He's also constantly inundated with arguments/accusations/requests from idiots. If you ever read through some of his Linux newsletter arguments, you can see why he'd develop a take-no-prisoners style of communicating.

He could take the killfile approach to lkml... of course that would get him just as much hate. It's really a lose-lose situation, no good way of getting people to stop with the "im right and won't stop pestering you even though it's your project" stuff without pissing off one half of the internet or the other.

The oddest thing is, as I read random lkml threads that interest me, it seems that Linus isn't a jerk immediately when people bring up stuff - even well hashed stuff. When he is wrong, he admits it, and seems reasonable about patches that aren't perfect/up to snuff, particularly from newer devs who are trying to do well. It's when the threads go on and on that it seems to degrade. (or when a topic that is famous for being a no-go is brought up... but i think it's fair to at least understand the thing you are trying to contribute to, particularly as well documented as linux).

To illustrate the point here is a 'complex' email from him on the 16th: https://lkml.org/lkml/2012/5/19/97

Now here is him on an un-related issue apologising: https://lkml.org/lkml/2012/5/21/230

For linus the github 'discussion' was an everyday thing. He speaks his mind and will apologise if he is wrong, such is the order of the day on the kernel mailing list.