Dave Hudson is the author of the Hashingit.com blog. He is also the VP of software architecture at immutable ledger firm Peernova, and a long-time designer of operating systems, distributed systems, network stacks, compilers and databases.
In this CoinDesk 2016 in Review special feature, Hudson reviews the various struggles Internet of Things (IoT) innovators have faced since the 1990s – a cautionary tale of what might be in store for blockchain.
2009 saw Satoshi Nakamoto deploy the first bitcoin node, and within five years its blockchain had become a large-scale industry.
But while new applications and commercial opportunities seemed only a short step away, in 2016, we realized that industrial blockchains weren’t going to be so straightforward.
Early enthusiasm for new technologies is nothing new. With most, an initial wave of excitement sees new ideas touted as solutions to a huge range of problems, the hype fades, gives way to skepticism, and ultimately, real applications.
In the late 1990s, the idea of Internet-connecting every electronic device seemed inescapable.
Every vending machine, coffee pot, toaster, refrigerator, microwave, and TV, would be cabled to the “net”, and a utopian sharing of data would improve life for everyone.
The reality for what we now term the “Internet of Things”, or IoT, was a little different.
It’s all about money
The original theory of IoT was that data would make everything better.
Microwave ovens might scan cooking instructions and thus not make mistakes, refrigerators might reorder milk, etc. Automation would liberate users of these appliances, and give them time for other things.
Unfortunately, the theory hadn’t been worked out fully.
Adding Internet connectivity to a device is never free-of-charge. In most cases this was a realm of small, low-CPU-powered devices, with no connectivity, so making them Internet-connected was going to cost money.
In the 20 years since those original ideas, little has changed.
Let’s consider the microwave oven example. A microwave would need a pretty simple IoT hardware design, so perhaps $5 in parts cost. The first problem is that the $5 turns into nearer $15 by the time we add the margins for the company making the circuit boards, the company building the product and the retailer that sells it.
Our next problem is that just having the hardware in our microwave oven isn’t sufficient. We need to make it communicate with servers that know how long, and at what power level, each new type of frozen pizza needs to cook. That implies servers, it implies dev-ops teams, it implies software engineers, and it implies business development people who persuade pizza manufacturers to provide cooking details for each new product they design.
The infrastructure side has perhaps cost us another $10 per unit.
Smart devices are a little like smart contracts.
They’re great when they “just work”, but not so great when people make mistakes. The 1990s vision of IoT involved lots of network cables, but then we got Wi-Fi, and the wires could go away.
Anyone who knows the technology understands that microwave ovens and 2.4 GHz Wi-Fi don’t play nicely together. Similarly, 5 GHz Wi-Fi and solid walls don’t play nicely together.
While our IoT microwave oven might connect to a home router just fine in 95% of homes, the other 5% wouldn’t work reliably, if at all. Unlike software that tends to be notoriously unreliable, microwave ovens pretty much just work.
If they don’t, then customers get irate and phone the manufacturer (more costs), they return “faulty” devices, they leave bad reviews on Amazon and they vow never to buy that brand again.
The idea of a smart microwave might still look great on a PowerPoint slide, but the niggling details turn an interesting concept into a liability. It isn’t worth the setup time and $50 to a customer, and the trouble isn’t worth it for the manufacturer.
Same old story
We have the same challenges when thinking about uses for blockchains.
Not every problem needs a blockchain as a solution. Blockchains cost money in terms of processing, storage and replication technology. In the case of a decentralized cryptocurrency, such as bitcoin, the blockchain-like concept is an essential characteristic to build a viable design, but for other problems we need to ask if the blockchain features are doing something valuable.
If domestic microwaves aren’t an option, then maybe refrigerators might be? Domestic ones have many of the same problems as microwaves, but how about commercial refrigeration? What if we could connect these devices so that if they broke down we could avoid expensive losses?
A large industrial cold store might contain hundreds of thousands of dollars of refrigerated products, so signaling breakdowns and avoiding stock losses must be a valuable problem to solve?
The maths is compelling, but the problem is that it was 25 years ago, too.
While they might not have matched our IoT vision, many companies already found approaches to network these devices a long time ago.
This example has another subtlety. Food storage is generally subject to regulations, and many countries require that records are kept of the temperatures at which products were stored.
Without networking, there would be a need for someone to manually record temperatures every few hours, and this is both expensive and error-prone. Commercial refrigeration equipment also involves service companies and manufacturers providing on-site repairs, so we have more stakeholders for whom access to data is important.
A naïve view of the problem might well have ignored them. Unexpected stakeholders introduce unexpected costs, and may resist changes that do not also offer them substantial benefits.
The implications for blockchains are very similar.
If a problem is already being solved, then, even if a blockchain might be useful, we need to ask if it offers enough incremental advantages? Do we know what all the problems are, including the ones that might not be obvious unless we were domain experts? Are there stakeholders, such as network architects, security experts, data architects, dev-ops teams, etc, who must change existing systems to adopt a new one? Are there analytical needs that require big-data, relational, graph, or time-series, views of any data that is being processed?
Forever is a long time
Leaving aside specific uses of IoT for a moment, it’s worth considering an important characteristic of the devices that were supposed to become smart and connected. These devices don’t get replaced very quickly.
Most of our connected devices get replaced quite quickly. Vendors provide support for a few years but then expect users to discard them and buy new ones.
The problem is we don’t do this with most of our electrical items. We typically only replace them when they fail. By making them connected we introduce entirely new modes of failure.
One such problem is how do we keep older devices working? Typically, manufacturers don’t receive any form of revenue once a device is sold, so what is the incentive to keep providing software updates once those devices are out of warranty?
Another problem is that, even if we might want to pay for updates and bug fixes, it may not be economically feasible to provide them. Older devices will have less powerful hardware that may not lend itself to new features.
A final problem is that our manufacturer may not have considered the possibility of a device becoming compromised.
The recent Mirai botnet has undoubtedly highlighted these issues, but how many toaster manufacturers have the level of security engineering skill to secure, and continue to secure, an IoT device against advanced adversaries?
These are all governance problems. How will our IoT device, once installed, continue to function, and avoid becoming a problem?
The parallels for blockchains are, again, striking.
We have seen major concerns about the governance of the bitcoin and ethereum networks throughout 2016, with both having problems in terms defining operational rules in the face of users pushing the boundaries of the installed designs.
With bitcoin, the block size has seen miners incentivized to restrict block expansion to maximize mining rewards, while The DAO hack incentivized users to want their coins back.
When we consider the deployment of blockchains into other types of applications, then how are these sorts of governance issues to be reviewed and resolved? If we consider systems that might potentially operate for many years, then what does it mean to have immutable storage indefinitely? How will the inevitable mistakes of various human users be corrected? What are the incentives for participants to keep systems running correctly?
In the case of commercial deployments, what are the implications for rolling out updates and upgrades across organizations that have different priorities?
A new hype?
Our journey through the history of IoT has been somewhat cautionary, and there are many unanswered questions, but this is not the story of a lost war.
Twenty years ago, Internet radio stations had barely surfaced, TiVo had yet to produce a set-top box, and ideas of 4k video on-demand streaming were distant science fiction.
Fast forward 20 years later, and designers have leveraged advances in processing, power management, wide-area networking, wireless networking, storage, display technologies and distributed cloud storage, to construct new end-user experiences.
Smart TVs and smartphones are barely recognizable from earlier CRT TVs and crude mobile phones, and yet both have a clear lineage to the original idea of connected things.
IoT arrived but not quite as expected.
Business empires based on the concepts of VHS tapes and DVDs were displaced. Users gained access to far more content, with lower costs and dramatically improved convenience. IoT technologies were not used in isolation, but were combined to solve real problems for the people who ultimately pay for the solutions, customers.
This, then, is part of the challenge for blockchains.
The commercial refrigeration systems slowly changed too. Internet connectivity was a better approach than the ad-hoc methods used 20 years ago, and so replaced earlier designs when they reached natural replacement cycles. Likewise, mature and more capable blockchain designs may well have opportunities to replace other technologies in the future.
Bitcoin stands as the first example of a viable blockchain solution to a well-defined problem. As with many first-generation designs it has also served to highlight challenges, and its ultimate success or failure will depend on its ability to see them resolved.
The challenge for other blockchains might be similar, but won’t be the same.
Blockchain technology will be well served by recognizing, and confronting the hardest problems that we know about, rather than imagining that we can resolve them later. We know that issues such as security, privacy, deployment and governance need to be addressed.
At the same time, we must avoid the temptation to use blockchains, and blockchain ideas, where they are not the best solutions, and champion those where they are.
If we do these things, then 2017 should be a year where blockchain hype gives way to blockchain hope.
Cave paintings image via Shutterstock