cargo

/cargo/index/

Rust crates, over the sparse index cargo prefers

Point your client at it

toml

[source.crates-io]
replace-with = "libworld"

[source.libworld]
registry = "sparse+http://HOST:10220/cargo/index/"

Replace HOST with your mirror’s address. Port 10220 is the server’s default.

Upstream: https://index.crates.io
What is stored: Crates in the blob store and catalogued; index files on local disk.

How it works

Each .crate is checked against the cksum its line in the sparse index publishes, before the bytes are stored.
cargo used to resolve dependencies by cloning one enormous index repository. The sparse protocol replaced that with ordinary HTTP lookups per crate, and that is what LibWorld speaks — so resolving a handful of crates fetches a handful of small files rather than the whole registry.
A .crate file is immutable: crates.io does not permit republishing a version. So a crate is cached permanently on first fetch.
Index files are revalidated, because a crate gains versions over time.

Things that have cost real time

Written down because they were paid for once already.

Splitting <name>-<version>.crate is not a split on the dash. The boundary is the first - that is followed by a digit, which is the only rule that gets both foo-bar-1.0.0 and foo-1.0.0-alpha right. Anything simpler mis-catalogues every crate with a dashed name, and does it quietly.