I agree that a proof-of-work based consensus algorithm (like in bitcoin, and in my P2P exchange proposal) is energetically much more expensive than the Ripple consensus algorithm (which is described at https://ripple.com/wiki/Consensus ).

And I agree that there's no reason why a P2P exchange has to use a proof-of-work based consensus algorithm - it just needs to use a well-proven P2P consensus algorithm.

But ... the bitcoin consensus algorithm has been demonstrated to work well over the last four years.

How confident are people in the Ripple consensus algorithm (which hasn't been tested yet)?

What other consensus algorithms can be considered well-proven in the face of huge botnets and hostile clients?

There's plenty of market manipulation possible even on a fully accredited exchange - e.g. "ramping", "painting the tape", "ghosting" (and all kind of other tricks that were routine in the early days of stock exchanges!). It's hard to know if it market manipulation is happening or not - we just have to trust the "fully accredited" exchange and the regulators to keep the market honest.

One advantage of a P2P exchange with a global order book block chain is that all orders ever placed can be seen by all - making detecting market manipulation much easier.

There may also be differences in incentives - a P2P exchange can be run for the benefit of all, whereas a central exchange is run for its own benefit (and may choose to ignore manipulation as long as it is not too blatant). For example, stock exchanges could use a "discrete double action" system to provide fairer order matching and limit high frequency trading. Why don't they? Perhaps because that would mean less volume and hence less fees ...

On the other hand, when it comes to the "looking after the fiat money" side of things, yes, I would prefer the fully accredited exchange!

Ah, I hadn't noticed the "Project Development" board until my thread was moved here  

The proposal above has the following aspects of the ideas discussed on https://bitcointalk.org/index.php?topic=172705.0:

• one global order book (backed by consensus block chain)
• order matching occurs periodically (when order book block filled)
• multiple brokers (traders specify the broker(s) they trust)
• no coloured coins or new currency required

I didn't anticipate the need for trading between ClearingHouses / Brokers / Gateways - perhaps I missed somthing?

As freequant points out, yes, I broadly agree this proposal is a "subset" of Ripple.

But is that necessarily a bad thing - Ripple suffers from:

• being centralized (at present)
• being proprietary (at present)
• not being in widespread use (at present)
• solving a much bigger and more abstract problem
• introducing a new virtual currency with some dubious "pre-mining"

I just think it would be a shame to see all attempts to design a decentralized exchange abandoned in the hope that one day Ripple will take off ...

tumak - Yes, I should have said P2P rather than merely decentralized. I guess I will have to think about "conditional block withholding attacks"!

There's been quite a few suggestions about decentralized exchanges
but they've generally petered out into either "Ripple will solve all
our problems (one day)" or "a decentralized exchange is impossible".

Here's what I believe is a fairly novel (but perhaps workable) idea
for how a decentralized exchange might work technically ...

Our decentralized exchange is a P2P system like bitcoin. There is
no central authority required.

It has a bitcoin-style block chain consisting of OrderBookBlocks.

Like bitcoin, proof-of-work is used to ensure everyone agrees on
the contents of the OrderBook block chain.

Each OrderBookBlock in the chain represents a snapshot of the "order book" -
that is, a list of the open Orders waiting to be matched on the exchange.

An Order in an OrderBookBlock would look something like this:


Once an OrderBookBlock is complete, an agreed deterministic
algorithm can be used (by anyone) to determine which Orders
matched with which Orders. An Order can only match another
Order if its Pledge can be verified with one of the ClearingHouse
public keys listed in the other Order.

The traders can then collect their bitcoins or fiat from the
Clearing Houses, using the OrderBookBlock as proof that they
are entitled to the pledged funds.

Thoughts? Obviously there would be lots of details to be worked
out (fees for OrderBookBlock miners, protocols for orders and
pledges, a fair matching algorithm, ...).

And another one here ...

https://bitcointalk.org/index.php?topic=158010.0

The demand is definitely there, anyway!

There's been a couple of threads on the topic of the lack of a UK-based exchange just in the last week:

The UK Public & Bitcoin
https://bitcointalk.org/index.php?topic=176717.0

How to solve UK Bitcoin Exchange Problem?
https://bitcointalk.org/index.php?topic=177059.0

Seems to be a general consensus that keeping the banks and the regulators happy may just be too hard/expensive.

Having said that, if buying bitcoins is treated broadly like sending foreign currency abroad then it occurs to me that there's tons of tiny little money remittance places all over London (often run out of the back of an ethnic food store / internet cafe) that let you send money internationally (often to somewhat unstable countries!) - surely they can't all have paid vast sums of money to register with the FCA?

Supposing a UK based exchange did manage to satisfy the banks and regulatory agencies that it was in compliance with all the relevant legislation, would anyone actually use it?

The Money Laundering / Terrorism / Payment Services regulations would seem to suggest that a compliant UK exchange would look something like this ...

Registration would involve:
- Sending a certified copy of Passport / Driving Licence (that is, a copy signed by e.g. an accountant / notary / bank manager).
- And a certified copy of a recent bank statement (not an online one!).
- Entering your address history for the last three years.
- Consenting to a credit check / electoral roll search.
- Waiting for a secret code to arrive by post and entering that to verify current address.

There would be no ability to send or receive bitcoins other than to/from a single bitcoin address you control. Changing that bitcoin address would probably require a lengthy process involving sending documents by post.

Deposits and withdrawals would involve paying a 1% fee to cover bank charges (faster payments aren't free for businesses). They would only be possible to/from your verified bank account (which must be in your name).

Additional fees would probably have to be levied to cover the cost of regulatory compliance - say 0.5% on every trade.

Any transaction you make (deposit, withdrawal, trade) may be suspended for up to a week if it looks a bit unusual or suspicious - but the exchange wouldn't be able to tell you why they suspended it, since that would be a criminal offence ("Tipping Off").

Deposits and Withdrawals above £10K (or that add up to £10K per month, or that look "a bit funny") would probably require such extensive due diligence that even talking about the due diligence is illegal.

Sounds fun, eh? Any ideas for domain names - I notice www.painfulbitcoins.co.uk is free.

Thanks ribuck (and matthewh3), your explanation of how Ripple can facilitate a banking system funds for bitcoin trade is very helpful.

I had contemplated a cash-less exchange, but I hadn't contemplated the possibility of a cash-less and bitcoin-less exchange! It is obvious though once you start breaking down what an exchange really is.

I do hope Ripple (or something like it) takes off, though I wonder if the concept will prove too abstract.

One possible criticism is that Ripple simplifies problems (which is good), but it also multiplies them; instead of my original problem of finding one trustworthy UK bitcoin exchange, with Ripple I now have three problems:

• How to find a trustworthy gbp -> ripple-gbp agent?
• How to find a trustworthy ripple-gbp -> ripple-btc agent?
• How to find a trustworthy ripple-btc -> btc agent?

Yes, TransferWise does seem rather good for sending money to the EU/US.

I guess they can't really help with selling Bitcoins for GBP though? Or can you withdraw from MTGOX to Transferwise who then send it to the UK somehow?

Besides, it does still seem like we should have a UK exchange - we don't have to wire money to the US to buy things from Amazon, do we? Why should we have to for bitcoins?

Yes, I guess so but perhaps I'm not understanding Ripple very well.

It all seems very general and a bit circular.

Let's suppose I have GBP in a bank account. I want to trade them for bitcoins.

I can issue an IOU for my GBP funds via Ripple.

jbloggs can issue an IOU for his bitcoins via Ripple.

That's great, now we can exchange IOUs via Ripple.

I now have jbloggs's IOU for bitcoins, and jbloggs has my IOU for my GBP funds.

So now I have GBP in my bank account (which I've promised to jbloggs via Ripple) and jbloggs has bitcoins (which he's promised to me via Ripple).

But now what do we do? We seem to mostly be back at square one!

But do they need to pay each other?

The exchange makes commissions in bitcoins from trades. It can sell bitcoins if it needs fiat money for operational costs.

Provided the exchange trusts the payment processors, it can allow their exchange balance to go negative (as would be needed if the processor has handled more deposits than withdrawals). In the long run, deposits and withdrawals should roughly balance (hopefully!).

If a payment processor's balance with the exchange becomes too negative, they could deposit funds via another payment processor. And similarly if their balance with the exchange becomes too high - they can withdraw via another payment processor. An agreement could be reached where payment processors do charge no or very low commission to other processors. This would also be necessary if a payment processor wants to cease processing payments.

What I'm calling "The UK Bitcoin Exchange Problem" is that currently, there are few options for buying/selling bitcoins for GBP bank transfers in the UK (see various posts on here).

It would be nice (and potentially lucrative!) to solve this problem.

There are a number of marketplaces such as localbitcoins and bitbargain that match individual UK buyers and sellers, but buy-sell spreads are very high (over 10%) and there is some level of risk involved for buyers.

There is the possibility of transferring money in EUR/USD to non-UK exchanges - e.g. bitcoin-24, mt-gox, but that's slow, fairly expensive and introduces currency risk.

Previously there have been several exchanges that accept GBP bank transfers but they all seem to close or stop accepting GBP bank transfers.

This is somewhat surprising since there is a reasonable amount of interest in bitcoins in the UK, and the UK "Faster Payments" system allows bank transfers to be sent in minutes at very low cost.

So why no UK exchanges?

Some folks claim this is because the UK banks are threatened by bitcoin, and rush to shut down accounts used by bitcoin exchanges.

However, I doubt the banks are that paranoid - it seems more likely to me that bitcoin exchanges have their bank accounts frozen due to their accounts showing up as the destination for stolen funds.

For example, if a criminal hijacks a UK bank account (a common problem in the UK) and transfers the stolen money to the bitcoin exchange account, there is a good chance that the banks will flag that exchange account as "potentially dodgy" since it received stolen money, and then freeze it once several transfers like it happen.

So what can be done?

The fundamental problem as I see it is that the bank account used by the exchange to receive funds is a single point of failure - if the banks freeze it, the exchange cannot operate since users cannot deposit or withdraw GBP funds.

The solution then is to move to a decentralized system where the exchange no longer relies on a single bank account.

The trading activity on the exchange - that is, exchanging bitcoins for cleared GBP funds - can still remain mostly centralized.

The bit that needs decentralizing is the conversion of GBP funds in bank accounts to/from cleared GBP funds that can be used on the exchange. This is sometimes called "clearing" - as in a financial "clearing house".

(There has been talk of the Ripple system as a really, really decentralized solution - but that's solving a much bigger and harder problem!).

So here's an idea for the start of a possible, simple solution ...

As normal, users have two balances with the exchange - a bitcoin balance and a GBP cleared funds balance.

However, the exchange is cashless - end-users can only deposit and withdraw funds via "Payment Processors" (which are independent of the exchange).

Suppose jbloggs wants to buy some bitcoins. Here's what happens:

jbloggs registers with the exchange.
jbloggs makes a bank transfer to a Payment Processor.
The Payment Processor instructs the exchange to transfer exchange funds from the Payment Processor's exchange account to jbloggs exchange account.
jbloggs can then buy bitcoins on the exchange.

Later, jbloggs sells bitcoins on the exchange and receives exchange funds in return.
jbloggs then instructs the exchange to transfer exchange funds from his exchange account to the Payment Processor's exchange account.
The Payment Processor makes a bank transfer to jbloggs.

In return for taking on fraud risk and looking after the funds, the Payment Processors take a small commission on all deposits and withdrawals.

In return for providing trading and account management services, the exchange takes a small commission (in bitcoins) on all trades.

I'd be interested to hear people's thoughts on this? Could it work? What would stop it working?

It's been tried (and mostly failed, apart from embedded smartcard devices) several times: http://en.wikipedia.org/wiki/Java_processor

IMO, the reason is this ...

Intel and AMD invest a huge amount of money into building the fastest possible general purpose CPUs. A modern chip fab costs literally 100s of millions of BTC. They also have decades of experience designing chips.

It's true that in theory a CPU designed to execute Java directly could be faster than an Intel/AMD CPU running a software JVM. But in practice, no-one else has the money and experience to build a CPU anywhere near as fast as Intel/AMD - so that advantage remains theoretical.

Where you can beat Intel/AMD is by concentrating on tasks that general purpose CPUs aren't very good at - such as tasks that benefit from massively parallelism. Things like graphics rendering, video decoding, SHA256 hashing, ...

However, executing Java bytecode isn't one of these tasks - it's a complex, unpredictable, non-parallelisable task.

A company called Azul http://en.wikipedia.org/wiki/Azul_Systems came close to succeeding with custom Java hardware - they used to produce special hardware (Vega Java Compute Appliances) which massively sped up Java garbage collection in particular. However, it was never quite compelling enough to make companies decide to go out and buy very expensive Azul servers instead of just buying more off the shelf Intel x86 servers ...

Thanks Bitcoinin.

I've also noticed that the bitcoind JSON-RPC response times do seem a bit high, especially if running on modest hardware. But I've not investigated further.

I wonder how large scale bitcoin sites (e.g. satoshidice, mtgox) perform operations such as checking for transactions and making transactions - do they use:

• the standard bitcoind JSON-RPC interface
• a lightly patched bitcoind
• a heavily customised bitcoind
• a proprietary client that speaks the bitcoin protocol
• an application that bypasses bitcoind and pokes around in the bitcoind database directly
• a mixture of the above!

Quick question ... do people mining as part of a pool directly participate in the Bitcoin network as peers, or do their mining apps just talk to the server running the pool? Or does it vary from pool to pool / app to app?

Neat idea!

Sorry to hear about your loss of BTC Steven, hope you manage to get something back!

I'd be interested to know what prevents online wallets like MtGox from:

• never giving an address used by one user to another user
• remembering all addresses used by a user and considering them all to belong to him

since that would seem to prevent these sort of problems.

Is it a technical reason - I thought addresses were cheap to
generate, but perhaps having to cope with lots of addresses
per user is tough?

Or is it more of a philosophical thing based on the original
intention that addresses aren't used over and over by the
same user (like how change goes to a different address)?

still down 

Thanks, very helpful suggestions.