The archived material is frozen. Do not edit it. Preserved copies of Ethereum Classic material that upstreams have deprecated, deleted, or will. None of it is maintained, corrected, or relabeled by this repository. It is the historical record, and its value is that it is exactly what upstream published. The documents that describe it, this one among them, are kept current.
Most entries are whole upstream repositories. Each is merged under <org>/<repo>/ at one ref,
without squashing, and its other refs are merged as history only, so every upstream commit keeps its
original hash and stays reachable from main. A fork that adds only a few commits to another
entry's history can be held as history only, with no folder. Where only part of a repository is
Ethereum Classic material, or the material was never a repository, the entry is an extraction
instead, recorded with the commit or revision it came from. AGENTS.md sets out each form.
PROVENANCE.md indexes every repository held here, and each one's record,
provenance/<org>/<repo>.md, gives its refs, dates and known defects. Read the record before using
anything here.
The archive is described below by what its material is. Each section says what it holds; the index
in PROVENANCE.md lists the repositories.
The consensus tests of each of this chain's eras, from the original Ethereum Classic suite to the
suites ETCDEV, ETC Labs and multi-geth carried, with the tests written for single ECIPs, tests that
later left their upstreams, the tools that filled and ran the suites, retesteth and hive among
them, and results recorded against the live network.
The clients that carried this chain, whole and with history. This chain has had core-dev churn
where Ethereum has had one client throughout, so its clients ARE the record of its eras, and a
subset cannot show how a client changed across one. Classic Geth was the first client specific to
Ethereum Classic. Parity followed, held at the DAO fork and at its last state supporting the chain,
beside Ethereum Classic's own fork of it; then multi-geth, and core-geth. Besu's support for the
chain and the ETC Cooperative's plugin for Nethermind are held too, and OpenEthereum as an
extraction of its chain specifications. EXTRACTION-historic-clients.md records why multi-geth,
Parity and OpenEthereum were frozen where they were, and what the OpenEthereum extraction holds.
The software this chain mines with, and where it came from: Ethash's reference implementation and its copies and forks across this chain's organizations; Etchash, ECIP-1099's change to it, in Go, C and C++, with the research behind it; the miners built on them; the open-ethereum-pool line of mining pools and its Ethereum Classic forks, stratum servers and mining proxies; and, from before Ethash, a 2014 prototype of the proof of work.
The virtual machine ETCDEV wrote in Rust for this chain's clients, whole and with history, and what
was built around it: SputnikVM itself, with its patches for the chains it ran on; etcommon, the
crates of Ethereum's common structures it builds on; its C and Go bindings, whose Go package Classic
Geth imports; a development node built on it, with a web interface; a callback wrapper; and the
forks that carried it on at ETC Labs and under the name evm.
What was built around the clients: JSON-RPC and OpenRPC specifications and the tools that generate them, serialization helpers, signers and signing services, a chain configuration specification, a block explorer, node operation and packaging, research and simulation work on the chain's economics and security, and the smaller sites and tools of its community.
Some entries are live build dependencies. ethereumclassic/core-geth builds against Go modules held
here, and a replace can point at this repository for each one. Branches named
dep-<org>-<repo>-<commit> point at commits the client pins, so the go command fetches that commit
rather than the whole archive. PROVENANCE.md gives the replace and go.sum lines, and
AGENTS.md the rules those branches keep.
Read the organization, not the bare name.
ethereumclassic/core-geth is deliberately absent. It is the live production client, so it
fails the test this archive applies: what disappears if the upstream vanishes tomorrow. Vendor it
when it is actually deprecated, not before.
etclabscore/core-geth is a different repository and is held here, frozen at the commit the
successor was created from. It holds the era the ETC Cooperative funded, from January 2022 until
that organization moved to maintenance mode at the end of 2024, and its record cites the
Cooperative's own published record for both ends of it. Its upstream is alive and has committed past
that commit; that continuation is deliberately unreachable here, and completing it would erase the
boundary the entry exists to record.
Parts of this material are wrong, and they stay wrong.
The clearest case: etclabscore's Ethereum Classic fork labels were produced by a text
substitution over Ethereum fixtures, which an Ethereum Classic client then filled again. The
substitution left gaps: no filled test carries a label for Spiral (ECIP-1109), and some files
still carry Ethereum's own Merge and Shanghai labels.
Do not fix them here. A corrected mirror is no longer a mirror: it cannot be compared against what upstream published, which is the only reason to keep a copy of a dead corpus. The defects are documented in each repository's record, so a reader knows what they are holding.
Build a suite outside this repository, using this as source material. Where a fixture here is sound, a suite can reference it. Where it is wrong, the suite carries the correct version, as its own work named its own way, with this archive left untouched as the record of what was inherited.
- No formatting tool runs here. A conformance vector's bytes are its meaning. This repository configures no hook, and any tool added later must report rather than rewrite.
- Mirrored names are kept exactly: directories, filenames, and the fork labels inside a fixture.
- Adding to the archive is expected; changing it is not. When an upstream deletes something else, extract it here and record the refs.
- A vendored tree is verified by TREE HASH, never by diff.
git subtreerelocates a corpus under a prefix, sogit diff <ref>..HEAD -- <path>reports every file as added and proves nothing. Comparegit rev-parse '<ref>^{tree}'againstgit rev-parse 'HEAD:<org>/<repo>'; two identical hashes is the whole proof. - Cloning this repository clones every entry's history. Everything is reachable from
main, so a consumer that needs one entry, such as a submodule pinned to one upstream commit, still receives the whole archive. The go command is the exception, for a commit adep-branch points at: it fetches that commit alone.PROVENANCE.mdsays what this means for each consumer it names.