Reusable rockets promise much easier testing because you should usually get them back, and you can debug as you go rather than having to get everything perfect the first time.
Supplying fuel for a Mars expedition from the lunar surface is often suggested, but it's hard to make it pay off - Moon bases are expensive, and just buying more rockets to launch fuel from Earth is relatively cheap.
Speaking of photography, while the Apollo 8 crew shot hundreds of photos, there was one that got everybody's attention: a blue-and-white Earth rising over a gray moonscape.
Sure, there were hopes that Constellation's systems could later be adapted to support more ambitious goals. But Apollo had those hopes, too. It didn't work in 1970, and it wasn't going to work in 2020.
The key virtue of orbital assembly is that it eliminates the tight connection between the size of the expedition and the size of the rockets used to launch it.
The original specifications for Apollo navigation called for the ability to fly a complete mission, including a lunar landing, with no help from Earth - none, not even voice communications.
The Orion capsule uses an escape system quite like that of the Apollo spacecraft in the 1960s and 70s: an 'escape tower' containing a solid-fuel rocket that will pull it up and away from Ares I in a pinch.
Trying to build a spaceship by making an aeroplane fly faster and higher is like trying to build an aeroplane by making locomotives faster and lighter - with a lot of effort, perhaps you could get something that more or less works, but it really isn't the right way to proceed.
Whether solid rockets are more or less likely to fail than liquid-fuel rockets is debatable. More serious, though, is that when they do fail, it's usually violent and spectacular.
The communications delays between Earth and Mars can be half an hour or more, so the people on the ground can't participate minute by minute in Mars surface activities.
Sometimes a malfunctioning test setup actually gives the tested system a chance to show what it can do in an unrehearsed emergency. During a test of an Apollo escape system in the 1960s, the escape system successfully got the capsule clear of a malfunctioning test rocket.
Past experience, on the shuttle and the Titan rockets, suggests that large multi-segment solid rockets have a probability of failure of 0.5 to 1 per cent.
One of the headaches of high-tech test programmes is having to debug the test arrangements before you can start debugging the things you're trying to test.
On the technical side, Apollo 8 was mainly a test flight for the Saturn V and the Apollo spacecraft. The main spacecraft system that needed testing on a real lunar flight was the onboard navigation system.
NASA has never had a problem finding capable people to be astronauts. NASA's problem was, and still is, finding ways to cut the list of capable applicants down to a manageable length.
My one concern is that when money gets tight, it's easy to cut R&D funding that isn't tied to a specific project - look at what's happened to NASA's aviation research.
Large solid rockets have never been a very good way to build launchers that might have crews on top, especially because of the problems in getting the crew away from a failing launcher.
It's true that Apollo 10's lander was overweight. Late in the craft's development, it became clear that its ballooning weight was endangering the whole mission.
In the long run, it's impossible to make progress without sometimes having setbacks, although people who get lucky on their first attempt sometimes forget this.
In the first few years, it was at least plausible to come in in the morning and read all the Usenet traffic that had come in, and 15 minutes later be off doing something useful.
