Economically Unspendable Outputs: A Problem On The Radar

[Explodicle]

Who knows, maybe provably fair gambling is the "killer app" for Bitcoin.

And I thought Bitcoin was a fiat-killer, the most powerful idea for the internet since the WWW.
I do hope some other people share this view...

Those things aren't mutually exclusive. Part of the advantage over fiat is that it's the market - not a bunch of lofty rhetoric - that decides the allocation of resources.

[1714098207]

1714098207

[1714098207]

1714098207

[1714098207]

1714098207

[misterbigg]

That "bloat" when maxed out will increase chain size linearly, while the cost of storage goes down exponentially. It's really an anomaly of young bitcoin that we had blocks below the cap this long - for most of bitcoin's future, the blocks should be nearly full.

Seems to me that as long as the community does not choose to fiddle with the block size, the fees will start to take care of SatoshiDice-class problems and it should remain practical for the system to be operational on the backs of a wide class of operators. 

You didn't read. Sure, the blocks are full but the problem is that the SatoshiDICE losing confirmations are unspendable and unprunable. Do you understand this distinction? A normal transaction output stays in the block chain until it is completely spent at which point it is eligible for pruning and can disappear forever. It only "bloats" the block chain until it is spent. Compare this with SD's losing confirmation, which can never be spent economically and therefore can never be pruned. Do you get it now? Why won't it read?

[Explodicle]

That "bloat" when maxed out will increase chain size linearly, while the cost of storage goes down exponentially. It's really an anomaly of young bitcoin that we had blocks below the cap this long - for most of bitcoin's future, the blocks should be nearly full.

Seems to me that as long as the community does not choose to fiddle with the block size, the fees will start to take care of SatoshiDice-class problems and it should remain practical for the system to be operational on the backs of a wide class of operators.  

You didn't read. Sure, the blocks are full but the problem is that the SatoshiDICE losing confirmations are unspendable and unprunable. Do you understand this distinction? A normal transaction output stays in the block chain until it is completely spent at which point it is eligible for pruning and can disappear forever. It only "bloats" the block chain until it is spent. Compare this with SD's losing confirmation, which can never be spent economically and therefore can never be pruned. Do you get it now? Why won't it read?

Even with no pruning at all the max chain size still goes up linearly over time. Do you understand that?

[tvbcof]

That "bloat" when maxed out will increase chain size linearly, while the cost of storage goes down exponentially. It's really an anomaly of young bitcoin that we had blocks below the cap this long - for most of bitcoin's future, the blocks should be nearly full.

Seems to me that as long as the community does not choose to fiddle with the block size, the fees will start to take care of SatoshiDice-class problems and it should remain practical for the system to be operational on the backs of a wide class of operators. 

You didn't read. Sure, the blocks are full but the problem is that the SatoshiDICE losing confirmations are unspendable and unprunable. Do you understand this distinction? A normal transaction output stays in the block chain until it is completely spent at which point it is eligible for pruning and can disappear forever. It only "bloats" the block chain until it is spent. Compare this with SD's losing confirmation, which can never be spent economically and therefore can never be pruned. Do you get it now? Why won't it read?

Ya, so they stay in the block chain, bloat it, and cause overhead and nuisance.  As long as it is predictable I don't see it as a terminal event.

Perhaps an attacker could use is as a DOS mechanism by deliberately leveraging the weakness at a significant monetary loss.  In that case it probably would be possible to make a fork which a significant chunk of the userbase would adopt (if that is what it took to deal with the issue.)  Or just shift to a crypto-currency solution which didn't have this weakness.

[zebedee]

You didn't read. Sure, the blocks are full but the problem is that the SatoshiDICE losing confirmations are unspendable and unprunable. Do you understand this distinction? A normal transaction output stays in the block chain until it is completely spent at which point it is eligible for pruning and can disappear forever. It only "bloats" the block chain until it is spent. Compare this with SD's losing confirmation, which can never be spent economically and therefore can never be pruned. Do you get it now? Why won't it read?

I've explained why this is false.  Why do you keep repeating a falsehood?  Ax(e) to grind?

[niko]

I've never heard of anyone except evorhees being involved with SatoshiDice. You should really be careful with throwing names around like that.

Perhaps mrbigg has privileged access to non-public information? Perhaps this access has something to do with his job of posting amazingly inflamatory idiotic statements such as

Bitcoin under ATTACK now: What YOU can do to help!

Blockchain.info Colluding with SatoshiDICE to spam the network?

?

[Sukrim]

In recent times mrbigg has started to troll hard with half-true information, similar to a user which name was a certain greek deity who carried the world on his shoulders...

If unprunable dust output becomes a problem, there IS incentive to include "collecting" transaction in the network - you have a big transaction input, but can prune each of these from your list. Once storing your list is more expensive than block space, you include these transactions. If it isn't, it's also not a problem to store them.

Also, there is a hard cap of 21 million times 10^8 Satoshis. You can do the maths how big the maximum set of pruned unspent outputs can get and this isn't even a VERY scary number (Facebook soon will store more data than that). This is the absolute maximum EVER that can be reached, no matter the block size, no matter how many people attack BTC.

By the way, I'm also pissed that SD is using several gigabytes(!) of my HDD space on my full node and only paying miners for including these (I personally believe there should be full nodes not only run by miners or at a loss) but instead of spreading information that I bend a bit to sound informed or plausible, I'd rather release a "no-SD" patch and maybe even binaries for pools to use. There STILL is no real bitcoin client that is tailored towards miners and their needs, most clients are written for end users (=transaction generators, not transaction confirmers).

Also this issue is not new at all, since day 1 SD has been under attack by people for creating 1 Satoshi bitdust. They didn't care back then and they won't change their operations now that they risk their business. Even the "let's track SD earnings" thread in the dev forum was also partly created to also see how much they bloat the block chain.

All you do with this thread is creating a Streisand effect - every publicity is good publicity...

[Come-from-Beyond]

Also, there is a hard cap of 21 million times 10^8 Satoshis.

I bet u r wrong. Some guys say that bitcoins can be infinitely divided.

[Sukrim]

This requires a hard fork, if you create more than 3 additional digits (smallest unit is 1/100th of a nBTC) you overflow an 64bit int and once this is necessary, I guess there are a LOT of other problems first.

Currently there is also a hard cap of 1 million bytes/block. There are ideas to change that, but as it currently stands these are the rules every block on the network must follow to be valid. Of course in the future BTC could be divided down to 100 decimal places, blocks could be 100GB per block etc. - all these things are (maybe?) desirable or wouldn't change the main idea behind bitcoin a lot, but still currently and without a widely accepted hard fork they are out of question.

[Killdozer]

I bet u r wrong. Some guys say that bitcoins can be infinitely divided.

How much do you bet?

[Come-from-Beyond]

I bet u r wrong. Some guys say that bitcoins can be infinitely divided.

How much do you bet?

21 million times 10^8 Satoshis + 1 satoshi more

 

[Peter Lambert]

If I have other coins, say 4 BTC, and want to spend 1.25 BTC, it's trivial to create a transaction with 4 BTC + 1 or more single-satoshi inputs, that pays 1.25 + 2.7500000x BTC change as outputs, and pay no fees.

So not only can they be spent, they can even be spent and pruned for free.  Yes, I've done it myself.

First of all your trick only works once, if the 4 BTC come from a transaction output that is sufficiently aged. So it's not "free", you're consuming the limited resource (old coins in this case). You could not, for example, use the same 4 BTC again to combine another single satoshi input for "free".

Second, assuming that you did not have aged coins you would have to pay a fee. The 36 bytes for the additional input (the 1 satoshi) will drive the fee up by more than one satoshi. This is what we mean by "economically unviable to spend."

People talk about "old" coins, and "aging" coins, to avoid a fee, but what is the actual meaning of "old"? How long do they have to sit, is it a coupe blocks, a couple days, a couple weeks?

[mobile4ever]

Yep, I agree and there is no question that SatoshiDICE has done some great things. But they are also pushing a burden of disproportionate costs onto all full nodes that exist or will ever exist in the future. No other Bitcoin enterprise behaves this way.


All that SatoshiDICE needs to do is pick one of several options for changing the way it sends unspendable outputs to notify gamblers of losing bets.

They could hold back those small payments and have them build up until such time as the receiver wanted to withdraw them.


Ditto on bitcoin is "still in beta" - we are all still "early adopters".

[Mageant]

I think this "problem" is only temproary will fix itself when the price for transactions go up.

This could happen when we start hitting the hard limit of 1 MB per block.

[crazy_rabbit]

Won't the problem solve itself when 1 bitcoin is worth 1 million dollars and that tiny little bit of transaction "spam" suddenly becomes very valuable, and miners lower fees in competition to handle the vast bitcoin traffic?

[Come-from-Beyond]

Newly Discovered Flaw, Could KILL Bitcoin!

I'm afraid Bitcoin #1 is already dead. The only questions we should discuss:

1. How long can we pretend that everything OK?
2. What is the plan for moving on to Bitcoin #2?

[conspirosphere.tk]

I'm afraid Bitcoin #1 is already dead. The only questions we should discuss:

1. How long can we pretend that everything OK?
2. What is the plan for moving on to Bitcoin #2?

3. When its price is going to crash? Deads in the water didn't use to be so expensive.

[SalvorHardin]

This is a re-post of my comment from the other thread.

I am not sure why no one has mentioned it here.  But starting today, Satoshi dice already changed its method of sending out 1 satoshi for confirming losing bet.  Instead, they are now sending out 0.5% of bet amount back for losing bets, regardless of bet size.  Given that minimum bet is 0.01 BTC, now all Satoshi Dice losing bet tx starts at 5000 satoshi and up.  I presume that SD changed its policy due to the introduction of SD filter patch.

Therefore, Satoshi Dice now absorbs all tx fee, and return 0.5% of bet amount, even for minimum bet.  Previously, it will deduct a tx fee first and send you 1 satoshi.  So you need to bet at least 0.2 BTC or so, to get 0.5 mbtc back.  But now even 0.01 BTC minimum bet gives you at least 5000 satoshi back.  I presume the change is made today because of all the heat and the patch here.

[Come-from-Beyond]

I'm afraid Bitcoin #1 is already dead. The only questions we should discuss:

1. How long can we pretend that everything OK?
2. What is the plan for moving on to Bitcoin #2?

3. When its price is going to crash? Deads in the water didn't use to be so expensive.

When? When most of traders realize that Bitcoin has a serious flaw.

[Soros Shorts]

This requires a hard fork, if you create more than 3 additional digits (smallest unit is 1/100th of a nBTC) you overflow an 64bit int and once this is necessary, I guess there are a LOT of other problems first.

If you plan for it now to happen in 20 years it does not need to be a hard fork if 99.9% of the clients have been updated to expect it. It is not as though there is a current need for something smaller than a Satoshi. Also if you are worried about performance problems, I would bet that by then the ALU in your watch can do native 1024-bit int arithmetic operations.