Primer for a P2P Distributed Exchange

[nii236]

But if you were an exchanger wanting to sell dUSD, getting 50% of the worth of BTC back does not give any incentive to become an exchanger!

[1715010036]

1715010036

[1715010036]

1715010036

[1715010036]

1715010036

[Mylon]

Hi BTCLuke,

When I initially read your post, (couple of days ago) I immediately saw the potential, the one thing that does not work however, is fiat -> BTC -> fiat, in the entire scheme you posted. (also really bonus kudos to the guy that basically already wrote the txn scheme on how to handle the p2p exchange, crypto only, as that is the way to go)

^^ We all know this can be done, and perfected as long as we are dealing with crypto currencies, with a blockchain to actually see if someone is able to spend that.

While I wanted to reply earlier, I didn't get to it being very busy. And then again I hadn't solved the fiat -> btc yet. Now I think I have gotten the clue from bitpay after their talk on the conference, and I truly believe that in combination with this p2p exchange it can work.

Forget instant transferring USD, EURO or any other fiat, it needs to digital and verifiable. (as long as there are bankers, they will not allow us to do this.)

So we gotta go outside the box... (bitpay's words)

Grab any gift card you can think of, that is automated verifiable, whether it is still valid or not.

For our example we will use Amazon and Ebay gift cards.

If I go to the store I can most likely buy an Amazon $50 Gift card for $30.

If I can check this card online whether it is still valid, then this cards code is worth $30. (despite being $50 Amazon gift card)

So I have a code, that i can check on the fly, and trade...

Now I hear you say, I don't trust Amazon. Ok fine then go Ebay.

Since there is such a huge market of gift cards, by different companies, all you have to trust is a company you already trust, that has these gift cards.

The most important part here is, to use the existing gift cards from major companies. (who people already trust, to be the intermediary between fiat and BTC)


The other thing to realize is, that crypto exchanges can do the exact same thing on this same platform.

While sharing the same platform, they can benefit from it (easier access to more crypto currencies, all of a sudden your not just exchanging bitcoin, you can exchange anything you have the client / plugin for.)

A BTC-E USD IOU, all of a sudden can be traded against a MtGox USD IOU, or MtGox Euro IOU, but at the same time vs an Amazon Gift card, BTC or LTC.


Any exchange can join, and give out IOU, depending on how much reputation they hold, their IOU's will be worth more or less. (you wanna facilitate trading of these IOU's to keep prices down)

Any Gift Card from a Major / Medium sized Company can join, as long as they are instantly verifiable as not spend. (aka, you're using your mutual trust in this company, without needing their consent)

Any Crypto Currency can join, as they are backed by a blockchain.

(And many other things that I haven't thought of, that are instantly verifiable over the internet, or have enough reputation, that people start trading their IOU's)

The combination of these 3 on a shared exchange platform, using the security of the bitcoin protocol, has the best chance of success in my opinion.


Last but not least, you could also build a sort of "localbitcoins" into this, though it would technically be nothing more than a message board, asking where to meet, with a reputation system in it. (someone looking to buy BTC, posts a message he wants to buy, the moment they meet and sell, the seller will sign the transaction as being completed, on which the system will see the transaction as completed as they see the bitcoins appear into the persoon looking to buy BTC, and anybody trying to up-rate himself, will be killed by the fees of the network upholding the exchange)

Looking forward on all input on this.

[CypressXM]

While I'm still more or less convinced that a P2P exchange of BTC=>Fiat is not feasible, what is feasible is very interesting.

What's feasible is a P2P exchange of BTC=>AlternaCoin

A totally distributed P2P exchange! This could be done now...There are no technical barriers that I can think of.

[justusranvier]

What's feasible is a P2P exchange of BTC=>AlternaCoin

A totally distributed P2P exchange! This could be done now...There are no technical barriers that I can think of.

Just the economic barrier that there's no market for this service because nobody cares about altcoins.

[CypressXM]

What's feasible is a P2P exchange of BTC=>AlternaCoin

A totally distributed P2P exchange! This could be done now...There are no technical barriers that I can think of.

Just the economic barrier that there's no market for this service because nobody cares about altcoins.

True, I would probably get a few LTC but that's about it.

[phillipsjk]

minted bars like this one designed to break apart in small groups of 1g bars or as individual 1g bar are great for our future p2p bitcoin gold exchange. gold is the solution. i read people waiting more than 20 days for wire transfer from mt gox or accounts frozen for id problems etc. that is crazy! nobody has the right to interfere with your private transactions. with the gold exchange there will be no such delays or interference. you can do it over the net or by visiting your nearby bullion dealer. nobody will get in your way.

fiat --> gold --> bitcoins
bitcoins --> gold --> fiat

Gold dealers are required to report large suspicious transactions in Canada, and likely other countries as well. Gold is also much easier to counterfeit than Bitcoin.

[BTCLuke]

Whew... Had to take the day off yesterday because this was just too much to absorb.

Now with a fresh head, I can see that all of the proposed solutions to the fiat problem still seem to forget that whatever it is that we'll be trading as Fiat on this exchange must still follow the economic requirements for money!

Remember, if it isn't convenient to trade a form of money, then it won't be used as money for long. If you want to argue this point please head over to Mises.org and argue with them. They have a trillion different arguments to counter yours with.

The closest thing mentioned on this thread (or a linked thread) to a Fiat replacement that a convenient P2P exchange can use is OT's Ricardian contracts. If Bearer-bond contracts were made in 1 USD or 1 yen or 1 euro allotments, then we can trade them fluidly in place of fiat.

I like to think of them as little cryptobonds.

1 USD + a fee is traded to some trusted source for a $1 cryptobond, which is a tradeable hash that can fit into the wallet on this P2P exchange and be traded as quickly or even quicker than a bitcoin can between clients. Of course making sources trusted but not centralized then becomes the problem... I'll leave that problem for another day.

But the important thing in this thread is the criteria. To rephrase the original criteria with the change in #6:

A P2P Distributed Exchange MUST:

1. Be without any central points of failure, since a government or two WILL be coming after it one day. I suggest a Bitorrent-like software schematic.
2. Show everyone a very large number of possible trades to choose from, (thousands?) so assets can form a stable price. (e.g. a Bitcoin is going for $120)
3. Transact trades pretty much INSTANTANEOUSLY, so when you're watching a graph and want to trade at a very specific time you can do so. (This is extremely important for arbitragers and other traders who help keep the price fluctuation down.)
4. Offer Graphs and APIs for for graphing like MtGox does.
5. Have three-user (trustless) trading, so a non-interested 3rd party always hosts the trade between the buyer and seller. (And should provide Escrow too!)
6. Hold and transfer a cryptosecurity that perfectly represents fiat and has all of the characteristics of real money. (Divisibility, fungibility, Malleability, Scarecity, etc...)


Again, my exchange idea is a work in process but these criteria are something I'm sure we will all want to agree on for your own projects or mine too... So please, think for a moment what else a working P2P Distributed exchange will need to be and do, and help me ensure we haven't missed a criterion on this list.

Thanks for all your hard work!

 

[bytemaster]

@btcluke

I want to start by saying I appreciate everyone who takes time to post thoughtful responses.  So I will respond to your post stating I fully believe I can satisfy all of your requirements.  So much so that I have posted another thread where I am looking to hire someone to help me and invest $20,000 or more in creating a working system based on my ideas.   But, as a sanity check I am also offering a 10 BTC bounty to anyone who can convince me that the idea will not work and save me time and money.    That said, let me attempt to explain my idea in greater detail.   I am also willing to discuss my ideas on skype with anyone interested in helping or attempting to claim the 1 BTC bounty.   (my handle isn't exactly anonymous because I have tied it to my website the-iland.net and my github account (also bytemaster)), send me a private message for skype (or ichat).

Now that I have your attention let me attempt to address your requirements by first stating that I am a very well read student of mises and austrian economics and thus am basing EVERYTHING on an austrian understanding of economics and will clarify anything I that you think I might misunderstand.  I am actively want to be proven wrong if I am wrong.

First of all we must understand that all forms of USD have different value despite being denominated in USD.  USD on Mt.Gox is less valuable than a paper note in my hand because it costs time and money to move a deposit from Mt. Gox and convert it into a bill in my hand.  It is also less valuable because there is counter-party-risk where the counter-parties are Mt.Gox and various governments and banks that could all prevent me from redeeming my Mt.Gox USD.    Even money in my bank account is of lower value than USD in my hand (provided I need it now) because of ATM fees, daily withdraw limits, and of couse counter-party-risk.

We must also recognize that at other times paper-USD is of less value than Mt. Gox USD such as times when you want to buy BTC NOW to take advantage of the price movement.   Therefore the price of Mt. Gox USD is above or below parity depending upon the relative demand for deposits vs withdraws from Mt. Gox as well as the ease of trading balances within Mt. Gox.

With that understanding we can clearly see that it is only an illusion to claim redeem-ability at 'face value' with any issuer.  Every issuer has a different trust level with every individual and therefore USD notes issued by different issuers are not perfectly fungible nor free of counter-party-risk.  The issuer is always subject to seizure from the government and could default regardless of how honest he intended to be.   The conclusion we can draw from this is that there might as well be a market for crypto-USD that fluctuates AROUND actual USD prices give-or-take a couple of percent similar to the price variance between Mt. Gox USD and paper USD which is what Bitinstant's entire business model is designed to capitalize on.  Thus if we can create a crypto-USD currency that has VALUE recognized by the market yet without any redeemable IOU like contract (like Bitcoin) yet trades at or near parity of paper USD base solely on market forces (not price fixing) then we will have a workable system.

Another aspect of depositing (trusting) money with a 3rd party is the expectation of interest payments in return for the risk associated with lending your money to the bank.  Any bearer-bond contract that does not pay interest will always be subject to immediate demands for redemption and thus put heavy stress on the most expensive / difficult part of converting fiat to crypto-fiat (redemption).

1. Be without any central points of failure
  - a crypto-USD that can be traded via systems like localbitcoins at prices near parity with the dollar would lack any central points of failure.  Assuming you could provide a way to imbue value into the crypto-USD independent of any backer or price-fixing.  (It can be done, more on this later)

2. Show everyone a very large number of possible trades to choose from
  - by separating the process of converting paper-USD to crypto-USD and trading crypto-USD for BTC or crypto-USD you enable all trades to occur through broadcasts on the network or offers in the blockchain.

3. Transact trades pretty much INSTANTANEOUSLY
  - assuming there exists crypto-USD and BTC then an exchange could be run entirely from behind a TOR node.  However, there is still value in trades that can occur at the same speed as bitcoin transactions for most people who are not attempting arbitrage.  On the other hand if the exchange were a natively transparent part of a single blockchain then there probably would be much need for arbitrage.  

4. Offer Graphs and APIs
  - considering the entire market would be part of the same blockchain this could easily be added into any desktop client.

5. Have three-user (trustless) trading
  - how about 2-user trust-less trading between crypto-Fiat and 'BTC' and using escrow for remote exchanges of crypto-Fiat for bank-Fiat?

6. Hold and transfer a cryptosecurity that perfectly represents fiat
  - crypto-USD is more divisible, fungible, malleable, and scarce than paper-USD.


I think the #1 criteria is to eliminate counter-party-risk from holding of crypto-FIAT.  

I believe I have done this with my system which I will attempt to describe with more clarity and to PROVE that crypto-USD is not an IOU, has value of its own (like BTC) and that the value is automatically moved toward paper-USD parity by market forces, and that the value of crypto-USD is not at risk regardless of 99% changes in the value of 'BTC'.

More in a follow on post...

[bytemaster]

So the challenge we have before us is to create a crypto-USD that has market value of its own at or near parity with real USD with no issuer 'backing' the crypto-USD.  I will now explain how this can be done.

First we must establish that bitcoins already establish market-value without an issuer.   So we must now find a way of using 'bitcoins' to back crypto-USD in such a way that crypto-USD maintains a value near parity with paper-USD. 

To do this there must be some way to make holding crypto-USD pay interest in BTC and have the amount of interest paid automatically adjust with the exchange rate.  So lets try to set up such a system.

1) Assume for a moment that holding bitcoins paid a dividend in bitcoin proportional half of the mining fees (miners get the other half).   This would turn bitcoins into BitShares where a BitShare can be thought of as a share in the exchange.   Suppose Mt. Gox issued shares in the exchange that paid dividends based upon profits from the exchange.  In this case the 'exchange' is the BitShare network and the Shares are 'bitcoins' and the 'fees' are the transaction fees + mining reword.

2) With the above scenario owners of BitShares own them for 2 reasons, appreciation and dividends.  The dividends encourage saving.

3) If someone wanted to issue crypto-USD they could divert their dividend payments from their BitShare to crypto-USD issued at the current exchange rate.  The result would be that crypto-USD pays dividends proportional to the exchange rate and therefore has value derived from BitShares that is close to parity.

4) Why would someone convert BitShares to crypto-USD?   Because there is a buyer with paper-USD that is willing to pay a premium to above parity to covert paper-USD into crypto-USD.  This premium would have to be high enough to justify giving up BitShare dividends and cash out into paper-USD.   This will increase the supply of crypto-USD with each conversion and each conversion will be done at the current market price of crypto-USD vs BitShares and thus the ratio of BitShare dividends paid to crypto-USD holders would follow the exchange rate.   

5) Why would someone redeem crypto-USD for with paper-USD on this market?  They would do this because they own some BitShares that are not paying dividends because they are used as collateral for $100 crypto-USD.  If they can buy crypto-USD for $95 worth of paper USD then they can profit by redeeming $100-crypto-US for a mere $95.  This profit comes because $95 paper USD allows them to free BitShares that pay dividends equal to $100 crypto-USD.  They also know that eventually the crypto-USD market will swing back the other direction.

6) The reason the price of crypto-USD vs paper-USD fluctuates between $95 and $105 is based upon the relative demand for deposits and withdraws.  When demand is 'equal' the price is approximately at parity.   This market behavior is no different than the actual price fluctuations you would see with people willing want to buy or sell Mt.Gox USD.

7) Because crypto-USD pays dividends in BitShares (proportional to their issuance), it creates an incentive / demand for people seeking a return to deposit paper-USD.  This demand for deposits of paper-USD is a source of supply for those wishing to withdraw.  Thus most of the time 'redemption' is not doen by issuing or paying off of BitShare bonds, but via the direct exchange of crypto-USD for paper-USD at near parity prices between depositors and withdrawers.

In conclusion, I believe I have identified all of the market forces required to create a bitcoin like crypto-USD that derives its value from the same source that Bitcoin does yet does not suffer exchange rate fluctuations beyond those already found in the traditional banking system.  I would even submit that due to open market competition that market forces would drive the exchange rate fluctuation to be much close to 0 than is currently provided by the closed, regulated, slow banking systems.

[Pzi4nk]

I get that what you're trying to do here is much bigger than facilitate USD to crypto-USD but I think you bring up a good point about the relative value of a dollar based on our desired ends.

The simplest way to think of this is the Mt Gox example you gave where USD ready to be exchanged for BTC (or other cryptocurrency) immediately is worth more than USD in my hand. This means there is a market for crypto-USD sitting in an exchange ready to be utilized. If there is a market why isn't it happening?

I think the reason it's not already happening is the trust issue. The fact that regardless of how much I offer you to let me use your crypto-USD in exchange for the USD in my hand, you have to believe you'll actually get my USD before you'll even set a price for your crypto-USD.

To me the central issue is trust. If you solve that problem the market will adjust to the relative value of USD in different forms and to changes in demand.

[Mylon]

To me the central issue is trust. If you solve that problem the market will adjust to the relative value of USD in different forms and to changes in demand.

I have the same problem, which why I don't like the above idea. I already have to trust the government that my fiat is worth something. Beyond that point, all I want is choice, a big huge mess of anything that get cryptographically secured and bought at a fixed rated with fiat, and combined into one big exchange, trying to determine btc -> usd and usd -> btc. Due to the mass amount of choice, nobody has to trust a central point, you just choose who to trust. (getting fiat into bitcoin always needs some trust, the only thing we can do is make sure you get enough choice in who to trust) Me for example, am already thinking about WoW Gametime cards, as one of those options that you could trade on it. Keep in mind I'm looking for 100's of different things, who are already digitally stored "money/value" that are easily cryptographically securable, to prevent double spending / scamming etc. Its my believe that this is one of the very few ways, to get an p2p exchange of the ground, that is big enough to meet everyones depends and able to get us a stable price.

It also makes things a lot easier, as when I wanna spend BTC at for example Amazon, and they don't accept it, but their gift cards are being traded on the exchange, all of a sudden I make someone happy with BTC, and he already paid Amazon the USD for the things I wanna buy.

[c0rw1n]

We need a distributed third-party.

We need to figure out a way to distribute escrow, and to keep the escrowers honest.

I see that with a swarm of gateways that just check if the fiat exists, and sign that with their own btc. I mean like a dominant assurance contract.

I'm still sorting that out...

[ivanol]

We need a distributed third-party.

We need to figure out a way to distribute escrow, and to keep the escrowers honest.

Can you just keep escrowers honest based on a reputation system? In order to build reputation an individual uses the same address for all their transactions in the market. Based on signed feedback from the addresses of counterparties they trade with they build a track record of being trustworthy, and you can make a judgement based on the amount of business they have done already as to what their reputation is worth to them and therefore how much they can be trusted with in escrow.

A way to build initial trust to start up would be to use a public address on the network that was expensive to generate, and that you are less likely to want to throw away in order to defraud someone. Two ways to do this: either use a vanity address starting with for example a string of 1s as a proof of work, or make a couple of transactions from your address with large mining fees which is then the cost of that address to you.

3. Transact trades pretty much INSTANTANEOUSLY

Is this necessary for all trades though? It helps to have some instant trades, but there is no reason that p2p localbitcoins type transactions couldn't exist within the same system. A seller of bitcoins could advertise a fiat payment method along with the price. A trade would be entered on the p2p network as "agreed" when an agreement is made, as "paid" when the buyer confirms they have dispatched fiat payment, and as "settled" when the fiat payment settles and the BTC are sent.

The settlement date could be anything from a few minutes later for a wire transfer, to several days later when cash arrives in the mail, but the prices can still be broadcast to the market to help determine a market price. It is up to the buyer/seller to decide what risks they wish to take in the trade based on each others reputation and the risks of the payment method (cash vs paypal vs wire transfer etc).

[bytemaster]

I get that what you're trying to do here is much bigger than facilitate USD to crypto-USD but I think you bring up a good point about the relative value of a dollar based on our desired ends.

The simplest way to think of this is the Mt Gox example you gave where USD ready to be exchanged for BTC (or other cryptocurrency) immediately is worth more than USD in my hand. This means there is a market for crypto-USD sitting in an exchange ready to be utilized. If there is a market why isn't it happening?

I think the reason it's not already happening is the trust issue. The fact that regardless of how much I offer you to let me use your crypto-USD in exchange for the USD in my hand, you have to believe you'll actually get my USD before you'll even set a price for your crypto-USD.

To me the central issue is trust. If you solve that problem the market will adjust to the relative value of USD in different forms and to changes in demand.

Such a thing already exists today, it is called Bitinstant.  They charge a fee for 'instant funds' in Mt. Gox and the reason they can charge that fee is price difference.

Actually, you can set a BitShare price for crypto-USD without having to trade with anyone but yourself.  It would lock in the 'exchange rate' and if you were the ONE AND ONLY issuer and held all crypto-USD then your dividend payment would be EQUAL to holding just BitShares and thus there is NO RISK in placing bids without having $USD in hand because the value you receive is proportional to the DIVIDENDS received which is the same both immediately before and immediately after your bid is accepted.   Now if the price changes while you hold-crypto-UDS then you could lose money due to exchange rate changes between crypto-usd and bitshares, but that is what markets do.

I am responding to you in another thread because I want to get all discussion on this idea in one place so I do not have to repeat myself...

[phillipsjk]

...
FIAT is toxic to an automated distributed P2P exchange.
...

I am not sure point 3 is possible without virtual tokens and the resulting central point of failure. If you want to wait for 1 confirmation while trading with bitcoin: the wait will be anything from seconds to over and hour; with the expected wait time being 10 minutes.

I have a 7th criterion to add:
7. It must not be assumed that the individual users have access to less information than the network as a whole.

As mentioned, the are numerous P2P exchange threads popping up. The ones that sound the most plausible on the surface are also the most flawed: That the network as a whole can somehow authorize things that individuals can't. This violates the laws of the universe, cause and effect, The second law of thermodynamics, if wishes were horseshoes, or whatever you want to call the DRM problem.

I am embarrassed to say I was convinced by the CV token thread for just over 24 hours.

[bytemaster]

...
FIAT is toxic to an automated distributed P2P exchange.
...

I am not sure point 3 is possible without virtual tokens and the resulting central point of failure. If you want to wait for 1 confirmation while trading with bitcoin: the wait will be anything from seconds to over and hour; with the expected wait time being 10 minutes.

I have a 7th criterion to add:
7. It must not be assumed that the individual users have access to less information than the network as a whole.

As mentioned, the are numerous P2P exchange threads popping up. The ones that sound the most plausible on the surface are also the most flawed: That the network as a whole can somehow authorize things that individuals can't. This violates the laws of the universe, cause and effect, The second law of thermodynamics, if wishes were horseshoes, or whatever you want to call the DRM problem.

I am embarrassed to say I was convinced by the CV token thread for just over 24 hours.

I think that 7th criterion is entirely reasonable.  I do not think my approach violates this criterion, but would be interested in your opinion.   

[Chaoskampf]

Is it just me, or is this a good idea?


https://bitcointalk.org/index.php?topic=212490.0;all

[eMansipater]

...
FIAT is toxic to an automated distributed P2P exchange.
...

I am not sure point 3 is possible without virtual tokens and the resulting central point of failure. If you want to wait for 1 confirmation while trading with bitcoin: the wait will be anything from seconds to over and hour; with the expected wait time being 10 minutes.

I have a 7th criterion to add:
7. It must not be assumed that the individual users have access to less information than the network as a whole.

As mentioned, the are numerous P2P exchange threads popping up. The ones that sound the most plausible on the surface are also the most flawed: That the network as a whole can somehow authorize things that individuals can't. This violates the laws of the universe, cause and effect, The second law of thermodynamics, if wishes were horseshoes, or whatever you want to call the DRM problem.

I am embarrassed to say I was convinced by the CV token thread for just over 24 hours.

I'm someone still convinced by the CV tokens thing.  What made you change your mind?  I would like to know about it.

[phillipsjk]

I'm someone still convinced by the CV tokens thing.  What made you change your mind?  I would like to know about it.

The way I understand Bitcoin, value is transferred around by crytographically signing a transaction to prove that you have the private key. This requirement should prevent the CV Token network from "collapsing back into bitcoin" when the backers fail to act.

For CV tokens to work, they must do one of two things:

• Store private keys in the network. This is bad because a bad actor can simply use those private keys to send coins to themself.
• Somehow transfer value in and out of the bitcoin network without using private keys. This is bad because it breaks (at least my) fundamental assumptions about how the protocol works.

I admit, I am not as familiar with the gory details of the Bitcoin protocol as I would like. It is possible that there is something that I am missing. So far Croesus has refused to give concrete details, citing complexity.

I think that 7th criterion is entirely reasonable.  I do not think my approach violates this criterion, but would be interested in your opinion.  

I have not had time to review version 2.0 of your proposal. I don't think it violates that criterion.

What concerns me about your proposal is the lack of fungibility. If different DS coins are worth varying amounts because each transaction has its own exchange rate: price discovery is going to be difficult when converting to another crypto-currency like Bitcoin.

[eMansipater]

I'm someone still convinced by the CV tokens thing.  What made you change your mind?  I would like to know about it.

The way I understand Bitcoin, value is transferred around by crytographically signing a transaction to prove that you have the private key. This requirement should prevent the CV Token network from "collapsing back into bitcoin" when the backers fail to act.

For CV tokens to work, they must do one of two things:

• Store private keys in the network. This is bad because a bad actor can simply use those private keys to send coins to themself.
• Somehow transfer value in and out of the bitcoin network without using private keys. This is bad because it breaks (at least my) fundamental assumptions about how the protocol works.

I admit, I am not as familiar with the gory details of the Bitcoin protocol as I would like. It is possible that there is something that I am missing. So far Croesus has refused to give concrete details, citing complexity.

There are definitely more ways than that to transfer value around in Bitcoin, though they aren't as well known.  You can use scripts to build complex conditions which must be satisfied, and then anyone who can satisfy them can spend the coins.  Just off the top of my head:

-you can build transactions that anyone can claim (has actually been used)
-you can build transactions claimable by some subset of a group of keys ("m of n" signing, actually in use)
-you can build transactions which can be claimed by anyone who knows a secret password or piece of data (could be a file, a word, or an answer to a math problem, not just a private key corresponding to a public key or its hash, and has also actually been used)
-you can build transactions which go to the first person performing some amount of computational work, similar to how mining works (but separate from it)
-you can build transactions that no one knows the output of until they are spent (pay to script hash, where the transaction only has a hash of the script and the person collecting the transaction has to provide a script that hashes to that hash when spending it.  I believe this is already in use, but I'm not sure)
-you could build a transaction that could only be claimed by the owner of a certain pgp key (might be a big script, don't know how efficiently you could implement RSA signature verification in transaction scripting)
-you can build a transaction that isn't valid until after a certain time
-you can build a transaction which only becomes valid when someone else adds more money to it
-you can build a transaction with a sequence number which can later be replaced with a later sequence number
-you can build transactions where a portion of it can be modified later without requiring a new signature as long as the part that was signed remains unchanged
-you can build transactions that can be updated repeatedly without publishing them to the blockchain, and then only publish the final copy when all is said and done
and lots more I'm sure I'm forgetting.  Satoshi built a ton of cool features into Bitcoin--we just have most of them turned off right now because there aren't any commonly used use cases so it isn't worth the risk that the feature has been improperly implemented in one or more clients/miners and might result in a security exploit.

croesus has been light on details, but what she/he describes seems technically possible to me.  Just take a look at the contracts page on the wiki and you will find a whole slew of things that can be done with transaction scripting.