The answer for why this needs to be attached to the space station is because it's making use of Dextre, the very large/expensive/awesome robotic arm attached to the ISS. The initial experiments may also likely need a human finely controlling the robotic arm or conducting extra-vehicular activities to set things up. There's some more details in this article.

It also wasn't mentioned in the summary, but a big part of why this is so challenging is that the tech is ultimately intended for satellites which weren't immediately designed to be refueled. There's a -lot- of old satellites out there with their fuel supplies winding down, and this could be potentially useful for quite a few of them.

You don't even need to look outside NASA to see ridiculous spending to compare to. The same House appropriations bill with the $431M JWST cut includes $2B for the Space Launch System (SLS) and $1B for the Orion/MPCV capsule. The SLS is basically Congress's mandate to NASA to build a heavy-lift rocket out of Shuttle-legacy components capable of competing with SpaceX's Falcon Heavy rocket. The $2B is only for the first year of SLS funding, for a rocket which isn't expected to have its first launch until 2017 or later. Mind that this is for a rocket that NASA didn't even want in the first place.

The amount being spent here seems to be a whole lot, until you consider how much is going to be poured down the "back-up insurance plan" with the SLS program just in case the commercial spaceflight approach doesn't work. I've heard estimates of about $3-4 billion being spent just on that one program, something that still has yet to even be figured out in terms of who is even going to build it in the first place.

It's kind of terrifying when one realizes that the combined budget for SLS and the MPCV capsule is $2.5B/year, and it's expected to be at least 6 years before it's ready for first launch. You could buy quite a few SpaceX's for that much money.

The actual cost numbers don't match your assertion: Mir cost about $4.3B to build, while the ISS cost around $100B. (ISS's pressurized volume is only about twice Mir's)

If a self-driven car is able to pass a normal driver's test (perhaps even including natural-language processing of the DMV employee's requests), do you think it should be allowed to drive? What's if it's able to pass several driver's tests, with a success rate significantly higher than the average current driver?

Curiously enough, there used to be a Caltech project class based on pretty much exactly that, although it's unfortunately no longer offered:

http://www.its.caltech.edu/~musiclab/

As a bit of trivia, Caltech alum Bill Gross actually ended up founding GNP Audio based on an engineering project he did as a student. He later went on to co-found, like, a gajillion other companies.

I'm by no means an expert, but my understanding is that there's a few different advantages L2 has over LLO:

* Lower delta-V to reach it from LEO (3.43 km/s vs. 4.04 km/s), and -much- lower delta-V to go from there to Earth escape orbit (0.14 km/s vs. 1.4 km/s). This makes it much more practical for sending missions/probes to Mars, the outer planets, or just about anywhere else in the solar system.
* In that likely case that you're using hydrogen/oxygen propellant, boil-off is going to be your primary long-time storage concern. In LEO (and presumably lower orbit) you not only have to worry about shielding a depot from the Sun, but also have to worry about shielding the thermal emissions from a nearby constantly-moving terrestrial body. If you're in EML2, all you need is a sun shield to keep the temperature down.
* I suspect it's much more difficult to dock with a constantly-moving target in lunar orbit than with a more stationary target at a Lagrange point, both in terms of actual maneuvers and mission scheduling.
* There's substantial gravitational anomalies in the Moon, adding stationkeeping costs for maintaining a consistent lunar orbit.

Absolutely. Remember that development costs tend to be very important when it comes to rockets. For example, the recently-cancelled heavy-lift Ares V rocket NASA was building was projected to cost at least $32B to develop (ignoring operations costs). This was for a rocket with 188mt capacity. By comparison, SpaceX recently announced a smaller rocket (53mt capacity) which will launch at $100M/flight starting in 2013. Instead of spending $32B to develop a bigger 188mt rocket, NASA could instead spend that money to launch fuel and payloads on 320 Falcon Heavies (16,960mt total payload). This of course ignores the greater economics of scale that could be obtained if you were launching a rocket that many times.

In an optimistic scenario, the Ares V would launch once or twice per year. If you assumed that the Ares V launched twice per year and was completely free to operate (which is false, as it actually would've cost billions more to operate), it would take 45 years before the Ares V would have launched more payload than you could've launched spending the same money on Falcon Heavies.

L1, L2, and L3 (the ones in line with the primary and secondary bodies) are dynamically unstable. It's not like you can park there. L4 and L5 points are much better because they are dynamically stable points, however nobody talks of placing a fuel depot there.

Actually, assuming you're talking about a hydrogen/oxygen fuel depot, you'll have a few pounds of propellant boil-off every day (out of several tons total). You can redirect the boil-off for station-keeping, and it pretty much meets the requirements for station-keeping at L2. There's more details in this ULA publication on depot architectures:

http://www.ulalaunch.com/site/docs/publications/AffordableExplorationArchitecture2009.pdf

Actually, not only is it not politically sexy, but it's outright politically dangerous. Having fuel depots allows you to use existing rockets for exploration beyond low-Earth orbit, alleviating the need to develop heavy-lift rockets. A number of politically-powerful congressional districts (and congressmen) are heavily banked on NASA building a heavy-lift rocket from Shuttle-legacy components, while that isn't the case for fuel depots. I predict it won't be long before this particular effort is squashed by Congress, perhaps even outright banning it like they did with the TransHab inflatable modules.

Huge not done. Mars doesn't have enough atmosphere for the chutes, they'd impact more than land. At the same time it's got enough atmosphere to make thrusters very difficult.

Keep in mind though that SpaceX is already planning a thruster-only (no parachutes) system for returning from Earth orbit. I wonder if SpaceX is plotting something along the lines of supersonic retropropulsion for the descent to Mars.

That's what gives me pause... My gut reaction is to think this is too big of a job for one company, but Musk seems genuinely intent on this goal, and seems to be marking all the early steps toward that goal. (Heavy lift? Check. Man-rated? Check...) Even so, that's just a start. They're going to have to step up their current development trend by an order of magnitude, at least, in order to reach Mars, and that's a tall order for such a short timespan.

Actually, just as a thought experiment, here's my guess at Elon Musk's to-do list for the next 10-20 years before he'll be able to start sending people (including himself) and supplies on one-way trips to Mars:

• rocket capable of launching crew to orbit: done, already launched 2 Falcon 9 rockets
• landing capsule: Done, with their Dragon capsule sent into orbit and brought back for water landing last year. The heat shield and parachute system are apparently have much more capability than needed for mere return from Earth orbit, although not known yet how they'd deal with Mars.
• heavier cargo-launching rocket: Falcon Heavy currently under development with first launch in 2013
• in-space docking/assembly: SpaceX will be gaining experiencing in some parts of this with Dragon and ISS, but will need more
• in-space restartable propulsion stage: Raptor hydrogen/oxygen stage under development
• propulsive landing system: under development, many shared elements with the launch abort system currently scheduled for testing in the next couple of years
• in-space habitat: Could use a Bigelow habitat. Could also potentially bring the center stage of a Falcon Heavy all the way to orbit and convert it into a habitat. For periods of high radiation, can shelter in water-shielded area.
• surface habitat: It seems like adapting one of Bigelow's lunar surface designs would be the best option here.

What am I missing?

In TFA, he doesn't mention a return trip. Is that intentional? A one way trip to mars makes a lot of sense.

Musk has stated on a number of occasions that he plans to retire on Mars, so he expects to make a one-way trip himself. Elon Musk is 39 years old now, so in about 20 years he'll be looking to retire. He'll probably have time to groom a younger successor by then to head the company, and I wouldn't be surprised to see him on the very first one-way Mars colonization trip himself.