There isn’t actually a distinction between in, out, and inout. Is this intended for backward-compatibility or future-proofing?

[ParadoxV5]

• Struct.ptr does not utilize the optional argument flags (not even checking if it’s a valid choice).
  

  ffi/lib/ffi/struct.rb

  Lines 131 to 133 in f2d3846

  	def self.ptr(flags = :inout)
  	@ref_data_type ||= Type::Mapped.new(StructByReference.new(self))
  	end

• Struct and Buffer’s . alloc/new _ inout/in/out are all either an alias of, or is nothing more than a delegate to, .new.
  

  ffi/ext/ffi_c/Struct.c

  Lines 859 to 864 in f2d3846

  	rb_define_alias(rb_singleton_class(StructClass), "alloc_in", "new");
  	rb_define_alias(rb_singleton_class(StructClass), "alloc_out", "new");
  	rb_define_alias(rb_singleton_class(StructClass), "alloc_inout", "new");
  	rb_define_alias(rb_singleton_class(StructClass), "new_in", "new");
  	rb_define_alias(rb_singleton_class(StructClass), "new_out", "new");
  	rb_define_alias(rb_singleton_class(StructClass), "new_inout", "new");

  
  

  ffi/ext/ffi_c/Buffer.c

  Lines 194 to 198 in f2d3846

  	static VALUE
  	buffer_alloc_inout(int argc, VALUE* argv, VALUE klass)
  	{
  	return buffer_initialize(argc, argv, buffer_allocate(klass));
  	}

  
  

  ffi/ext/ffi_c/Buffer.c

  Lines 387 to 407 in f2d3846

  	/*
  	* Document-method: alloc_inout
  	* call-seq: alloc_inout(*args)
  	* Create a new Buffer for in and out arguments (alias : <i>new_inout</i>).
  	*/
  	rb_define_singleton_method(BufferClass, "alloc_inout", buffer_alloc_inout, -1);
  	/*
  	* Document-method: alloc_out
  	* call-seq: alloc_out(*args)
  	* Create a new Buffer for out arguments (alias : <i>new_out</i>).
  	*/
  	rb_define_singleton_method(BufferClass, "alloc_out", buffer_alloc_inout, -1);
  	/*
  	* Document-method: alloc_in
  	* call-seq: alloc_in(*args)
  	* Create a new Buffer for in arguments (alias : <i>new_in</i>).
  	*/
  	rb_define_singleton_method(BufferClass, "alloc_in", buffer_alloc_inout, -1);
  	rb_define_alias(rb_singleton_class(BufferClass), "new_in", "alloc_in");
  	rb_define_alias(rb_singleton_class(BufferClass), "new_out", "alloc_out");
  	rb_define_alias(rb_singleton_class(BufferClass), "new_inout", "alloc_inout");

[larskanis]

I think it is future-proofing. I know the declaration of the direction of buffer usage from OpenGL, where the memory location can be different according to this information.

Also our Wiki describes them as a performance optimization: https://github.com/ffi/ffi/wiki/Types

[larskanis]

The very first commit with in/out buffers is 2592f27 .