Editing Thoughts on blockchain scalability (section)

= SGX-based chains =

Something that seems to have been overlooked in current consensus design is the possibility of using trusted hardware to solve trust problems. I know, I know, I said “trusted” hardware. But if you think about it blockchains are already ''practically'' trusted given that it’s beyond our current resources to ''practically'' verify them every time we use them.

<blockquote>Protip: See “[http://wiki.c2.com/?TheKenThompsonHack Ken Thompson hack]”
</blockquote>
In reality the blockchains we use today rely on a mountain of trust-debt that accumulates to form an informal reputation system. Reputation systems like this are potentially harmful because we don’t know exactly who has what influence, nor how that influence may be exploited in the future…

Because of this, I find it curious that whenever people try talk about trusted systems in the blockchain space, we act like we’re not already using one. In this light trusted hardware doesn’t seem like such a bad option.

I’m not about to say that trusted hardware is some quantum leap forwards in consensus design (it’s not)… But I am saying that it has benefits worth considering. '''Here are some of the things you can do with [https://software.intel.com/en-us/sgx Intel’s SGX]:'''

* Run trusted code on untrusted hosts. Use this for a decentralized [https://shapeshift.io/#/coins shapeshift], low-trust dead-mans switch, escrow agents, and so on.
* More elegant soft forks since you can prove who is running what.
* A way to upgrade any blockchains with any rules you like, since you can write rules that are only run by protected routines based on your requirements and there is no trusted, third-party left to hack.
* Distributed, cross-blockchain smart contracts written in any language.
* Scale any existing blockchain by having provable transaction attestations.
* Improve privacy in payment protocols.
* Use it to create decentralized, autonomous agents on public infrastructure with very low trust. You can distribute agents that [https://blog.ethereum.org/2014/05/06/daos-dacs-das-and-more-an-incomplete-terminology-guide/ run entire corporations] this way that can be arbitrarily complex, even with private information.
* Use it to share public resources with private organizations, leading to greater decentralization, privacy, along with more efficient markets.
* Solve several unsolved trust problems in decentralized storage systems.
* Poormans [https://bitcoinmagazine.com/articles/cryptographic-code-obfuscation-decentralized-autonomous-organizations-huge-leap-forward-1391849871/ Indistinguishability Obfuscation].
* Run near bullet-proof, anonymous, decentralized marketplaces on public infrastructure and store it across obfuscated blackboxes.
* Reduce attacks in the Lightning Network ([http://hackingdistributed.com/2016/12/22/scaling-bitcoin-with-secure-hardware/ see Teechan].)
* Potentially a better way to do ICOs as companies building software on [http://roberts.pm/permissioned_resource_coins decentralized community-run infrastructure] would have to depend on their community for hosting so the relationship would be more equal.
* '''Practical multi-party computation.''' Not just 1 + 1 between hosts (sorry researchers but that’s really underwhelming.)
* [Your blockchain problem here?]

'''Yes, indeed, you can do far more with trusted computing in distributed systems then you can with blockchains.''' So I think it would be a waste to dismiss it completely because of conspiracy theories about Intel.

'''Tl; dr:''' To continue to make progress we can’t let dogmatic thinking overcome us, and most of all: we need to be honest about how these systems work in the real world and not just rely on wishful thinking.