Originally, POSIX didn't specify the O_EXCL flag, using much the same reasoning you're using now. After all, what harm could come from some other process creating a file in between checking if it already exists and actually performing the open? Then, someone (and I can't find the link to the story right now, sorry) demonstrated that this race condition opened up a security hole. They could trick a legitimate program into overwriting a sensitive file that the attacker wouldn't normally have access to. After a lot of back-and-forth, the Gods of POSIX eventually relented and added the O_EXCL flag to the open(2) function. Now, I get that the odds of this actually opening an exploitable security hole are remote. But they aren't zero. And if you pay attention to computer security news, you'll see that "remote chance of a security hole" exploits crop up all the time. It's enough to make any developer (such as myself) just a little bit paranoid about my code. But, the thing is, there are absolutely no downsides to allowing exclusive open whatsoever. So, no matter how small the improvement may or may not be, if it is **an** improvement with no drawbacks, why **wouldn't** you implement it?