It seems to me that it is the obscurity of source code that enables copyright violation.
If software could not be copyright unless it was registered with the library of congress, then the library could run a filter on all new arrivals and check they were different from previously posted source from other sources.
What if two people independently arrive at thesame algorithm, say, an optimal implementation with the least possible amount of code in a given language. Is it still a copyright violation? You can't copyright math, can you?
Legally that is not a copyright violation. There are examples where this is likely, and companies make efforts to prove this no copying occurred. I think this is called a 'clean room' implementation.
If software could not be copyright unless it was registered with the library of congress, then the library could run a filter on all new arrivals and check they were different from previously posted source from other sources.