Menu

Tau-Chain – a programmer’s perspective

May 31, 2016 - Blockchain
Tau-Chain – a programmer’s perspective

 

Piotr Piasecki is a master of science and a double bachelor in the field of computer science and informatics. He is currently an Advisor for the Crypto Currency Certification Consortium, and Chief Scientist at Provable Inc.

 

Living in the Bitcoin land, you never know what you might come across next. It could be as benign as someone issuing a currency backed by pre-1965 silver US dimes, as geeky as someone creating a blockchain to mine for prime numbers, or it could be as convoluted as BitShares with the many iterations it had over the years (as someone put it – “BitSharesX – An Alt Coin That Is Impossible To Understand“). Over the last few months, I’ve been seeing a lot of spam about Tau-Chain, along with its many extravagant claims, and figured it might be interesting to try to understand it.

Disclaimer – the project appears to be delving really deep into the theoretical computer science that almost borders on philosophy. While I do have a masters degree in computer science, I can’t claim I fully understand some of the topics Tau-Chain touches on or their implications. I will instead focus on more practical aspect of Tau and how it presents itself as a piece of software with practical use.

What is Tau-Chain?

So, what is Tau-Chain? Well, it’s quite simple, just look at this graph from the founder of Tau:
TauChain 1A simple explanation of Tau-Chain…

Okay, it’s not simple at all. This graph represents what sort of confusing things we’re dealing with here…

From what I gathered looking at the project’s website, its whitepaper, roadmap, some articles on it, listening to a LTB interview, viewing some other resources and talking briefly to the founder of Tau-Chain, I think we are dealing with two components here – Tau and Tau-Chain. Unfortunately, it seems the people involved in the project like to use those terms interchangeably and confuse everyone further.

Tau appears to be a new programming language, apparently similar to Idris. Unlike most traditional languages most programmers deal with on a daily basis, it is not turing-complete. Instead, it is a decidable programming language. What this means is that it avoids the halting problem, while still being able to do anything a finite turing machine can do. Since in practice we don’t have infinite turing machines, from what I understand it should be able to do anything a turing-complete language could do. Apparently, this approachmight be more secure. On top of that, Tau “has built-in P2P and blockchain“.

Tau-Chain on the other hand, appears to be a sidechain-enabled blockchain that can run the Tau language. It seems to be similar to Ethereum with its contracts – both have a growing library of code embedded in it that anyone can call upon to build their code on. As I understand however, Ethereum’s code can be more risky to use as you might not always be able to predict what the contract might do without its source code at hand, while Tau the language is more predictable in its execution?

The project also appears to have another component to it – the Agoras. As far as I can tell, they seem to be smart oracles that can execute various contracts and other Tau code. They appear to be able to interact with the Tau-Chain, as well as with one another directly. All in all, they remind me a lot of Codius, especially if you consider that that project aimed to be able to prove what code is being executed and so on. Not a bad feature, but there doesn’t appear to be much new to talk about there.

What Tau-Chain promises

While initially researching Tau-Chain, one will stumble upon their promotional video:

Tau-Chain, solving all of your software development problems apparently…
Which lists a few outlandish claims about what Tau / Tau-Chain can deliver:

As a software developer, I would take all of that with a huge grain of salt. Then again, it might be my turing-complete attitude talking and things might be different in the decidable language space. If this video was talking about traditional software languages, I would put my money on the video being about test-driven development – an approach to software development that starts with test cases (what the code should and should not do), and then developing the code to fulfil those tests. In theory this could mean that the software has no more bugs, it does what it is supposed to and can be verified automatically when new code is checked in. So while it would fulfil most of the listed requirements, in practice I would not expect it to be *the* solution to all problems – writing good test cases can be as hard and time consuming as writing good code, and I doubt 99% of the clients purchasing software would be able to use that. If the test cases aren’t sufficiently complex, we might run into the problem of software being built just to tick the checkboxes and not much else. After all, any program operating on a sufficiently small domain could be replaced by a lookup table

I am also very sceptical of how the software will decide what are the stored facts and how those will be handled and proven. Even more so when we’re talking about “facts” about the real world and social norms. How do you prove you should not put used forks on the table, from a software perspective? How do you handle a problem having multiple contradictory answers (an infinite sum of (1-1+1-1…) can be proven to equal 0, 1, 0.5, -0.5, etc…)?

Some other claims I stumbled upon from other sources (1, 2, 3):

While there are more claims, lets just limit ourselves to those few (a lot more can be found in the LTB interview).

The Tau / Tau-Chain’s feature of embedding how the network operates in the blocks themselves is rather unique feature as far as blockchains go, but at the same time it can be one of the more dangerous thing out there. It certainly offers the network more apparent freedom from Bitcoin-like hardfork stalemate, although in reality Bitcoin’s hardfork problems are never about the code being hard to change, but about the people you need to convince. It might also impair some thin clients if they are applicable to the chain (how can you just run the chain from a given length if you don’t know what the rules are from all of the previous blocks?). The definition of who the “users” in the system are (one-vote-per-person / machine / CPU?), as well as what the rules for hard-forks will be will probably shape the network very drastically early on. I wonder whether anyone will try to change the code of how the blocks are executed to “stop execution, return 0″…

The token presale doesn’t appear to be anything new in the crypto world – it’s the paradox of presales all over again. Tau the language and network doesn’t need a new coin, it would probably operate better without it, but the developers need money to develop the language / network, so they sell tokens to speculators. Looks to me like another Bob Surplus-esque coin looking for a problem.

As for the last claim, and a few similar marketing blurbs, I think they deserve a section all of their own…

False equivalence, false dichotomy, eating your own dog food

The quote about basically being able to replace Google appears to be a false equivalence fallacy. There are many problems with trying to say you can basically be like Google:

All in all, statements like that are just red flags if someone also asks you for money. At best, they are marketing superlatives. So while sure, if we’re talking about Tau the language, someone might use it to implement a Google-like service with it and so on, but the same could be said about computers based on cogs and wheels (after all, any turing machine is equivalent to another). All in all – false equivalence – your software is not even comparable to Google.

Now, lets finish this discussion with a subtle false dichotomy. I stumbled upon this marketing blurb about Tau from some of the spam I see pasted in a few chats I visit:

TauChain 2

It compares how Tau-Chain is different from Ethereum, and links to a blog post by Peter Vessenes criticising how buggy some of the Ethereum smart contracts can be. He makes a lot of valid points – since you can’t upgrade and fix the contract code post-launch, you either need a good failsafe, or write perfect code not to lose people’s money. However, what I take slight annoyance with, is how this sort of marketing might misrepresent the situation – “Tau is different from Ethereum, here are a few reasons why. Here is someone criticising Ethereum (while not talking about Tau)”, implying that since Tau was not criticised and it is presented as Ethereum’s competitor, it somehow doesn’t have those flaws. No Tau, criticism of your competitor does not mean you don’t / won’t have those problems yourself.

Lastly, I find it really amusing that Tau apparently doesn’t like the taste of its own dog food – for all of its criticism of turing-complete languages, saying how Tau is a much better language and all of that, in the endthey develop their code in C++. I did bring this point up to Tau’s creator and he made valid points as to why that is – they want to develop the software in an efficient language to make it operate efficiently and in the future they might implement Tau-Chain in Tau. Understandably, software development takes a lot of resources and time, and you want to release early, release often, but this somehow doesn’t fill me with confidence that Tau will be usable for any commercial-grade software any time soon…

Conclusions

While Tau appears to be an interesting development of a new programming language and its creator certainly sounds very knowledgeable in his field, Tau-Chain looks like a project looking for a problem. Bootstrapping a new token to run a blockchain to use a new programming language for smart contracts that don’t halt seems like a very complicated way of reinventing everything just because you want to change a few things. I am highly sceptical of how the network will handle everything it promises, especially when it comes to dealing with things in the real world. It could be as mundane as a different flavour of Ethereum with a non-turing complete language, some smart oracles, etc., or something potentially new – only time will tell. Until Tau-Chain is released, I remain unconvinced.

Amusingly enough, the Tau-Chain video contains an Escher-like perpetual motion water mill at 1:40. I wonder if this is telling that the project is trying to invent something impossible?

 

 

Sursa: bravenewcoin.com