I argue that the OS itself must be retarded to guarantee that reads are matched to writes. The OS isn't doing that. read(1), the user program, does that; other programs can do as they like.
ur tarded. You can't implement read(1) guaranteeing atomic `read` on an OS that doesn't guarantee atomicness of reads and writes. The fact that they flaunt the tarded decision by implementing userspace programs that reinforce the guarantee means that they're all retarded.
Again, you don't have to do what read(1) does. It would be insane to suggest otherwise.
Not me, nobody has to do that, nobody wants to do that, and yet they guarantee that.
Caller-allocates-memory is standard practice for C programs, and for a language that operates at a systems level it's the only sane thing to do. The read routine should not be responsible for allocating memory when the caller knows best where the read buffer ought to be.
Why do you explain the obvious? Do you believe that if you explain the obvious, no, wait a second
Also, saying that it's impossible to use read properly is just total bullshit. Allocate a fixed size buffer, call read with the size of the buffer, check the return value. That can't overflow, ever.
Ah! You are retarded in truth! Like, I'm not trying to offend you, but I realize that you're actually retarded and unilaterally terminate the discussion.