Blog > 2017 > November

Ethereum Classic community comes together at successful Hong Kong summit

Principles and plans for growth unite participants at event organised by Grayscale

22 November 2017 Jeremy Wood 10 mins read

Ethereum Classic community comes togeth

Since day one, Ethereum Classic has been driven by principles. What began as a disparate group of people who came together after the DAO hack and the consequent hard fork on Ethereum has now become a strong and growing community, united by a belief in immutability and that code is law. So it was tremendously exciting to see everyone gather at the first Ethereum Classic Summit held last week and watch the community move beyond the realm of instant messaging to interacting in person and making new friends. There were many highlights of the two day summit in Hong Kong. It was a moment to reflect on how much Ethereum Classic has achieved in a little over a year, developing its identity, building its technology and securing its future. Grayscale, which provides Bitcoin and Ethereum Classic investment trusts, has provided much momentum by organising this event.

Barry Silbert, founder, explained that Grayscale will put a third of its ETC fees for three years into supporting the growth of ETC though development, marketing and the community, and the ETC Cooperative has been established to do this. Barry told IOHK: “We wanted to do our part and help support increased communication, collaboration and bring together all the various stakeholders at least once a year. I’ve been super excited to see the participation, the level of engagement, the quality of the speakers and participants and think this year has been a huge success.”

Charles Hoskinson giving his keynote speech

What might come a surprise to some is that Ethereum Classic is not a direct competitor to Ethereum. In his keynote speech on the first day of the summit, Charles Hoskinson, founder and CEO of IOHK, and former CEO of Ethereum, said: “Frankly, we compete far more with Bitcoin. To me, Ethereum Classic is basically a digital commodity, it’s digital gold. It’s really exciting because it can give us a very interesting design space that we can explore that’s totally independent of the design space that Ethereum is pursuing. We don’t have to embrace proof of stake, Plasma or the fork of the week. Instead we can start with principles and we can grow from principles.”

Charles argued that two separate philosophies had always sat uncomfortably together in Ethereum, even before the community split. “We actually had two Ethereums from the beginning but we didn’t know that. We had two philosophies at the same time. One was this notion that code is law, that code can’t care, it doesn’t matter what you’re running and it doesn’t matter about the consequences. But there was this other philosophy, the notion of a world computer, the idea that you could have universal infrastructure that would be a beautiful computation layer that could be added to the internet. But the fundamental problem is that these things are philosophically so different they can’t coexist. When you’re the world computer you have to fork, you have to embrace mutability and have a lot of flexibility in your platform.”

Now more than a year after the split, Ethereum Classic has developed its own identity. It had a great asset in its principled and warm spirited community, Charles pointed out and championed the development work being done. “We see a lot of great things coming down the pipeline. It’s got to a point where we can prove beyond reasonable doubt that we can keep up with Ethereum Foundation and in some cases we’re writing better software. As a comparison, Bitcoin Core has about 110,000 to 120,000 lines of C++ code. The Mantis client does more and it’s 10,000 lines of code. It’s twelve times more concise. “The Sputnik VM, pulling the VM out and having it as a standalone, is a phenomenal idea, it’s a great piece of engineering. It’s far better than anything coming out of the Ethereum Foundation.”

A highlight of the conference was hearing about the steady pace of development of the ETC protocol. The eight-strong team of engineers that make up ETCDEV team are working on implementing the monetary policy, a flexible software developer kit that can be used to build applications on Ethereum Classic, and making performance and reliability improvements to the Ethereum Virtual Machine. Isaac Ardis, Go developer from ETC Dev, outlined the direction and purpose of the Emerald Project, which is not just a wallet but a library and suite of tools for third party developers. Licensing projects under Apache 2.0 makes them commercial-friendly, and easy to implement.

Igor Artamanov, CTO of ETCDEV, told IOHK that he believed Ethereum Classic would rise to be in the top three cryptocurrencies. “We’ll continue our work on open tools for developers who will build apps on top of ETC blockchain. Our goal is to set up development processes and help with building various tools for engineers and open source libraries around ETC.” A theme of the conference was formal verification, and on that topic Igor said: “It will be cool to have. We are not experts in formal verification, but if someone will decide to introduce it to SputnikVM, we’ll appreciate that.” On monetary policy, Matt Mazur, advisor to the ETC Dev team, gave a clearly explained presentation on ECIP 1017, the proposal that contains changes such as capping the amount of ETC that can be created, in contrast to Ethereum’s uncapped supply.

Alan Verbner speaking at ETC Summit
Alan Verbner speaking at ETC Summit
The [ETC monetary policy](/blog/ethereum-classic/a-joint-statement-on-ethereum-classics-monetary-policy/ "ETC monetary policy,") is modelled on Bitcoin’s monetary policy, and is designed to reduce risk, create simplicity, and encourage investors. Every 5 million blocks, there is a 20 per cent reduction in block rewards. The question of how Ethereum Classic will scale has always been an issue of debate. Igor shared insights on how sidechains could be deployed, making use of public and private sidechains and having the main chain for security. Cody Burns, ETCDEV adviser, talked about cross-chain interoperability and how [cross-chain atomic swaps]( "Cody Burns, Medium") could allow people to exchange coins in a trustless manner.

There was news from the Grothendieck team, the IOHK developers that are dedicated to ETC and who have over the past year built the Mantis client for Ethereum Classic. Alan McSherry, team lead, gave details about integration with the Daedalus wallet. It will be fantastic for the community to have another option for a wallet, where they can securely store their ETC. “The Daedalus release is going to happen hopefully in December. That’s taking the Daedalus wallet, the UI that was built for Cardano, and putting it on top of our Mantis client, our ETC node, so you’re able to use a familiar UI with Ethereum Classic.”

Alan McSherry also offered a look at what the team are thinking about in terms of Mantis development in 2018. With regard to K Framework, the important research to create formal semantics of the EVM done by Professor Grigore Rosu at the University of Illinois, and funded by IOHK, he said: “With any luck, that will result in the possibility of writing Scala contracts that can be compiled to run on the EVM, or a slightly different version of the EVM, and we’ll have the formal verification and it’ll be a question of trying to create formal proofs around those contracts.” IOHK research into sidechains could allow the possibility of creating contracts to allow interactions between ETC and Ethereum and ETC and Bitcoin, while other features on the horizon for ETC include zero knowledge proofs and anonymous addresses, he said.

And Alan Verbner, developer on the team, gave a reminder of why the team use Scala. “Functional programming allows you to create less and more secure code and that’s why we are trying to build in Mantis compact code that is safe and you can check its correctness. Our client has 10,000 lines of code, it’s easier to read and easier to access. We are building open source software so we like people to look into it and analyse it to look for bugs and security issues. You have fewer lines of code to test.”

The theme of principles arose again when it came to hearing from Ethereum Classic miners. We heard how mining pools chose ETC because of the core principles of the community, proving the attractiveness of the code is law philosophy. There was good news from the 91Pool, the first group that ever mined on ETC, when Meicy Mei announced the group has purchased a building in Shanghai to serve as a dedicated hub for the ETC community in China, with a cafe and bar that will host meet ups.

ETC Summit key stakeholders on stage
ETC Summit key stakeholders on stage
A final reminder of the principles that brought the Ethereum Classic together came during an impassioned keynote speech from Meredith Patterson, an expert in programming language security. She recounted the scorn that Ethereum had responded with when formal methods of software development were suggested as a good idea. She argued that Ethereum Classic still had the chance to adopt formal methods – an approach that IOHK is a strong advocate of and actively pursues during its development process. For code to be law, she said, the meaning of the code must be unambiguous, and users must have the same understanding of it that machines do. Steps to get there include adopting a rigorous formal semantics, like KEVM. Deficits in test suites should be found and the problem corrected. Solidity should be replaced.

The importance of coming together as a community to debate these issues cannot be underestimated. The results of these discussions will drive the development of Ethereum Classic. Sharing this view, Barry told IOHK: “I think in the digital currency space there’s a little too much interaction done through social media and through anonymous forums and it’s important to get people together to meet each other, challenge each other, and do it in a way that is respectful and that fosters communication and collaboration. So I do think it’s important to have summits like this. I absolutely expect that we’ll do it again next year and hopefully the year after that will become an annual event that everybody in the community looks forward to.”

Writing a High-Assurance Blockchain Implementation

Guest blog from Edsko de Vries, who is working on the high assurance implementation of the Ouroboros blockchain protocol

3 November 2017 Edsko de Vries 9 mins read

Writing a High-Assurance Blockchain Implementation - Input Output

Writing a High-Assurance Blockchain Implementation

In our previous blog post Cryptocurrencies need a safeguard to prevent another DAO disaster we discussed the need for high assurance development of cryptocurrencies and their underlying blockchain protocols. We also sketched one way in which one might go about this, but we did not give much detail. In this follow-up blog post we delve into computer science theory a bit more, and in particular we will take a closer look at process algebras. We will have no choice but omit a lot of detail still, but hopefully we can develop some intuition and provide a starting point for reading more.


When we want to study computer programs as mathematical objects that we can manipulate and analyze, we need to write them in a programming language for which we have a well understood mathematical theory. If this mathematical theory inherits much of what we know from elementary mathematics, that's an additional benefit. For example, we learn in school that:

for every number n, n + n = 2 * n

Shockingly, in many programming languages even this most basic of equations does not hold. For example, in the language C, when we define

int f() {
 return 5;
int g() {
 int x;
 scanf("%d", &x);
 return x;

we have that while f() + f() is the same as 2 f() (both equate to 10), g() + g() is certainly not the same as 2 g(): the former asks the user for two numbers and adds them together, the latter asks the user for a single number and multiplies it by two. In the language Haskell this distinction between an honest-to-goodness number, and a program that returns a number is explicit (this is what we mean when we say that Haskell is a pure language), and for numbers basic mathematical equations such as the above hold true. Knowing that for every number n, n + n = 2 n in Haskell would still not be particularly useful without another property of Haskell: when we see the expression n + n anywhere in a program, we can (almost) always replace it with 2 n and know that our program still behaves in the same way: we say that Haskell is referentially transparent. This is key: it means that we can analyze bits of our program at a time and know that our conclusions are also correct in the wider context of the whole program. In other words, it means that we can do compositional reasoning: we can understand the larger program by understanding its components individually. This is essential when dealing with anything but toy examples.

Process Algebra

Of course, even in Haskell we can read from files, write to network sockets, etc. and while the type system makes sure that such programs are distinguished from pure values, reasoning about such programs is nonetheless more difficult because we do not have a good mathematical theory that covers all of these "effects". It is therefore useful to identify a subset of effects for which we do have a mathematical theory. Different kinds of applications need different subsets of effects. For some applications a process algebra is a good match. A process algebra is a programming language in which we can express systems of concurrent programs that communicate with each other, along with a mathematical theory. There are many different kinds of process algebras, and they differ in the details. For example, some provide synchronous point to point communication, others asynchronous; a few provide broadcast communication. So what would be an analogue to a law like n + n = 2 * n in such a language? Well, there are multiple answers, but one answer is bisimilarity, which intuitively allows us to say that "two processes behave the same". For example, consider the two following processes

p1 m = forever $ do
        bcast m
        bcast m
p2 m = forever $ do
        bcast m

Both p1 and p2 consist of an infinite loop; p1 broadcasts message m twice within that loop, and p2 only once. While these two processes may look different, they really aren't: both broadcast message m indefinitely. To make "they aren't really different" more precise, we can try to prove that p1 and p2 are bisimilar. Intuitively, this means that we have to show that any action that p1 can take, p2 can also take, and they are still bisimilar after taking those actions. The proof for p1 and p2 would look something like this, where a red line indicates "can do the same action here":

Process Algebra

So why is this useful? When two processes are bisimilar, you know three things: they can do the same actions, they will continue to be able to do the same actions, and most importantly, this is compositional: if p1 is bisimilar to p2 then we can replace p1 with p2 in (most) contexts. As before, this is essential: it means we can understand and analyze a larger program by understanding and analyzing its components.


Process algebra is a large field in computer science, and many different kinds of algebras have been studied and are being studied. Some of the earliest and perhaps most well-known are the Calculus of Communicating Systems (CCS) and Communicating Sequential Processes (CSP). These days one of the most important process algebras is the pi calculus; argueably, the pi calculus is to concurrent programming what the lambda calculus is to functional programming. While the lambda calculus forms the theoretical basis for functional programming, nobody actually writes any code in the lambda calculus; it would be akin to writing in assembly language. Instead we write programs in a much more sophisticated functional language like Haskell. Similarly, the pi calculus is a "core" calculus, good for foundational research but not so good for actually writing programs. Instead, we will use the psi calculus which is a generalization of the pi calculus with more advanced features, but still with a well understood theory. Let's consider an example of bisimilarity in the psi calculus, similar to the previous example but slightly more interesting. Do you think we should regard processes p3 and p4 as "behaving the same"?

p3 m = forever $ do
        k  <- newKey ; bcast (encrypt k  m)
        k' <- newKey ; bcast (encrypt k' m)

p4 m = forever $ do
        k  <- newKey ; bcast (encrypt k m)

What about process p5 and p3?

p5 m = forever $ do
        k <- newKey
        bcast (encrypt k m)
        bcast (encrypt k m)

It turns out process p3 and p4 are bisimilar, and the proof is almost the same as before

Psi Calculus

They both send out an infinite number of messages, each message encrypted with a new key. Although p3 and p4 will generate different keys, from the outside we cannot tell them apart because we don't know anything about those keys. Process p5 and p4 are however not bisimilar:

Psi Calculus

After p5 sends out a message encrypted with some key k, it then sends that message again encrypted with the same key; p4 cannot follow.

Ouroboros Praos

One of the nice things about the psi calculus is that we can embed it as a domain-specific language (DSL) in Haskell. We use this to define a model of the Ouroboros Praos blockchain protocol, which we can both execute and run (it's just Haskell after all), and even test using QuickCheck (Haskell's randomized testing tool), but moreover is also amenable to formal verification using the psi calculus metatheory.

We start by expressing the algorithm as it is defined in the Ouroboros Praos paper in our Haskell psi calculus embedding. Although we cannot formally verify that this algorithm matches the one in the paper, the correspondence is close enough that it can easily be verified by hand. We can run this implementation and experiment with it. This first implementation will do various things that, while we can implement them, we would not want to do in a real implementation. So we then make a number of small, incremental adjustments, that get us closer and closer to an actual high performance implementation. Each of the individual adjustments should be small enough that a human reader can understand how we go from one implementation to the next and convince themselves that we haven't changed anything fundamental.

Each of these algorithms can be run, so that we can test that all of the adjustments we make to the original algorithm don't change any fundamental properties. Since we express the algorithm in terms of a well-understood process calculus, we can also take some of these adjustments and formally prove that we haven't changed anything fundamental. Bisimulation is but one tool in the theoretical toolbox here; there are many others. The end result is that we have a bridge from the high level specification of the algorithm in the academic literature to the actual low level implementation that we implement, with high assurance that what we actually run indeed corresponds to the algorithm defined and proved correct by the cryptographers.

Download and Contributing

If you want to follow the work in progress, you can find the psi calculus development of the Praos blockchain protocol at Praos Formalization. One way in which you could contribute would be to read the document, compare it to the Praos algorithm in the literature, and convince yourself that indeed they do the same thing; after all, this is a step we cannot formally prove. Although the subsequent adjustments are in principle formally provable, in practice we may only have resources to do that for a few steps; so here too the more people read the document and try to poke holes in the arguments, the better. If you want to do more than verify what we have done, we would be happy to discuss the possibilities; contact us here. We look forward to hearing from you!