The ability to use Windows in principle was likened to being able to also use a raw Turing machine; i.e. that it could have worked but would have been completely inappropriate.

I think that Unix-alikes are going to dominate any sort of high-performance computing; and I think it's more than just the network effects of the current user base. The scriptability and composability that permeates Unix-alikes is essential for automating tasks across thousands of nodes. That's not to say that it's impossible in Windows, but it's not as much in Windows' culture. Being able to ssh to any node, occasionally draw graphics remotely, all these things are hugely important. The main advantage of Windows, point-and-click administration and lots of existing software, are not advantages when you can't point-and-click thousands of machines and you have to write all your software anyway.

Not so sure I agree with that in general. What about the LHC? That is just a tool, too. Is that too not worth mentioning?

I did not say that tools are never noteworthy. They are important but they are also a means to an end. The LHC is hardly an average run of the mill tool; it is rather exceptionary. So I'm not sure why you think it applies "in the general case."

Well, because you claimed pretty strongly this:

"Tools are tools. Whether its a hammer or a Scientific Linux Cluster the significant factor is the user, the carpenter or the physicist"

I agree with you that tools are a means to an end and that obviously the LHC is a lot more important that the choice of OS. I just think you were a bit too dismissive of tools in your post, though perhaps that wasn't your intent. If someone is going to rah-rah a tool, in this case Linux, they should be pointing out what role it played and what about its use was essential, not just that it was used.