And for those of us on the outside, we wouldn't tell a random developer that he should crank out 40 push-ups to prove his worth, like we might in the military. In the military there is what we'd call a bona-fide occupational qualification behind assessing the psychological durability of recruits.
Is there such as bona-fide requirement to develop software patches? Should there be?
I'll give Linus a pass on this much; his behavior here is well within the norms of the Linux kernel development community and his own previous behavior. He's the Captain of that particular ship and he's not doing anything illegal, nor does he intend personal injury, so at some point it's easier to concede that he will run that ship the way he wants and put on asbestos underwear, than to try to change both his behavior and the resultant mailing list culture.
But I think it's important to keep in mind that his kind of behavior must be very much the exception to the rule for software development, open or otherwise. If you breed a culture of aggressiveness then you must be ready to handle the people you draw in from that. It works for the military, it seems to work for the LKML, but I would be hesitant to hold it up as a general model.
And the military doesn't breed "aggressiveness" outside of specific services and roles that require it. It does demand an inordinate "attention to detail" due to the fact lives are on the line, whether that be on ships, planes or just normal activities.
This "attention to detail" is also seen among other professions, but seldom in software outside of NASA/avionics/medical/etc...
Linus has laid out guidelines and expectations which while demanding, can be followed. If you can't test your code nor follow a checklist, then maybe you shouldn't submit patches to him. It's the military equivalent of passing barracks/quarters inspections, it should be elementary for professionals.
Don't make it a strawman, as if the only alternative to treating each others like assholes is to be passive-aggressive backstabbers. Even the military has those, the Navy calls them SWOs :P.
Even Linus doesn't make that claim directly, instead he says that the other cultures you might try to enforce lead eventually to the culture of PC passive-aggressive backstabbing, not that those are the only two options.
I do agree that the Linux kernel has a much higher need for attention to detail than many other user-space projects. But I don't think every such detail-oriented project necessarily requires this type of culture. How is the environment developing SQLite, or Postgres?
And even if we say detail-oriented projects need alpha personalities to enforce these checklists, where does that leave the majority of open-source development?