^{Originally posted by houseme October 3, 2025}

Summary

The internal-logs feature in the opentelemetry-otlp crate currently enables "opentelemetry-http/internal-logs". However, since opentelemetry-http itself depends on "opentelemetry/internal-logs", this indirection introduces an unnecessary dependency on opentelemetry-http when using grpc-tonic without default or http features.

Current Behavior

In opentelemetry-otlp/Cargo.toml:

[features]
internal-logs = ["tracing", "opentelemetry_sdk/internal-logs", "opentelemetry-http/internal-logs"]

This forces users to depend on opentelemetry-http even when:

• Using grpc-tonic transport
• Not using http or default features
• Only needing internal-logs functionality

Proposed Change

Directly depend on "opentelemetry/internal-logs" instead:

[features]
internal-logs = ["tracing", "opentelemetry_sdk/internal-logs", "opentelemetry/internal-logs"]

Benefits

1. Eliminates unnecessary dependencies
   Removes forced opentelemetry-http dependency for grpc-tonic users
2. Aligns with dependency hierarchy
   Since opentelemetry-http already depends on opentelemetry/internal-logs, this change respects the existing dependency chain
3. Improves compile times
   Particularly for projects using grpc-tonic with default-features = false
4. Maintains compatibility
   No breaking changes - all existing functionality preserved

Verification Steps

1. Check opentelemetry-http/Cargo.toml confirms:

   [features]
   internal-logs = ["opentelemetry/internal-logs"]

2. Test with:

   opentelemetry-otlp = { version = "*", default-features = false, features = ["grpc-tonic", "internal-logs"] }

   Should not pull opentelemetry-http

Additional Context

This micro-optimization supports the project's goal of maintaining a lean dependency tree, especially important for users carefully managing their dependency footprint in performance-sensitive applications.

Would you like me to submit a PR for this change?
I can prepare a minimal PR updating the feature flag if the maintainers agree this is a welcome optimization.

cc @lalitb @bantonsson