>> for something he calls adversarial interoperability

Cory Doctorow has had a lot of interesting ideas, but he didn't coin the term adveserial interoperability. It's a concept and term that predates the Web, and is absolutely the appropriate solution to many of our current predicaments.

All of this has happened before and it will all happen again.

> make it unenforceable for a TOS to ban reverse-engineering for compatibility reasons

While I appreciate that (and since I don't appreciate my work being locked up in proprietary formats, I primarily use either plain-text or LaTeX for documents), there's a whole can of worms around that.

[note that I do not know the specific details of .PSD files, nor the patent status of things below. It's an example]

As an example off the top of my head, Photoshop has a feature called "Content-aware Fill". This lets you delete a region of an image and have it be replaced by auto-generated "reasonable" context (if you delete someone standing on a hill with the sky behind them, it'll interpolate grass and sky for you).

So now I'm reverse-engineering the .PSD format, and discover that a content-aware fill operation is stored in the file as the pixel coordinates and CAF parameters; not the pixels themselves. The CAF algorithm is patented. What now? Under your proposal I'd be justified in carrying on figuring out how to make an interoperable implementation, but... I couldn't distribute that implementation without violating Adobe's patent.

When Indesign was new and everyone was pissed at Quark, Adobe made it possible to import Quark files. It wasn’t perfect, but it was good enough to get you over the transition. Something like this would surely be possible with Indesign files.

The DMCA has a provision allowing that. It is explicitly legal to reverse engineer file formats for the purpose of interoperability.