My real world AWS sdk s3 experience: I have an uploader interface (I only needed to upload). The concrete uploader has the s3 code. The fake uploader has any custom result/err I need to test against.