Topic: The DEX (Decentralized Exchange) Thread - page 10. (Read 24439 times)

There is little reason to buy an asset from an anonymous counterparty (stocks or bonds require trust in the issuer). Cash and assets are entirely different things. Capitalism requires more than cash. Credit and reputation are strongly related. Almost every action in the economy relies on some previous record of an individual or organisation and involves more than 2 parties. Although Bitcoin minimizes trust in a third party, still that's doesn't change that 2 counterparties interact in some way based on a range of assumptions. If Alice sends Bob 1 BTC she expects something in return. Escrow solves that only in a limited way (although that concept could probably generalised to work in many more contexts).

But in terms of pure DEX (and ignoring the very hard problem of consensus on assets): instead of just talking about HFT, its more fruitful IMO to think about price determination and total order of events in general. Blockchains implement only partial order, so by default auction systems on top of them are not really going to work in a meaningful way. A solution would be desirable for determining prices of transactions themselves (fee market).

That's what I said up thread, its not possible to get that fast enough in a decentralized system.


That's not entirely true IMO.

If the frequency of submitted trades are low, then the likelihood of a conflicting trade being presented at the same time is also low, so the millisecond settle time is not as important with regard to conflicts.

Even if you are referring to the issuance of high frequency trades that want to take advantage of the minute price movements that happen over the course of a second, then without the absolute volume of trades overall these tiny differences in price over the course of a second won't occur either, as there aren't enough traders making multiple trades per second on that resource anyway to effect the price enough for anyone to take advantage of.  Thus the millisecond settle time again is mitigated.

If the frequency of trades are low, then any trades you make in the above manner are just trading on the last buy/sell, which is likely yourself.

I don't suppose it matters much to argue about it though, as its not possible in a decentralized, or even semi-decentralized system to get anywhere near these millisecond settlement times.

High frequency trading requires very, very fast response times; that's the key part. It doesn't necessarily follow that trades are submitted at a high frequency in HFT.

OK. Distributed systems tend to be slower to reach consensus.

Confusion of context, I was referring to a high volume of transactions, not value, which I guess is the same as high frequency.

IMO, this is a very bad idea. When identity generation is trivial, there is no recourse for the long con attack; attackers create an identity, build the required trust and then pull off a huge scam and repeat the process until the system as a whole is worthless.

Here is another project

Lykke

website: http://www.lykkex.com
whitepaper: https://wiki.lykkex.com/colored_coin_exchange_white_paper

Lykke is a more of a mix of decentral and central exchange targeting a large scale market and also institutions. The authors have significant expertise in the traditional financial markets and I think that is a valuable perspective. As outlined its not the same as the idea of decentral exchange in the Cryptocurrency sense.

It is good to begin with the word exchange and think backwards from there. To physically exchange goods you need to be in the same location. To transport goods around the world one needs permission from those who control the flow of goods. Then one has the problem that one good might be legal in jurisdiction A and illegal in jurisdiction B. Part of a marketplace is also the definition and registration on what goods should be traded (markets where the cost of creating a digital asset is zero removes friction, but also leads to spam).

If both sides of the trade are not in the same jurisdiction its not an exchange in any legal sense - no legal protection for the buyer and seller. There is no purchase contract underlying the transaction, and no legal recourse. Note that the way the economic system currently functions is that we have large institutions which determine the outcomes of these processes and provide the market infrastructure. If you for instance want to buy a stock certificate in a developed country you have many different institutions involved (stock exchange, market-makers, brokers, settlement). The simpler, but still non-trivial case is a warehouse receipt. If two parties exchange 1 bar of gold for instance usually one has some expert acknowledging the quality of the metal. There are only 20 or so commodities traded on the large markets. To exchange a good or service is far more complicated than it seems. If one is exchanging physical goods for instance there is the question of quality and location. Undeveloped countries often lack financial infrastructure and need to build this on their own. In the case where such countries build up a new exchange market one can see what is involved with developing a marketplace. And this doesn't touch even on the complexity of equity and bond markets. A lot more research could be done in this area to show what things are possible and how Crypto-economics could lead to new types of markets.

high volume and high frequency are two things. You could attract users to make infrequent large trade and get high daily volume. B&C allows user to trade between their own wallets and can potentially attract high value traders to trade often.

Load handling isn't the same as settle or confirmation time, which is a completely different kettle of fish

It comes down to good old CAP and consensus.   Assume there is a trade T with an ask of 100.  User A submits a buy with a price of 100, User B also submits a buy with a price of 100 at exactly the same time but at the opposite end of the network.  Which is the correct buy to apply?  Some nodes will see A's buy first, apply it, and B's will be invalid, others will see B's first, apply it, and A's will be invalid.  Then you end up in a conflict, half think A was first and is the correct buy to match that ask, the other half think B.

So while the initial acknowledgement can be very fast (sub 1 second), the duration to come to consensus is much longer and in some circumstances the original acknowledgement that A and B see may not be the one the network ultimately agrees on.

In the case of Bitshares, it has what I consider to be semi-centralization in the form of DPOS, which makes this task much easier to do quickly.  Not only that, I doubt that Bitshares is seeing trading loads high enough yet to be triggering that scenario in the first place.

Bitshares also has very fast block times IIRC, and coupled with DPOS, they mitigate a lot of the problems that you have to deal with when doing the same in a fully decentralized system.

Is the Bitshares approach the right or wrong way?  That depends what you want from the platform, if it does what it says on the tin and both they and the users are happy with it.  If everyone is happy, then sure, why the hell not

Also I agree that HFT DEXs will probably never be required, but I spent a little time pushing the limits just so I knew where they were

That thread has way too many sock puppets and shills for me to take it seriously. I will look into it though.. I remember seeing this thread but I never looked into elCoin.

Don't take this the wrong way, as I like you and your eMunie project, but I don't get this statement. You claim eMunie can process way more TPS than Bitshares, yet you can't get trades to settle on your DEX in less than 10 seconds? That doesn't make sense to me. Bitshares settles trades in less than 4 seconds reliably. In fact, 4 seconds is the maximum amount of time and it usually takes around 2 seconds, but it does depend on when you click the confirmation but in relation to when the last block was produced. I suggest trying Bitshares' DEX. I think you will be impressed with the speed of settlement.

Also, I don't think anyone expects DEXs (at least as the technology is today) to realistically compete with the NYSE or other large HFT exchanges. I only expect them to compete with cryptocurrency exchanges in the near to immediate future. You can however trade stocks and commodities on Bitshares which is a benefit compared to most cryptocurrency exchanges.

I can agree that users don't seem to care about decentralization... yet. I think that centralized exchange users will keep getting "goxxed", and eventually everyone will understand why DEXs are so important. It is a long term play that requires the inevitable to continue happening (exchanges going bankrupt). Let me comment on your list of most important things for an exchange:

I can only use Bitshares as a measuring stick, but it is very convenient compared to most exchanges. Simply entering a username and password is much easier than most exchanges where you at the very least need to confirm your email address.

The fees can be adjusted, and I think soon you will see a competitive fee structure on the Bitshares DEX. Low percentage-based fees are inevitable because the community mostly agrees that is the only way to proceed.

As far as speed, Bitshares settles trades in less than 4 seconds, which sounds slow on paper but is actually pretty fast when you actually use it. It is hardly an inconvenience and is not that much slower than centralized exchanges. Maybe high volume day traders would not like the lag, but for casual traders or "buy and hold" types the lag is not a problem. That is my opinion of course, but I am in the latter group and the lag doesn't bother me a bit- 4 seconds or less goes by much faster than you'd think... 4 seconds is the maximum amount of time if you happen to submit the trade right after a block was produced, so in most cases it will be more like 2 seconds.

Security is great with DEXs.. that is the point. There is some systematic risks involved with decentralized derivatives, but I see them as being much more unlikely to collapse than a centralized exchange going belly up.

There are better ways to do this, however I doubt a fully anonymous exchange is even desirable. In the end its about how one thinks about law, risk and identity. The failures of MtGox and Crytpsy are a bit over-rated since other exchanges have been operating for years. I think its in part that Bitcoin is just not good for holding deposits - its not that deposits are always undesirable. It should be possible to trust others if one chooses to. One can create reputation and credit systems on top of trustless systems. From a regulatory point of view consensus on assets is 100x harder then consensus on money. Imagine a fully P2P world-wide exchange. To get proper clearance for assets will be almost impossible for a long time. Otherwise one is operating in a legal greyzone. Anonymous digital cash and anonymous exchange are very different things. So one has to think about very clearly what kind of activity what wants to facilitate. It is impossible to touch fiat without KYC/AML, but that is not even desirable. To transmit fiat money on electronic networks one needs a banking license. Fiat money ultimately settles in the Internetbank market. A completely decentral exchange, IMO makes most sense for swaps between Altcoins and commodities.

Correct, users don't care.

Plus, trade speed will be the critical problem that will prevent any DEX from seeing high volume, especially in the future.

I've spent quite a lot of time looking at ways to achieve high frequency trading (HFT) that is both secure and reliable (trades don't go missing, no "queue jumping"), and I don't think its possible to get < 5 second easily without some form of semi-centralization.  When you compare traditional HFT exchanges such as the NYSE and others which are able to process trades sub-millisecond, well, our decentralized counterparts are never going to be able to compete as a serious alternative.

The fastest I've been able to get trades in our DEX with sufficient reliability is about 10 seconds, but that comes with quite a bit of overhead across the board in terms of bandwidth, processing and other costs.

The real problem with the idea of a decentralised exchange is not the technology, its the users.

Users do not care about decentralisation at all. The only thing they care about is convenience, fees, speed and security. DEX has great security, but fees will always be higher and speed will always be slower.

funny nobody seems to want to mention the ELEPHANT in the room:

https://bitcointalksearch.org/topic/elephant-platform-announces-ethereum-based-cryptocurrency-elcoin-1309597

Looks like a scam that could give Ethereum a bad name.

Agree, this is currently the best decentralized exchange solution in development. There is also a project going on for Ethereum platform.
I personally think that centralized exchanges do not have a future (look at MtGox, Cryptsy,...).

i don't think jl777 has been dumping anything. he put up some sell walls to prove a point, but he later pulled them. afaik nxt is still utilized in the project. as i mentioned earlier in this thread supernet and nxt are two seperate entities. all this happened a few month ago with the update to 1.6.2 on October 31 2015. things have cooled down now and both projects continue to be developed. nothing to get hung up on now imo.

Supernet still being developed on top of NXT?  We gathered that the lead dev is dropping NXT due to some API issues in one of the upgrades.  And that the Supernet lead dev was the one responsible for dumping all that NXT to its lows.