His points may be valid but his language is that of a 15 year old. It's a wonder he has a job.
Anyway, here are the problems:
1. All Mac users should be using Homebrew now - that ship has sailed. Fink and Ports have advantages, but not in the typical situations.
2. Cost? Get over it. Any employee's salary plus expenses is an order or two of magnitude higher than the hardware he or she is working on.
3. Staging environments solve the OS problem. Just have staging and integration be on the exact same OS as the production system. This is the only realistic solution. Even if you're running Ubuntu 10.04 desktop and the production system is 10.04 server, there are still going to be not inconsequential differences.
4. In a Python world, this just doesn't matter. PIP should be used to manage libraries and freeze versions.
ThinkPad is the best computer out there aside from the MBP - the only problem is with the OS typically found on the ThinkPad. My time isn't worth saving $100 with Windows - sorry.
It scales well but the pool of people with the experience and inclination to run a major Internet site using .NET is small and more importantly, getting smaller. Nobody graduating with a CS degree in the States right now is excited to work with MS - it just isn't sexy right now.
Sexy isn't everything, but it's not nothing either.
I love the idea of slabs too large to move outside of the rooms in which they'll be placed. I also love that the slabs will have extra space for people in the future to write something to those in the far future.
Nonetheless, I think 'danger' is a bad thing to express. Any group able to access these tunnels will be advanced enough to measure radioactivity and understand the consequences of exposure to it. Some sort of pastoral Planet of the Apes group would simply walk over it none the worse for wear.
The simplest thing to do is to state what's inside (U238 or whatever), how much of there is, etc... Don't seal up the rooms with the details of the area, let them be accessible so what's inside can be updated. U238 should be understandable for at least 2-3k years and then they can worry about it.
I like the idea of making decisions for seven generations ahead (a few hundred years). The point of that way of thinking is not just to think ahead at all (7 generations vs. only this generation), but also not to be overburdened by the too distant future (people a thousand years from now might be excited to find all that easily mined radioactive material). We can't let perfection be the enemy of the good. A solution for a few hundred years is totally fine.
If you have that much time on your hands, just use github as your main repository source and Redmine+Gitosis as a secondary? Everybody with a git clone has a full copy of everything.
s3 is a representation of 3 geographically diverse mirrored stores. In addition, versioning is available to address the issue of a rogue credentialed service deleting objects.
Your point is still valid. Vendor and application level redundancy is still desirable if one can afford it.