‘DLT and Physics’: consensus, time, and data structure
This article is a topic I have been keeping in mind for many years, even dating back to 2017 when I just entered the crypto world. Before 2020, I used to be impressed by the article “In Math We Trust” (Zhang Shousheng). After 2021, I was completely daunted by IOTA’s core researcher Hans Moog’s series of articles (especially The Trust Machine Part III, which I highly recommend to read a few times). It’s hard to find anyone in the cryptocurrency community who cites relativity (synchronization problem) or quantum mechanics (measurement problem) in an article, especially if their understanding of physics is profound and correct.
(Readers are welcome to send me good articles on first principles or suggest which researcher I should subscribe to.)
Anyway, I wanted to write down my thoughts this time because I posted a series of tweets back and forth with Hans on April 19, discussing DLT consensus and physics related issues. To put it this way, many people may think that physics is a basic science and information science is an applied science, but this is not the case. Information science is “basic science” (pure programming is not information science research). If information (bits) are considered as fundamental as atoms/energy, it should be no surprise that the dialogue between the two fields can go very deep.
Blockchain and global time
Talking about time, one of the most enigmatic phenomena in the physical world, reminds me of the classic physics joke (probably from Feynman or Wheeler) that “time exists to avoid simultaneous events.” The fundamental law of the universe is not relativity or quantum mechanics, but the law of contradiction. The easiest way to avoid inconsistencies (the equivalent of a 51% attack in the blockchain world) is to sort event by time: as long as it happens with an order, you can avoid inconsistencies.
On the other hand, the core of decentralized ledger technology (DLT) is not only about digital scarcity, but more importantly how decentralized nodes can reach consensus . In the world of blockchain, time is defined by the block rather than the objective time of the outside world. Events packed into chunks by miners push time forward, so along the lines of the Nakamoto Consensus, the longest chain wins by voting on time. The nodes choosing the longest version from a variety of possible reality can be said to be efficient (or brutally straightforward).
The well-known impossibility trilemma of blockchain can be said to point out that the consensus mechanism will face the dilemma of decentralization and efficiency under the premise of maintaining security. However, this dilemma is not necessarily a trade-off (there is no impossible theorem similar to CAP theorem), and there are opportunities for architectural innovation to break through. The trade-off between decentralization and efficiency can be illustrated by BNB Chain (previously called Binance Smart Chain). Fewer nodes and higher hardware requirement can indeed have higher TPS under EVM architecture. On the other hand, architectural innovations exemplified by Solana’s history demonstrate that POH: It is Solana’s design to allow each verification node to have an objective clock of local calibration (i.e. using hash function as VDF), so that the verification node can achieve global synchronization (and then vote) in a shorter time when transmitting messages. It is Solana’s philosophy to achieve scaling through speed and engineering optimization. This kind of philosophy makes me think about the objective time in Newtonian universe. (As for how Solana became the downtime chain, it is because the block order can be calculated in advance, and at the same time, there is no fee market. So it was vulnerable to DDOS.)
In a nutshell, if the blockchain nodes reach consensus by agreeing on the causal structure of global events and voting for the longest chain version, there is inevitably an architectural bottleneck: users need to wait for the verifier nodes to be packaged and spend time waiting for the nodes to confirm consensus. So if decentralized accounts can reach consensus “without global ordering”, it is a ground breakthrough paradigm shift. Physical world might give us some inspiration: for example, the theory of relativity is not containing the global time as sorting mechanism to achieve no contradiction. We can fully understand the two events are objective in time, but there is no order structure:the disappearance of simultaneously time let two distant events in the universe can happen at the same time, A earlier than B, (or B earlier than A) based on different observers, but it will not create conflict.
DAG and string theory
String theory provides an interesting insight into how fuzzy/nonglobal order of events can achieve a causal structure without contradiction through superior data structures (the above relativity deals with the simultaneity of two events, while the lower string theory deals with the collision of two particles). In the contemporary world view of particle physics, two point particle collision events (above left) for any observer is a point in time and space, the corresponding results are produce infinite gravitational interactions can not use elegant mathematical methods to eliminate (the method to deal with electromagnetic field didn’t work for gravity); In contrast, in the string theory worldview, the collision of two 1D strings does not occur at any unique point in space-time, and the different slicing angles of this 2D tubular structure represent different views of the same event in space-time by different observers. This unique slice of space-time solves the infinite gravitational energy conundrum of particle physics (it is only in theory, of course, the whole string theory still has no experimental evidence). This part of the description is basically quoted from the classic popular science work “the elegant universe” chapter 6.
In the field of DLT, the data structure of asynchronous ledger without global order is mostly in the form of DAG (directed acyclic graph). Since nodes do not need to sort transactions and wait for verification, the ledger can be updated in parallel. DAG projects this can achieve higher TPS, and they don’t require any fees (each user is a verifier, which complicates the conflicts of interest between consumer and verifier in the blockchain architecture). Therefore, it can be seen that “the impossible trilemma of blockchain is a special case of the impossible quadrulemma of DLT” : it can not have order, security, decentralization and efficiency at the same time. Since DAG (such as IOTA Tangle) has no sorting ability (as can be seen from mesh structure in the figure above), the remaining three good properties can be obtained at the same time. However, we have only talked about the data structure of DAG(IOTA) so far, and have not discussed how to reach consensus among nodes. After all, in the chain architecture, the ledger grows linearly and in the DAG architecture, the network grows in parallel. The longest chain wins method obviously cannot be applied directly. While IOTA’s mainnet 1.0 white paper in 2017 had Tangle(DAG) data structure, but the consensus mechanism originally envisaged could not be used at all (tragic).
PS: The limitation of how to scale consensus is the reason why I did not mention NANO when discussing DAG. IOTA(Tangle) and NANO (Block Lattice) are BOTH DAG in terms of data structure
IOTA and TIQM: A non-temporal consensus
Double spending events are more likely to occur in the network transaction structure of DAG parallel transmission than in the chain structure, so IOTA needs to establish a more efficient consensus mechanism when absolute time does not exist (also because absolute time does not exist, node voting cannot be conducted after the conflict occurs). The multiverse consensus mechanism proposed by Hans deeply solves this problem, in which consensus can be reached “atemporal” : To put it simply, in addition to the three information of sender, sender and UTXO, each IOTA node added version control when sending a transaction, so that the node had already voted when sending a transaction to Tangle (each node has its own reality). If there is a conflict occurs, the node is simply recount the vote without actual vote again, the check-in process is “atemporal”(Because consensus is not established at the time of the vote, or at the recount after a clash of branches.), the node can, according to public information, reproduce consensus without the need to establish causal relationship between branches.
I initially thought the multiverse consensus mechanism was a reference to Git, but Hans reminded me that TIQM (trader’s interpretation) in the quantum mechanical interpretation resonates more with IOTA’s “non-temporal” architecture. TIQM core point is the evolution of any event can produce normal bid wave moving forward in time, as well as the reverse time evolution confirm wave, events (that is, people often say that quantum mechanics to observe, superposition state collapse) is indeed a interference results in time and space between these two kind of waves, but we can’t point out that it is in which time is determined. It could be argued that the physical world does not tolerate conflict, but consensus may not take time. If you think of IOTA’s multi-universe consensus mechanism in terms of TIQM rather than GIT, then the nodes sending transactions are sending bid waves of normal evolution into the future, and the consensus of any node trying to unscramble the entire ledger is a confirmation wave of cutting the ledger branch forward in reverse time. So a conflict of possibilities (such as A double spending transaction) is like the collapse of the wave function observed in quantum mechanics. An observation that does have an objective outcome (each node agrees on whether A is included in the book or B), but perhaps that objective outcome is determined by a time-free/atemporal consensus.
Note: Different interpretations of quantum mechanics give roughly the same physical predictions, but we can get insights from them.
Summary: The ultimate architecture of consensus
As someone with an academic background in physics, I am always excited by the intellectual excitement of using first principles to break through the human understanding of time and space. IOTA’s paradigm gets rid of time through data structure (fuzzy absolute time) and consensus mechanism may become the ultimate form of the whole Crypto/DLT field. This realization was the second such shock moment in my life. The first time was when I was in high school reading about ‘the elegant universe’ and being exposed to string theory. Unfortunately, the elegant theory of high-energy physics may not show any sign of being tested experimentally for the next hundred to thousand years. In contrast, it took only less than 10 years (even from 2015) for IOTA to launch all of its grand designs. 10 years might be an eternity in cryptocurrency time, but that is just a flash in the sky to explore one of the final frontier of information science.
