Solizone EVMChapter 15 / 17
The next phase
Proving it on a live network, then hardening it.
The next major phase
The next major development phase is about making Solizone EVM look like an Ethereum network to external developer tooling.
The target flow is:
Foundry / MetaMask / ethers / viem
↓
Ethereum JSON-RPC
↓
signed Ethereum transaction
↓
mempool
↓
BlockProducer
↓
REVM
↓
Solizone EVM state
↓
canonical blockThat requires several components.
Raw Ethereum transactions
Today, the internal runtime constructs Solizone transactions directly.
The next step is:
raw signed transaction
↓
decode envelope
↓
recover sender
↓
validate nonce
↓
validate gas / fees
↓
admit transaction
↓
Solizone executionThis is the bridge between Ethereum tooling and the current execution engine.
Transaction pool
Once external users can submit transactions, those transactions need somewhere to wait before block production.
RPC
↓
transaction
↓
mempool
↓
BlockProducerThe mempool will become the boundary between transaction ingestion and block construction.
Ethereum JSON-RPC
The developer-facing goal is that familiar tools can talk to Solizone EVM using APIs they already understand.
Examples:
eth_sendRawTransaction
eth_call
eth_getBalance
eth_getTransactionCount
eth_getCode
eth_getTransactionReceiptThe exact supported RPC surface will grow gradually.
Re-execution validation
Another important future milestone is independent block execution.
A verifier should eventually be able to receive a Solizone block and do:
previous state
↓
block transactions
↓
REVM
↓
new state
↓
calculate commitments
↓
compare with blockThis strengthens independent verification and is important for a multi-node execution environment.
Canonical discovery
Today, Solizone EVM has local metadata that knows:
latest checkpoint CID
height
→ block hash
→ CIDThis works for the current node.
A distributed execution network needs more.
Nodes need a reliable way to discover:
What is the latest canonical Solizone head?
What is the latest accepted checkpoint?
Where is block N?
Which CID corresponds to the canonical block?This discovery mechanism is still future work.
Replication and durability
Content-addressed storage tells us what object we are retrieving.
It does not automatically answer:
How many nodes store it?
How long will they store it?
What happens if providers leave?
What replication factor do we require?Those are important protocol and operational questions.
Solizone EVM's current integration proves the storage lifecycle.
A production deployment will need an explicit replication strategy.
Crash consistency
The long-running node currently follows:
produce block
↓
store block
↓
save checkpointThis is the correct logical order because a checkpoint should never point to a block that does not exist.
But a production node must also handle failures between those steps.
For example:
block saved
↓
process crashes
↓
checkpoint not yet updatedFuture hardening needs reconciliation logic or an atomic commit design for this boundary.
A future Solizone EVM node
The longer-term node architecture looks roughly like this:
Ethereum tooling
↓
JSON-RPC
↓
mempool
↓
BlockProducer
↓
REVM
↓
EVM state
↓
Solizone block
/ \
/ \
↓ ↓
recent history publication
↓ ↓
local storage Logos
│
↓
historical archive
↓
Logos StorageA verifier could independently replay blocks and check resulting commitments.
That is the direction Solizone EVM is moving toward.
What's next?
The next major milestones for Solizone EVM are:
- raw signed Ethereum transaction support
- transaction decoding and sender recovery
- transaction admission rules
- Solizone transaction pool
- Ethereum-compatible JSON-RPC
- Foundry / ethers / viem / wallet connectivity
- independent block re-execution
- stronger canonical discovery
- production storage durability
- production node hardening
The goal remains simple:
make EVM development feel familiar while building it natively around the Logos architecture.