiate the process, it only acts as an input for confirmation.

Quote

[Multi-signature transactions] can't go into the blockchain half completed

Okay...let's assume that a thief discovers 1 of 2 private keys needed for a multi-signature transaction. Can he or she initiate a spend attempt using only one of the keys? If it's not going into the blockchain, how does the owner of both keys see this and realize that one is compromised?

Quote

Why would the second wallet sign something the owner didn't authorize? Why would it sign a transaction which is not valid? How can an anonymous public network decrease exposure to MITM than directly connecting the devices?

If you're ultimately asking about why a secondary network is necessary at all, it seemed very obvious to me that the bitcoin network itself would be incapable of transmitting partially decrypted transactions, and it would be a fruitless argument to try to convince anyone that it should do so. If the blind wallet has to communicate some portion of data to the spend wallet, then choose whatever method you think is best that wouldn't rely on a central authority. A decentralized network seemed more ideal because it would allow an owner to monitor the spend attempts from their wallet, anywhere in the world, without that information having to touch the bitcoin blockchain.

What this is ultimately trying to accomplish:
1) Creates an environment in which the user never has to enter their full password on a single device
2) Splits up the necessary password so that by design, it can be entered on multiple different devices
3) Allows the owner to see any attempt to move or spend the bitcoin at a single address before it completes, assuming all passwords have not been compromised

And it's trying to do all this without requiring any changes to the bitcoin protocol itself.