I know this is against the free market but this thread should be PINNED! A P2P Exchange is a MUST and gollum has started us in the right direction. This need the full focus of the entire community right now. Maybe we should even start a fund for the developers.

Wow, there are a lot of threads popping up today on new ideas for Decentralized Exchanges. (I can't imagine why...) I started one myself 5 hours ago.

Allow me to throw some ideas on the fire, as this thread seems to have gotten further than the others.

1. Your concerns about how to port in real-world cash are unfounded. There already exists an industry of businesses that will take cash in many forms and give virtual credits, including bitcoins, for them... BitInstant, Dwolla, OKPay, coinbase, ArumXchange, and many others can upload cash into this exchange in one or more forms. Let them deal with this headache; there are many of them so there is no central point of failure there.


2. To really replace MtGox and have real-time graphs and trades like we all want to see, the core of this beast must always determine the other clients running within a short enough pingtime's distance. (So trades can happen in milliseconds, not multiple seconds.)

If a client logs on, looks around for other clients, & notes all of their pingtimes, the client can be a real-time, graph-displaying trading platform for everyone within a decent pingtime of you!

True, this means that someone in Kansas isn't likely to see the trades of someone in Perth... But that's the nature of decentralization. The going price of a bitcoin would therefore not be global, but regional... Although the regions are likely to be continents, not cities. You'd trade quickly with those closer to you (pingtime-wise) and if you can't find your trade there, I guess it would be a good idea to include an OTC board in the package too, or port in #bitcoin-otc somehow.

(Note: VPNs could be used for people that live far away from population centers to reduce pingtimes, too!)


3. No marketing will be needed. (!) If each client for this exchange gives the fees to the person running the client where the trade happened, then it is self-incentivized to spread much like how bitcoin miners get paid to mine. I've got some very specific ideas about how to make this part work, but for now all you need to know is that everyone with an internet connection would want to run this client because it would literally pay them to just run on their PCs. (And they wouldn't need expensive GPUs.) More details on this step later.


4. No miners needed. A lot of the data that runs across these clients need not be encrypted... You want the world to know what the current price for each currency is, for instance... It's just the ownership of things on it you must secure, so no processing of SHAs here, just good, secure ledger keeping.

Every client should be both a place for you to log in and make trades, and a place for other people to trade on. If run this way, it will use a bit of bandwidth (think: bittorrent) but not any advanced hashing that requires special equipment like GPUs.


Enjoy the ideas; I'll add some more later.

Just throwing this out there:
Take an idea similar to bitmessage, where everyone generates their own address ID. Everyone also stores their own coins, and their own USD.
The order book is shared to everyone running this client, and can be modified by anyone to add and remove only their own transaction requests, signed by their own ID. Some proof of work could me included so it takes a second or more to place your order so mitigate spamming by tons of transactions
Anyone is free to put up an order, either sell or buy, with their own conditions, such as method of payment (Dwolla/PayPal/Wire), fees for cash transfer, etc. The order book organizes these so that the fee is included in the price being traded (e.g. all PayPal trades could be 10% above wire trades, etc).
Once a match is found, the transaction is locked and is accessible only to the people executing it. It is then up to the two people to get in touch (possibly using bitmessage type system build right into this) and actually carry out the trade off the system, e.g. sending btc to someone after they send cash, etc.
Once the trade is executed successfully, both parties must mark the transaction as successful, which increases the reputation record of the address IDs. This can be done by both addresses signing that the transaction completed successfully, and will be in both parties' interest to store a copy of. Since both have to sign, those can't be faked. Tying this to the OTC WOT (bitcoin-otc web of trust) would help, too. Someone could try to game the system by trading with themselves, but the risk is that their offer will be picked up by someone else, which means they'll have to carry out the trade, or get a bad rating.
Eventually, certain high reputation level nodes will emerge to be able to handle transactions for those more skiddish, possibly at a higher fee.

Concerns: still may be possible to game the system, will likely be easily succeptible to scams and chargebacks (at least until trusted nodes emerge), will not be able to handle high frequency trades, may be difficult to fill a large order that spans many smaller ones (instant buy on MtGox), and finally may be extremely inconvenient to have to ask for cash from someone directly, v.s. just having a cold anonymous system do it for you.

Of course this could also be expanded to easily include trades of MtGox codes for BtC, too, in which case, like was mentioned, big trusted companies could be the money in-out places.


It seems we have a lot of bases we can start with for the code already, be it Ripple, P2Pool, or Bitmessage...

1. Your concerns about how to port in real-world cash are unfounded. There already exists an industry of businesses that will take cash in many forms and give virtual credits, including bitcoins, for them... BitInstant, Dwolla, OKPay, coinbase, ArumXchange, and many others can upload cash into this exchange in one or more forms. Let them deal with this headache; there are many of them so there is no central point of failure there.

All of these ideas floating around here rely on already existent exchanges to facilitate a 'p2p exchange'. If this were p2p i could convert my fiat into coins and back without any sort of legal intermediary.

If coloured coins are used, then you can have a trade which is signed by both participants.  If it is included in the main bitcoin chain, then it counts, otherwise, it is fully reversed.

The buyer creates an offer transaction of the form:

Input 0: (0.000001, coloured) -> signed by BTC buyer with "SIGHASH_ANYONECANPAY + SIGHASH_SINGLE"
Output 0: (10) -> "Send to buyer's address"

This transaction is invalid, so can't be submitted to the main network.  However, the SIGHASH_ANYONECANPAY rule allows someone to add extra inputs to the transaction.  The SIGHASH_SINGLE rule allows adding of more outputs.

Input 0: (0.000001, coloured) -> signed by BTC buyer with "SIGHASH_ANYONECANPAY + SIGHASH_SINGLE"
Input 1: (10 + fee, coloured) -> signed by BTC seller (i.e. $ buyer)

Output 0: (10) -> "Send to buyer's address"
Output 1: (0.000001, coloured) -> "Send to seller's address"

It is now a valid transaction.  If the transaction makes it into the chain, then the trade occurs.  It not, then it is reversed.  It is atomic, it either fully works, or fully fails.

This system allows 2 people to trade.  It lets a buyer post "Will buy 10BTC for $1000".  However, for an exchange, you want to post "Will buy 10BTC for $1000 or market price, if lower".