The awkward Sia fork

Tonysheng.com is supported by the generosity of its members. In addition funding the free weekly essay (published on Mondays), members receive an additional members-only analysis that is timely and substantial. You can become a member for $12/month or $100/year. The following is a sample of a members-only analysis from Oct 09, 2018.

On Monday, I wrote about conflicts of interest. Since then, I caught up on Nebulous/ Sia’s controversial decision to hard-fork their network to brick third-party ASIC miners (Innosilicon and Bitmain).

Couldn’t have asked for a better example of conflicts of interest.

A brief summary of events:

• Mid-2015: a company, Nebulous, launched the Sia network–a file storage network
• September 2016: Nebulous had raised $1.25M
• June 2017: Nebulous announced they would be building their own ASICs: Obelisk and secured millions in presales of the hardware
• Soon thereafter: it become clear that third parties were also creating Sia ASICs, namely Innosilicon and Bitmain
• April 2018: Nebulous sold ~$1.5M in Siafunds¹, a tokenized security
• October 2018: Nebulous decided to initiate a hard-fork to brick third-party ASIC miners, using a “kill switch” programmed into their Obelisk miner

What a mess of incentives. The Nebulous team has many bosses:

1. Nebulous equity holders
2. Siafund owners
3. Obelisk equity holders
4. Siacoin owners
5. Obelisk ASIC owners
6. The Sia “community”

When interests compete, who wins? By law, Nebulous must make decision to create shareholder value for Nebulous, right? But don’t they have a similar fiduciary duty towards Siafund owners and Obelisk equity holders (which may be the same as Nebulous owners, I don’t know).

Does bricking third-party ASICs benefit the Sia network? Possibly. The rationale presented by Vorick seems plausible but pretty much impossible to confirm.

I’ll try my best to summarize their position using quotes from the post.

We believe that an ASIC manufacturer monopoly is strongly preferable to being a GPU mined coin. To the best of my knowledge, no ASIC mined coin has ever been hit by a 51% attack*. And yet, something like a dozen different GPU mined coins have been hit by 51% attacks, typically to steal from exchanges.

and

Cryptocurrencies with a block reward below hundreds of millions of dollars per year will not be able to profitably sustain multiple ASIC manufacturers. This cleanly explains why Bitcoin is the only cryptocurrency with multiple reasonably competitive machines.

and

failsafes will typically never be used. However, in the event of an unhealthy monopoly miner, or even an outright attack from a majority hashrate owner, these failsafes can be leveraged to break the attacker’s hardware without breaking all of the ASICs on the network, allowing the network to continue to be protected by ASICs, but escape the attacker or abuser.

In sum, with the end goal of securing the Sia network:

• ASICs, even an ASIC monopoly is preferred to GPU mining
• … but cryptocurrencies with relatively smaller block rewards cannot sustain a diverse ecosystem of ASIC manufacturers
• … so it’s important that the resulting monopoly ASIC manufacturer is a “healthy” one

The healthiness of an ASIC manufacturer appears to be a subjective judgement on whether they are creating the ASIC for the benefit of the network or not.

For the Sia network, an important line was crossed when secret ASIC projects superseded a public project that had substantial community investment. Though the Obelisk project got a few things wrong, largely Obelisk went about ASIC development and manufacturing in the right way. Obelisk was focused on adding value to Sia, to the Sia community, and to the cryptocurrency community as a whole, which is what we expect from a service provider.

Vorick takes the challengeable stance that because Obelisk was developed with community support, it is helathier than the third party ASICs which were developed in “secret”.

Now this might be true and an Obelisk monopoly may be the best choice for the Sia network. But one cannot ignore the massive conflicts of interest. Of course Nebulous supports Obelisk–they created it! Given their logic, if they did not brick third-party ASICs, the Obelisk would not be profitable and they would not only disappoint equity holders, but also Obelisk customers who are also Sia community members.

Again, I am not making a judgement on whether this was the right decision or not–it’s just so very clearly in the best interest of the equity entities that it’s not hard to imagine a scenario where it’s not best for the network, but it is best for Nebulous.

This is just the beginning of these conflicts. We will see more and more of them as teams integrate vertically and horizontally. Horizontally like this–building miners for their network–and vertically like Dharma or dydx building user-facing apps on top of their infrastructure. Lots of awkwardness to come.

I’m not sure what the lesson here is. It could be “don’t create conflicts of interest”, but maybe Nebulous will prove that these conflicts were unavoidable on the path to bootstrapping the Sia network.

A note on forking

This decision also offers a fascinating example on forking. In response to complaints that this decision to brick third-party ASICs was too “centralized,” Vorick and team emphasize that the fork is opt-in. You can decide not to join the fork chain:

This leaves people with the opportunity to reject the upgrade, and to instead continue using the old software and the old blockchain. If enough people rally around the old software, there could be a network split, and Sia could divide into two blockchains, in the same way that Ethereum split into Ethereum and Ethereum Classic, and in the same way that Bitcoin became Bitcoin and Bitcoin Cash.

At Nebulous, we view these cryptocurrency splits as one of the most powerful innovations of the blockchain space. Under traditional governance structures, a single decision gets made and everyone has to live with that decision. But when the network is able to split, you can get solutions where two groups of people with incompatible demands can both get what they want.

Of course, it’s awkward that Nebulous controls so much of the ecosystem. They develop the network, own most of the Siafunds, and also control the manufacturing of ASICs. Their control and influence in the community confounds the concept of an opt-in fork.

Yes, technically people could use the other fork, but it would require immense coordination to make that other fork useful. For a project in its early stages, it doesn’t feel worth the effort. If instead the fork was initiated by Innosilicon of Bitmain, then we would feel the ethos of opt-in. But when a fork is driven by the company at the center of the network, it feels more like a planned upgrade.

This whole thing is just awkward. Consider this sentence (posted yesterday) from a16z’s Scott Kupor on what a16z crypto is excited to invest in:

cryptonetworks are a new way to build digital services, where those services are owned and operated by a community of network participants rather than by a centralized corporation.

Wouldn’t it be nice if it were so simple.