Quote from: cunicula on January 26, 2012, 03:50:58 AM

Quote from: Maged on January 26, 2012, 03:38:58 AM

Yes, this actually would be possible. To do what you asked, specifically, you would create 2-of-3 multisig transaction addresses where the third key is owned by a third party (likely web-based) service. When you lose access to one of your two private keys, you could use this service to recover your funds at whatever predetermined rate you chose. Such a service could require a login, or not, depending on the security you desire.

Why add in a web-based service since it is not necessary in theory? Does it have something to do with the code lacking awareness of the blockchain? Are the core code modifications necessary to remedy this too difficult?

In my view, there should be a plan where txns with time-dependent spending limits without help from a third-party service become possible at some point. Maybe this is too tough to achieve right now, but it seems like a cleaner, more convenient, and intuitive solution to me. Time-dependent spending limits have such prevalence in consumer banking that I think people will be uncomfortable without them.

Third party service providers can cause lots of problems. Many people are already quite uncomfortable with them. "Don't lose two things at once" strategies also make things quite complicated. People get worried by stuff that is complicated. Who wants to spend the mental energy? I'm not saying I don't see value-added in these proposals, it is just that they do not seem like first-best solutions for users.

Should we view these proposals as temporary fixes?

You're right that it's possible in theory. However, we're talking about quite a few major changes that would have to be done.
1) Transactions are not aware of the blockchain at all. For this, they'd at least need to be made aware of the timestamp of the block they were put in. All such blockchain-aware transactions would need to be locked for at least 100 blocks after they are included in the chain before being spendable again. Even that might not be enough, in theory, although we'd likely be able to get away with it in practice.
2) Transactions would need to be aware of the output of the transaction in which they are redeemed. This is needed to find out how much is being spent and put conditional checks on it.

Neither of these are easy tasks. They might not ever be safe, either.