The design of Rust encourages having many crates because that's the only surefire way to have parallel and incremental compilation. If I were writing a large application in Rust, I would absolutely split things into as many crates as possible but keep it in a single repo for this reason.
The problem highlited here is not that there are distinct compilation and dependency units, bit that they are organized in a flat namespace.
Anybody can create a crate like "aws-sdk-cloud" and have a crate that looks official but is not published by AWS (and is one character edit away from a legitimate AWS SDK crate)
Comments
The design of Rust encourages having many crates because that's the only surefire way to have parallel and incremental compilation. If I were writing a large application in Rust, I would absolutely split things into as many crates as possible but keep it in a single repo for this reason.
The problem highlited here is not that there are distinct compilation and dependency units, bit that they are organized in a flat namespace.
Anybody can create a crate like "aws-sdk-cloud" and have a crate that looks official but is not published by AWS (and is one character edit away from a legitimate AWS SDK crate)
Does it mean that if I change a single line of a crate, then the whole crate must be recompiled?