Explicitly mention the generated file in rust/gen/README.md (#473)

rust/gen/README.md

@@ -1,10 +1,27 @@
 This crate is for maintenance purposes only. It is used to update the Rust library to a new model.
-It relies on 2 symbolic links to be coherent:
-- `rust/lib/src/model.onnx` should point to the model.
-- `rust/gen/data/config.json` should point to the definition of the target labels for that model.
+There are 2 files in the Rust library that depend on the model:
 
-Ideally if those 2 symbolic links are incoherent some test should fail, such that a PR updating the
-model without updating the config would not pass the continuous integration.
+- The model itself, `rust/lib/src/model.onnx`, which can either be a symbolic link to the source of
+  truth in this repository (publishing the crate will dereference the symbolic link), or a copy of
+  that source of truth (in which case there should be a test to make sure the file is an up-to-date
+  copy).
+- The labels describing the model output, `rust/lib/src/label.rs`, which is generated from the model
+  configuration, `rust/gen/data/config.json`, which can either be a symbolic link or a copy to the
+  source of truth in this repository, and should match the model above.
 
-There is another test to make sure the library code is in sync with the config. If this test fails,
-one simply needs to run `./sync.sh` from the `rust` directory to update the library.
+The purpose of this crate is to generate this second file. There is a test to make sure that this
+file is up-to-date. If the test fails, one simply needs to run `./sync.sh` from the `rust` directory
+to regenerate the file.
+
+An alternative design to generating the file before publishing the crate, would be to publish the
+model configuration and use a build script to generate the file during compilation. This has a few
+disadvantages:
+
+- We need to publish the model configuration which contains more information than needed to use the
+  library.
+- We need to use a build script, which is frown upon for security reasons, as the entity compiling
+  the library or CLI now needs to trust the build script, which can run arbitrary code. This only
+  matters when the entity compiling the library or CLI is not the same as the one running the
+  library or CLI (e.g. Debian maintainers), since the library and CLI too can run arbitrary code.
+- Using a build script also increases compilation time (and compilation complexity) instead of
+  having it factored before publishing.

rust/gen/src/main.rs

@@ -49,6 +49,8 @@ fn main() -> Result<()> {
     let header = header.split("\n\n").next().context("main.rs does not contain an empty line")?;
     writeln!(output, "{header}\n")?;
     writeln!(output, "use crate::output::Metadata;\n")?;
+    writeln!(output, "// DO NOT EDIT, see link below for more information:")?;
+    writeln!(output, "// https://github.com/google/magika/tree/main/rust/gen\n")?;
     writeln!(output, "/// Content type of a file.")?;
     writeln!(output, "#[derive(Debug, Copy, Clone, PartialEq, Eq)]\n#[repr(u32)]")?;
     writeln!(output, "pub enum Label {{")?;

rust/lib/src/label.rs

@@ -14,6 +14,9 @@
 
 use crate::output::Metadata;
 
+// DO NOT EDIT, see link below for more information:
+// https://github.com/google/magika/tree/main/rust/gen
+
 /// Content type of a file.
 #[derive(Debug, Copy, Clone, PartialEq, Eq)]
 #[repr(u32)]