He made a strong case on a single argument against bitcoin several times - which is the need to wait for the better part of half an hour to confirm a transaction.
I still don't understand how bitcoin will ever overcome that obstacle for casual transactions (as opposed to ones in which a 30 minute delay isn't a problem)
That is a mediocre argument against one narrow use case.
A valid transaction which is widely seen in the network will be part of the blockchain with a high degree of certainty, even if it is just a few seconds old. That is not perfect, but the risk of accepting unverified transactions with low volume is very low.
It is important to remember that most systems that we use today have serious flaws, including payment systems like credit cards and cash. Bitcoin doesn't have to be perfect in order to succeed.
Every single person I've discussed Bitcoin with has identified the requirement to wait 3 confirmations, or 30 minutes as the fatal flaw for casual commerce. In fact, we spend quite a bit of time spitballing coming up with systems to overcome it (Giving your wallet some sort of "reputation" is common - but then requires you always spend from the same wallet, and also leaves the system open to a wholesale attack in which a common wallet app might double spend from a lot of people at the same time)
If someone could come up with a response to that single flaw, I think most of the others (increasing transactions per second, size of the block chain, etc...) can be overcome.
Re: "How likely is it that a valid transaction that is seen by a large part of the network won't be included in the blockchain?"
My understanding is that if you don't wait for multiple confirmations, then it's not particularly challenging to double spend bitcoins. That's why vendors don't accept a transaction as "good" until multiple transactions have occurred.
And waiting for 30 minutes to pay your restaurant bill would kind of suck.
> My understanding is that if you don't wait for multiple confirmations, then it's not particularly challenging to double spend bitcoins.
That's incorrect, there is a significant cost associated with executing a double-spend attack. For small transactions, it's uneconomical to mine an alternate chain in order to execute a double-spend attack.
If you trust the miner a little bit, you can assume that a zero-confirmation transaction be part of the common blockchain. Now it's not advisable to do that when selling somebody a Ferrari and letting them drive off, but it's probably less risky than accepting a credit card payment.
Regarding reputation: I don't quite understand the point. You'd just have to sign your tx with a private key that has a corresponding public key with some reputation attached to it. This is pretty independent of bitcoin itself. You could even build a system where you deposit some amount with a third party (or multiple third parties using a MULTISIG tx), which will reimburse the vendor with that amount if a double spend is executed.
Regarding reputation - the theory would be that you would be willing to accept transactions (that were on the blockchain, of course), from a particular wallet, after that person had some type of reputation (either because they had provided identification, or had a track record, or some other method of proving that transactions from this wallet don't always have to be confirmed three times, particularly for small amounts of BTC.
But I think your system would work as well - but it then needs more than just a bitcoin transaction from my wallet.
I'm just trying to think of how we can encourage Restaurants and Grocery stores to start accepting BTC, without getting scammed by double-spends.
I still don't understand how bitcoin will ever overcome that obstacle for casual transactions (as opposed to ones in which a 30 minute delay isn't a problem)