I can travel indefinitely, as a military doc, with a week worth of casual civvies, a suit, a 'service dress' uniform, working uniforms, study materials, and 'doctor stuff' like whit coat, stethoscope, etc, with a folding garmet bag, a 25" rolling suitcase, and a standard messanger bag for carry on. Volume is about the same as a navy 'seabag' duffle. Tested over four- and six-week trips, doing laundry once a week.
My dad recently did 2 weeks in Japan with a folding garmet bag. A rolling folding garmet is definitely my go-to piece of luggage for city-to-city travel, flying or driving. For travel on ships, I recommend the seabag, as it takes very little room and can double as a backpack in a pinch. Obviously, if you're walking inland, your backpack can double as airport luggage. I tend to use an XL timbuk2 for overnighters.
I once managed to make it through a meeting with a huge, gaping hole in the backside of the pair of pants I had on, caused by a spring poking through a chair. I think the client wondered why I was so attached to the bag I was carrying, but didn't say anything.
Carrying backup stuff is a good idea.
This is all probably great advice when you're talking about being a tourist.
As someone who does business traveling a lot, though, it's really hard to argue some of these points.
I, personally, would rather carry a little extra junk through an airport because I decided to include a backup shirt and slacks in case I spill some soup.
When we're talking about business travel, you're talking about things that affect your career or your own company. I take what I need to be 99% sure I'll be successful at the destination-- and I don't worry so much about the burden. Of course often you can go out and get what you need, but you don't want to have to do that (especially at the last minute) if you don't have to. That extra bit of confidence of feeling prepared can make all the difference to let your mind focus on getting business done to the best of your ability.
All this aside, I think the analogy about software is a great suggestion. You really don't know what you should be taking with you (i.e. features) until you get there (i.e. release it to users).
I liken that to a civil engineer building a bridge remotely and then delivering all the pieces without having ever visited the site himself. He or she would have a much worse idea of the dimensions, concerns, or obstacles-- and might have to re-engineer a lot of what they built.
Of course software is more flexible than that, but it's still a great idea to take a few, and not all, of your "best guess" features to the customer until you get their usage/feedback to build on.