Your analogy doesn't quite fit. Copyright covers both the source code, which is more analogous to the schematics than the manual, and the binaries.

If a physical machine is not under patent protection you can strait up copy it and start manufacturing your own version. With software you have to go through an extra "clean room" step to get around copyright. In practice the clean room reverse engineering process is rarely utilized because it is expensive and fraught with the risk of a lawsuit from the copyright owner.

Yes, this is a major issue in copyright law. In practice, there is no distinction between the patent-eligible invention (the machine) and the copyright-eligible "creative work" (the manual, or the code, in this case).

That means that copyrights can be exerted in very menacing ways, like requiring users to bypass a copy protection mechanism in order to do required maintenance, as John Deere has done. That enables John Deere to pursue anyone who doesn't use an official dealer for maintenance under the DMCA's measures criminalizing the circumvention of copy protections.

Disclaimer: Not a lawyer, probably wrong.

Saying that there is no distinction isn't correct.

The scope of a patent is much broader than a single coded implementation. The copyright doesn't protect the functional aspects of the code, it only protects it as a written work. However, companies have managed to twist the concept of infringement via reproduction, and bypassing of copy protection.

Accessing a copy of code stored in memory is now considered creation of an unauthorized reproduction, bypassing copy protection under DMCA, and/or the like.

This type of protection shouldn't exist apart from preventing partial and wholesale copying and redistribution, paralleling the treatment of novels, etc. If I own a novel, I can do whatever I want to do with that book. I can write on it, annotate the margins, remove pages, make photocopies of parts of the book for personal use, etc.

With code, I can't do this? Why?

Yeah, I agree. That's what I'm getting at; in the physical world, there is a bright line between "an invention" and "a copy". You buy a reproduction of an invention, or a physical reproduction of a copyrighted work, and that reproduction is a discrete unit owned by the purchaser and not really subject to the limitations of IP (only insofar as the fundamental components are directly copied to deprive the inventor/author of the proceeds).

In cyberspace, many things that should be considered "inventions" for IP purposes, and thus subject to the much more limited extent of patent protection and for which each individual unit is an unprotected entity, are instead considered "new copies" of a copyrighted work, under which substantially more aggressive licensing restrictions are typical, and under which the definition of things like "infringement" and "derivative" are much more restrictive.

A software program should be an invention under patent law, and its copies should be considered discrete physical units. The code itself may be copyrighted, just as the plans for an invention may be, but the produced binaries probably should not be. Rather, they are units of the invention regulated as normal goods.

This has become even harder as we have gone to mostly-digital software distribution. You used to be able to easily resell software if you had a CD. Now you can't even do that (which, by the way, is no small motive for the companies; it completely obliterates the secondary market for software).

IANAL