> the burden of proof would be on anyone claiming that a large binary is direct creation
That's not how any of this works. Once I've submitted my binary to the escrow, then it's an original work worthy of copyright until challenged. If someone breaks my copyright, I simply assert my originality and get an injunction against it by default. If the adversary claims my work is a derivative, then the legal burden is on them to prove it - because they are the ones rising it as defense against infringement. So I've already quashed 90% of adversaries by this point just by legal intimidation.
Also, you are attacking a strawman version of this problem. In practice, what will be submitted to the repository are binary object files for the core proprietary sections that are unlikely to be changed, for example the file format definitions to preclude interoperability. Anything else can be source, the final linker step is automated etc. The market will also offer tools for binary randomization and compiler signature obfuscation.
So you are left with a technological arms race that needs to be settled in court, on a case by case basis, using expert testimony where the burden of proof belongs to the infringer. It's just absurd to think anything like this could ever work in practice to promote source availability, or there would be public benefits to put such a highly litigious system in place.
> That's not how any of this works. Once I've submitted my binary to the escrow, then it's an original work worthy of copyright until challenged
Are you referring to some existing system? Because otherwise we're discussing in the abstract how this could work. And it would be straightforward for the copyright office to have a rule prima facie rejecting machine-executable binaries being claimed as original works, unless compelling proof were submitted that it had been directly created. Because as I said, show me any substantial executable that has been directly created with a hex editor.
The idea of a company choosing a few critical parts that they then (honestly or dishonestly) claim as original binary works is interesting. But note that this still would go a long way towards making the overall work part of the public domain when it expired - such blobs are highly unlikely to contain references to libraries with API churn and whatnot. And even if they somehow do, since they are small enough to be directly worked on, they can be directly fixed.
That's not how any of this works. Once I've submitted my binary to the escrow, then it's an original work worthy of copyright until challenged. If someone breaks my copyright, I simply assert my originality and get an injunction against it by default. If the adversary claims my work is a derivative, then the legal burden is on them to prove it - because they are the ones rising it as defense against infringement. So I've already quashed 90% of adversaries by this point just by legal intimidation.
Also, you are attacking a strawman version of this problem. In practice, what will be submitted to the repository are binary object files for the core proprietary sections that are unlikely to be changed, for example the file format definitions to preclude interoperability. Anything else can be source, the final linker step is automated etc. The market will also offer tools for binary randomization and compiler signature obfuscation.
So you are left with a technological arms race that needs to be settled in court, on a case by case basis, using expert testimony where the burden of proof belongs to the infringer. It's just absurd to think anything like this could ever work in practice to promote source availability, or there would be public benefits to put such a highly litigious system in place.