Bitcoin snack machine (fast transaction problem)

[cadmium]

accept 0 confirm, but have a camera taking buyer face, and having show ID, and have an algor to check if both match

[oakpacific]

to do a SUCCESSFUL double spend requires alot of speed and hardware. but to send out 2 transactions to 2 different addresses using 2 clients but only using one private key supply does not require expensive equipment. A successful double spend requires a confirm on both sides. meaning the 10 minute wait. but sending out 2 transactions that don't require waiting the 10 minutes is piss easy.

accepting unconfirmed transactions should only be allowed on products with delayed delivery. such as bitpay which can inform merchants that the payment has failed in 10 minutes and the merchant still has time to email the customer to attempt the payment again before losing their next-day delivery slot.

it is never EVER acceptable for instant access content.

examples.
if i sent
0.1BTC from 1HomeMachineAddress to 1VendMachineAddress
0.1BTC from 1HomeMachineAddress to 1VendMachineAddress

one of those transactions would instantly be ignored as it would appear as just an echo of the other, where many nodes pass around the same transaction (to avoid repeats)

but if i sent
0.1BTC from 1HomeMachineAddress to 1HomeMachineAddress
0.1BTC from 1HomeMachineAddress to 1VendMachineAddress

it would appear as 2 different transactions and would require a confirm to sort out the mess (10 minute wait)

Right but you are assuming that the receiver is not checking the balance on the address and that the user has spent in excess of the balance.  That is not hard to do...


If my address has 1 btc and I have 2 tx for 1 btc each... the receiver can tell me to wait 10 minutes....  honest actors are not going to do double spends...

I still think an outright rejection for addresses with unconfirmed transactions is the most pragmatic way to go, buyers should be warned with notice that refund cannot be processed immediately.

[prophetx]

At some point there is some level of trust that comes into play... and I know having merchants in my own family that they are not going to care beyond checking that an address has enough to spend with my unconfirmed tx when it comes to small ticket items like a magazine or pack of cigs.  Plus most of their customers are repeat...

Even with physical money the merchant trusts that the money is not counterfeit, and with credit cards that the card was not stolen and unreported.  So the only way to really pull this off is combined with a 51% attack and anyone who owns equipment to do that would be jeopardizing their own investment...

For larger transactions they take so much time that by the time the tx is done the confirms have come in.  Have you ever bought a house or car or really nice jewelry in 5 minutes?  No that stuff all has paperwork and packaging that happens and can take some time.

[prophetx]

to do a SUCCESSFUL double spend requires alot of speed and hardware. but to send out 2 transactions to 2 different addresses using 2 clients but only using one private key supply does not require expensive equipment. A successful double spend requires a confirm on both sides. meaning the 10 minute wait. but sending out 2 transactions that don't require waiting the 10 minutes is piss easy.

accepting unconfirmed transactions should only be allowed on products with delayed delivery. such as bitpay which can inform merchants that the payment has failed in 10 minutes and the merchant still has time to email the customer to attempt the payment again before losing their next-day delivery slot.

it is never EVER acceptable for instant access content.

examples.
if i sent
0.1BTC from 1HomeMachineAddress to 1VendMachineAddress
0.1BTC from 1HomeMachineAddress to 1VendMachineAddress

one of those transactions would instantly be ignored as it would appear as just an echo of the other, where many nodes pass around the same transaction (to avoid repeats)

but if i sent
0.1BTC from 1HomeMachineAddress to 1HomeMachineAddress
0.1BTC from 1HomeMachineAddress to 1VendMachineAddress

it would appear as 2 different transactions and would require a confirm to sort out the mess (10 minute wait)

Right but you are assuming that the receiver is not checking the balance on the address and that the user has spent in excess of the balance.  That is not hard to do...


If my address has 1 btc and I have 2 tx for 1 btc each... the receiver can tell me to wait 10 minutes....  honest actors are not going to do double spends...

I still think an outright rejection for addresses with unconfirmed transactions is the most pragmatic way to go, buyers should be warned with notice that refund cannot be processed immediately.

Yea I agree that way the machine is not in a locked state waiting for the confirms

[MoonShadow]

to do a SUCCESSFUL double spend requires alot of speed and hardware. but to send out 2 transactions to 2 different addresses using 2 clients but only using one private key supply does not require expensive equipment. A successful double spend requires a confirm on both sides. meaning the 10 minute wait. but sending out 2 transactions that don't require waiting the 10 minutes is piss easy.

Not easy.  Let me se you try it.

accepting unconfirmed transactions should only be allowed on products with delayed delivery. such as bitpay which can inform merchants that the payment has failed in 10 minutes and the merchant still has time to email the customer to attempt the payment again before losing their next-day delivery slot.

it is never EVER acceptable for instant access content.

It's fine for a great many things.  A movie streamed over the Internet, for example, because it can be cut off if a double spend is detected.  Game or digital content licences (on Steam or iTunes for example) can be reversed.  Online services (monthly access, or even daily access) can be reversed.  All of this with a simple transaction, not using blockchain enforcable contracts.

https://en.bitcoin.it/wiki/Contracts

Using contracts, particularly example #7, opens up some rather creative solutions beyond the simple transaction.  The "loyalty card" system can be implimented within an android phone, without a actual card, and the balance revoked at will, so one could use it as a running balance for a vending machine at work as easily as one could use it at av ending machine that they never expect to visit again.  One could use #7 to pay for gasoline at the pump, again without a 10 minute wait.  Even the most basic contract is significantly more difficult to "double spend", and may actually be mathmatically impossible under certain conditions.

[Kluge]

accept 0 confirm, but have a camera taking buyer face, and having show ID, and have an algor to check if both match

Who the Hell submits to being pictured and giving away a photo of their ID for a freakin' candy bar? I don't like people looking at my ID, much less getting a scanned copy. I'm still uneasy about having submitted ID to sites like Gox. Anyway - the cost of adding all that would be pretty ridiculous for a candy vending machine.

Anyway, I compared this to personal checks in a (newer, actually) thread. Many stores accept personal checks (usually up to $xxx), and many also do their own version of KYC (see MoonShadow above). If a customer turns out to be a frequently-thieving assbag, you ban them from your store. In the case of vending machines, I'd guess they'll just have to accept the risk or use ZipConf-like insurance. For workplace vending machines, you can do something like lock out the machine until a tech visits it when a double-spend is detected, so the thief will gain the ire of their coworkers, and pressure to stop doing it. If it continues and company management does nothing, the vendors can simply remove the machines and put them to work somewhere else. I think most physical and low-cost/high-volume companies will find the risk exceptionally low.

Companies don't need to, and frequently do not, operate on the premise that their customers are thieving assholes, with the exception of many IP distributors, and if a company thinks Sony BMG is a good model to emulate, keeping them out with uneasiness over risk of double spends is fine by me. 

[luv2drnkbr]

Two things.  First, I seriously doubt that double spends on a $2 candy bar are going to be common enough that they'll be a problem.  At worst, the seller bumps the price up 10 cents to cover the cost of fraud and nobody really cares.

Second thing is a simply solution is a phone wallet that automatically creates a transaction to split up its funds into addresses each with only a small amount in them, and then what you would do is send the machine a QR of your private key for one of the addresses and a public key for returning change, and it creates the transaction and sends you the change.  This way it knows the transaction is valid, and it can wait 5 or 10 seconds and see if a double spend transaction shows up on the network.  At least this way, it will be somewhat more confident than it would be by you send it the transaction, but obviously, the only true confidence is a couple of confirmations.

[stslimited]

Ripple, obviously. It is credit while the underlying currency settles.

[LeTanque]

Ripple, obviously. It is credit while the underlying currency settles.

Yes, true. Ripple solves this problem.

For the sake of the Bitcoin only argument, though, I think the idea of verifying unconfirmed transactions would be useful.

[stslimited]

Ripple, obviously. It is credit while the underlying currency settles.

Yes, true. Ripple solves this problem.

For the sake of the Bitcoin only argument, though, I think the idea of verifying unconfirmed transactions would be useful.

My company is considering blocksplit insurance. basically allowing everyone to accept 0 confirmations and be insured if a bad actor began double spending or going off on their own chain

just not sure what people would pay for that

[MoonShadow]

In truth, there are several ways to solve this problem.  I expect that most will be tried in some form or another, and the best solutions will come to dominate, and most clients will evolve to be able to use those methods.

[MoonShadow]

Ripple, obviously. It is credit while the underlying currency settles.

Yes, true. Ripple solves this problem.

For the sake of the Bitcoin only argument, though, I think the idea of verifying unconfirmed transactions would be useful.

My company is considering blocksplit insurance. basically allowing everyone to accept 0 confirmations and be insured if a bad actor began double spending or going off on their own chain

just not sure what people would pay for that

I'd say that one half to one full percent of the purchase price would be about right.  That's about what the credit card companies charge to deal with fraud, and I wouldn't expect that zero-conf frauds would amount to greater rates of fraud.  Start at one percent, and see how profitable you are.  If it's profitable enough, some competitor will come along and force you to lower your rates.  If it's not profitable, you'll stop doing it too.

[biggbox]

Sorry for bringing up this very old discussion. Just wanted to see if there are any new solutions to this problem. I don't mind buying a can of coke from the vending machine with my mobile wallet.

[Lukas_Jackson]

Sorry for bringing up this very old discussion. Just wanted to see if there are any new solutions to this problem. I don't mind buying a can of coke from the vending machine with my mobile wallet.

Yes, there is a solution without 3rd party payment processor. Finally.
http://www.dashndrink.com

[madonnino]

Sorry for bringing up this very old discussion. Just wanted to see if there are any new solutions to this problem. I don't mind buying a can of coke from the vending machine with my mobile wallet.

Yes, there is a solution without 3rd party payment processor. Finally.
http://www.dashndrink.com

This is nice but the machine works whit dash and no whit bitcoin, does InstantX works whit bitcoin? at most we will use to change shapeshift btc in dash before purchase

[Lukas_Jackson]

Sorry for bringing up this very old discussion. Just wanted to see if there are any new solutions to this problem. I don't mind buying a can of coke from the vending machine with my mobile wallet.

Yes, there is a solution without 3rd party payment processor. Finally.
http://www.dashndrink.com

This is nice but the machine works whit dash and no whit bitcoin, does InstantX works whit bitcoin? at most we will use to change shapeshift btc in dash before purchase

Well, satoshi said bitcoin technology is an experiment. Dash uses this tech and explore ideas which would work in decentrilised way.

[Instaclasher56]

Do not say this is impossible but This could be cool!

There would be this website with its own bitcoin card that you could load with your bitcoins! This card will be loaded once the confirmations were confirmed and then you can spend your bitcoins anywhere without your phone! They wouldn't need confirmations but it could take up to 2 minutes so the machine sees what your balance is and then deducts what you bought (plus the 2 transaction fees (#2 being the card company he's some extra money from that purchase)) from your balance and then you walk away with your item!

Yes more stuff would be needed but it's a great idea!

[Windpower]

For low value items like that you would normally accept the transaction without any confirmations.  If the bitcoin node that the vending machine is talking to accepted the transaction then it's probably valid.  I think the only danger in that is if it was somehow double spent in a short time window - then it may never get into any blocks.  I'm not sure if that's really possible, maybe it is, but it seems like it would be hard to pull it off reliably.

Yeah exactly. They would probably just put a QR code on it and you scan it and send the Bitcoin.

[Kakmakr]

Just use a off-chain solution like Xapo with instant transaction confirmation. You get people to sign up for a Xapo account and then have them use their Xapo debit card to pay for the stuff they want to purchase from these vending machines.

Once this takes off, Xapo could form a partnership with these vending manufacturers and arrange for discounts on fees for these micro transactions. At this stage all in card spending <off-chain> is free. Correct me if I am wrong.

[ranochigo]

Just use a off-chain solution like Xapo with instant transaction confirmation. You get people to sign up for a Xapo account and then have them use their Xapo debit card to pay for the stuff they want to purchase from these vending machines.

Once this takes off, Xapo could form a partnership with these vending manufacturers and arrange for discounts on fees for these micro transactions. At this stage all in card spending <off-chain> is free. Correct me if I am wrong.

There's a huge problem with this solution. We're using Bitcoin, not services that uses Bitcoin like a debit card. By giving your coins to Xapo, you are allowing them to monitor your spending habits and even control your coins fully. There are major privacy and security concerns regarding this. Xapo debit cards uses Visa and is thus not fully a Bitcoin solution. Offchain transfers are usually free for everyone. However, a better solution would be to implement counter measures against double spending and still accept 0 conf transactions. The risk is relatively low.