Updated thoughts on “modular blockchains”

Polynya
5 min readMar 25, 2022

--

I have been guilty of shilling the “modular blockchains” meme last year, of course alongside more influential players like Bankless, Celestia, The Daily Gwei and others who brought it to mainstream attention. This year, I have not really used the term “modular blockchains”.

To be very clear, modularizing absolutely remains orders of magnitude more efficient than monolithic chains. And the many meltdowns of monolithic chains when tested with bare minimum stress has made this clearer than ever. Modular execution layers also have a ton of work to do, but they have significant headstart versus monolithic execution layers.

Where I was wrong was not the technical, but the socioeconomic aspects. Ali Atiia and Justin Drake have highlighted this before. Consider the gold standard, an Ethereum rollup:

  • Execution: rollup
  • Settlement: ethereum
  • Data: ethereum

By the way, lately I’ve seen people call execution layers that settle data elsewhere a “rollup”. This is not a rollup. A rollup must settle on the same settlement and DA layer. A validity proven execution layer that post data on a different layer is called a validium and comes with additional assumptions. You can brand them zkPorter, celestium, whatever, but please don’t call these a rollup. The fraud proof situation is more complicated, so I’ll skip that for now. The point is, if the data availability doesn’t come to consensus on the same protocol that’s verifying state transitions - it’s not a rollup. /rant

A properly implemented rollup means you don’t have to trust the rollup at all, and can exit to ethereum with your funds at any time. However, this is not entirely bulletproof as you’re assuming that the rollup will be properly implemented like so. In reality, you may have rollups with different security models and standards, though I definitely expect all major rollups to offer a clear exit mechanism of some kind isolated from trusting rollup sequencers.

Different rollups come with different assumptions. You could have an immutable or enshrined rollup that offers identical security to ethereum — barring no bugs. To upgrade the rollup, you must use the EIP process; or deploy a new instance entirely. Many rollups will opt for upgradeability, which will be driven by token voting. This is a similar economic assumption to how proof-of-stake L1s upgrade, though rollups can experiment with novel upgrade mechanisms without having a token at all. There are other interesting risks which I won’t cover here — see Justin’s and Ali’s comments above. Personally, I’m not worried about some of these — I believe a well-implemented rollup can be 99% as perfect as an enshrined rollup, but there’ll certainly could be some rollups with not insignificant assumptions.

With EIP-4844 and later danksharding, we add a fourth layer: expired history. This is a pretty easy 1-of-N trust assumption, but I’ll add it to the mix anyway. Other data layers may choose to not expire history. So, now you have:

  • Execution: rollup
  • Settlement: ethereum
  • Data: ethereum
  • History: rollup & others

The ideal solution, instead, should be:

  • Execution: ethereum
  • Settlement: ethereum
  • Data: ethereum
  • History: ethereum

It doesn’t have to be ethereum, it can be bitcoin, but the idea is whatever offers robust security. Now, please don’t mistake this for a monolithic solution. This will be modular, but all enshrined within one protocol:

  • Execution: enshrined rollups (e.g. zkEVM, or optimistic enshrined rollups post-statelessness) [Note: I used to call these “canonical rollups” in my older articles before “enshrined rollups” was coined by Justin Drake.]
  • Settlement: enshrined settlement layer (e.g. EL)
  • Data: enshrined data layer (e.g. danksharding)
  • History: enshrined history layer (e.g. enshrined Portal Network?)

This leaves you with only one protocol with the least assumptions, and maximum security coalescing onto one protocol. AFAIK, Tezos is the only project currently building with this approach; though it’s likely Ethereum will also have its own enshrined rollup(s) years down the line. There will certainly be options with external rollups on top of this. So, you get the best of both worlds: maximum socioeconomic security + experimentation & diversity which in turn will also influence progressing the enshrined layers.

So why not have many “modular L1s”, as described above? You’d want to accumulate rather than divide security & liquidity. Having many “modular L1s” is going to be a very fragmented and insecure mess. However, I think having 2 or 3 modular L1s is the ideal outcome. You have at least one modular L1 that accrues nation state level maximal security, and a couple of others that keeps it in check. There could be a long tail of niche modular L1s. Indeed, this is how most industries consolidate over time — with 2 or 3 major players. The pressures are even more significant for proof-of-stake blockchains because of security accumulation mentioned above.

I have been very interested in volitions in the past, but it quickly becomes a mess of assumptions:

  • Execution: volition
  • Settlement: ethereum
  • Data: ethereum, zkPorter, Celestia, Polygon Avail, adamantium etc.
  • History: the data layers’ respective solutions, the volition, or not

You can see how this is clearly not as elegant as having one modular L1 — although if you are in rollup mode it’s just as good as a rollup. There are many nuances here, but to put it simply, you’re now trusting an additional entity — the DA layer. The most interesting solution, that still interests me, is adamantium. Here, by taking custody of your own data, or choosing your data provider, you forego trusting a different and weaker honest-majority consensus entirely. Honest minority consensus for validium DA layers hasn’t been investigated yet, AFAIK, but I believe holds strong potential as well. (One could say DAC falls into this category, but the problem is it’s permissioned. Also, note that honest minority DA will not work for a rollup — only for validiums that are verifying state transitions on an honest majority settlement layer.)

In reality, however, it’s very likely we are full steam ahead into a world where we see rollups, volitions, validiums, etc. With EIP-4844 and danksharding, there’ll be ample data capacity to settle on Ethereum, but if this blockchain stuff sees exponential growth we’ll see additional data layers handle the surplus demand. We will also see some monolithic chains persist through network effects they are building now and strong bizdev & marketing going forward, despite the inherent insecurity of monolithic bridges and crippling inefficiency of monolithic execution layers. However, I’m absolutely convinced even the most stubborn monolithic project will eventually start pivoting modular components. You simply can’t say no to 1,000x gains in efficiency! Unless, of course, you don’t need scale or innovation.

In an ideal world, though, we can thus form a clear hierarchy of execution layers by security:

Enshrined rollups > rollups >> validiums > AnyTrust >> monolithic sidechains & alt-L1s

--

--

Polynya

Rants and musings on blockchain tech. All content here in the public domain, please feel free to share/adapt/republish.