Another unpopular idea about blockchains, from the comments

1 – My favorite unpopular blockchain ideas: 99% of corporate experiments regarding blockchains are better handled with Apache Kafka and multiple archivers. Anything that attempts to be a fast, global ledger has to accept the reality that global ordering is a limitation, not a feature, and instead use logical clocks. The intersection between blockchain enthusiast and distributed system researchers is close to zero. When we look back 100 years, Bitcoin itself will be seen as far more relevant in retrospect than blockchain technologies.

That is from MR reader Bob.


Yeah, or just a database in some cases.

The point of block chain is decentralization, particularly not needing a central trusted source.

However, often there isn't really an intractable trust problem, just the lack of a standard. Using a blockchain solution doesn't remove the need to create a standard and convince parties to adopt it.

And frequently the trust problem can't be truly solved anyway, as you have to trust someone. See Bitcoin exchanges.

Tickle me there...

Tickle me in my underwear ;)

Clearly the technology is immature and has to separated from the central idea. While blockchain isn't the universal solution some seem to claim, there are applications in the growing sweet spot, where closed blockchains supporting transactions of reasonable volume and complexity will be an important driver of disintermediation. The technology to use blockchain to put my employer out of business will exist in the next ten years. Industry acceptance of the change will take much longer.

For now we're in the gold rush phase. Later, when the hype dies down and it's either a household word in the dull sense or only technicians remember the term, we can use it to get things done where it makes sense.

I see this claim regularly, but no one ever provides a credible application that isn't better done in the traditional way.

Your stuck on the world we have now. Try to imagine the future. In the 1980s the very idea of Facebook made no sense. In the early nineties, indexing the Internet like Google does was thought to be impossible. IBM actually had a patent on a similar idea to Pagerank but it was too slow. Sergey and Larry changed that.

My first company, started in 1996, created a product to manage customer service over email. I used to go meetings with potential investors and half of them either didn't have email or didn't use it much. They didn't see the point of an enterprise class product of email customer service. They preferred to use the phone.

I dont see the original statement as being unpopular. Corporations have very little to gain from the classic blockchain use case, building open access systems that are not controlled by any authority. For corporations though "it's the latest thing" so they believe it will better solve their problems. Corporations thought the same about Big Data a few years ago, and think the same about AI today. Like all technologies, the limitations of reality will come with a bit of experience.
Then, the statement "the intersection between blockchain enthusiast and distributed system researchers is close to zero", could be interpreted as saying that distributed systems researchers are not interested in blockchain technologies. That would not be a correct interpretation. I see quite a bit of interest amoung researchers and in the distributed systems literature. People are trying to innovate on some of the basic ideas. And there is considerable interest in a whole range of associated, interesting scaling, security, and cryptographic issues. The intersection between distributed systems researchers interested in X and all distributed systems researches will alway be close to zero. The field is just that big.
And when we look back 100 years, who knows; bitcoin challenges the monopoly at the source of state power. Theoretically competition should benefit the people who dont own the monopoly.

Big Data and AI, though, are legitimately transformative developments that are already making a big impact, which will continue to grow--even though it's certainly true that rhetoric often outruns reality.

I think the same has yet to be demonstrated for blockchain. Will it actually matter beyond cryptocurrencies (and will they ever be more than a niche)?

Corporate executives see it as the latest thing because their developers told them so. So they let them launch a project to keep them happy. That lets the rank and file developers believe they are working in an intellectually viable place where they can aspire to greatness. The business also need the star developers around because the next next big thing might actually be relevant. Maybe the media attention is useful too.

So maybe it is a good decision.

Legacy companies don't see much use for it because their business model is based on the current paradigm. Innovators dilemma.

"The intersection between blockchain enthusiast and distributed system researchers is close to zero."

Such is life in Trump's America.

Is there an English version of this website?

Ha, thank you! I was just scrolling through the comments to try to work backward from there, to some vague grasp of the post. No luck so far. Back to that remedial "Letter to Jaime Dimon" TC posted a few weeks ago, I guess.

It's humbling to drop by for some light reading with the morning coffee, and sometimes feel like I am witnessing speciation in action. Reading the newspaper never made me feel that way. In fact, the Jumble and Family Circus often left me quite smug and confident.

TC linked to this list of unpopular ideas about blockchain a day or two ago:

Is Bitcoin the tulip of our time. Maybe. But so what? Bitcoin is no different from any other speculative investment; indeed, it may well be the premier speculative investment. Critics say that the Bitcoin bubble inevitably will bust. Defenders say, so what, even if there is a correction, the correction inevitably will be followed by new highs. Defenders have the better case. After all, consider the most recent bubbles, in real estate, stocks, collectibles, busts followed by new highs. I'm always amused by economists and their focus on classical theories of investing, while the real money is made in speculation. The pending tax bills in Congress are a cornucopia of opportunity for speculative investing. A recent article in the NYT suggested that every investor should consider Bitcoin. We are all speculators now. Me? I'm in the tulip market. If the market busts, I can always use the tulips to landscape my yard.

One big differentiating factor between tulip mania and bitcoin is how wealthy the uneducated class is is 2017. The west has never been dumber and wealthier. At least in Holland, only a few percent of people even had any wealth, and they tended to be quite educated.

Bob has some misunderstandings about both kafka and blockchain.

On the Kafka side, you need to have a system (zookeeper, that does all of the coordination of the distributed system. So, while the database itself is distributed, the administration of the system needs to be highly centralized. Basically, one single machine that is assumed to be an honest broker (the zookeeper leader) performs the logic that defines the cluster. Zookeeper allows for simple failures, but not most byzantine failures. This would disqualify it from most of the use cases proposed for bitcoin.

On the blockchain side, it already does use logical ordering in the form of block orders. Clocks never really enter into it. There is an optional timestamp on transactions but miners have no requirement to honor that timestamp in choosing to include a transaction into a block.

He is right however that bitcoin and blockchain are almost certainly not the optimal system for doing a lot of the things. Proof-of-work is a very expensive option to solve a very particular set of constraints: preventing double spend and defining a system that does not allow or require arbitration. If you need exactly those two things, blockchain is basically as good as it gets. If you are willing to relax the arbitration requirement even a little bit however, there become a host of consensus algorithms that work much more efficiently than proof-of-work.

Zookeeper does hardly more than keep a roster of what machines are participating in the larger distributed system. It maybe that a little cluster of 7 Zookeeper processes is a central point of failure (in the sense you need 4 of the 7 running) but that is not "highly centralized administration" because it administers so little.

You are very right about proof-of-work being an expensive way to get around a tiny amount of trust in some arbiter.

Two major figures in the cryptocurrency scene are Neha Narula of MIT Media Lab's Digital Currency Initiative, and Emin Gun Sirer from Cornell. Both are distributed systems and databases researchers. This is because, as KeivnH says, distributed systems research provides consensus models for constraints different than Bitcoin's domain.

Comments for this post are closed