EINTR's design is one of computing's absolute classics. To MIT and New Jersey, we should add the McDougals approach: "I cannot work under these conditions." When faced with the PC loser-ing issue, just don't implement the code in question.
McDougals resolves the apparent conflict between the other two. It blames the interrupt hardware as the root cause. It produces non-working, incomplete software. It's kind of a modest proposal.
However, it also produces no ripples in the design fabric. With MIT, the OS source is a maintenance nightmare. With NJ, modern software still has to deal with archaic idiosyncrasies like EINTR. With McDougals, all the "conflict-free" portions of the software advance, those that write themselves.
The result is likely immediately shelved, perhaps as an open source PoC. Over time, someone might write some inelegant glue that makes interrupts appear to behave nicely. Alternatively, the world might become perfect to match the software.
If nothing else, the software will have mimicked the way we learn. We use imperfect examples to draw the idealized conclusion. Even if it never gets to run, it will be more readable and more inspiring than either MIT or NJ.
McDougals resolves the apparent conflict between the other two. It blames the interrupt hardware as the root cause. It produces non-working, incomplete software. It's kind of a modest proposal.
However, it also produces no ripples in the design fabric. With MIT, the OS source is a maintenance nightmare. With NJ, modern software still has to deal with archaic idiosyncrasies like EINTR. With McDougals, all the "conflict-free" portions of the software advance, those that write themselves.
The result is likely immediately shelved, perhaps as an open source PoC. Over time, someone might write some inelegant glue that makes interrupts appear to behave nicely. Alternatively, the world might become perfect to match the software.
If nothing else, the software will have mimicked the way we learn. We use imperfect examples to draw the idealized conclusion. Even if it never gets to run, it will be more readable and more inspiring than either MIT or NJ.