In the late 1990s, the game industry was starting to tackle "best practices". Basic techniques common in the rest of the software world (and mostly taken for granted now) weren't getting much traction in the game business. I often heard fellow game developers complain that those practices wouldn't work for games, because the industry was "just different." A decade later, I think they were right--but not in the way they thought they were.
When will Duke Nukem Forever be released? When it's done. DNF is the most popular whipping boy for a game development schedule gone horribly wrong, but there have been plenty of similar examples (including the original Unreal). Game developers are notorious for scope creep and gold-plating, as well as massive mid-stream changes in design and architecture. Why is that?
I've become convinced it's not solely due to "unprofessionalism" or lack of effective project management. You see, if you're creating a business application, or a device driver, or control software, you can specify the requirements, from which you can create a design, from which you can create code, which you can trace back to requirements. It's not hard to write effective, measurable requirements for this type of software, which means it's not hard to justify or drop a given feature with objectivity.
But games have a damnable, hidden, unwritten Requirement Zero:
Thy Game Shall Be Fun.
Requirement Zero is a real killer. Because it's not objectively measurable, it's not easy to manage. You can pick on Daikatana because of its schedule slips, or because you don't like its designer, but the bottom line was it just wasn't a fun game, and for more reasons than the obvious "sidekicks get squished by doors" bugs.
I once had a conversation with Danielle Bunten Berry (of M.U.L.E. fame) and a colleague about a particular game. It was gorgeous, free of technical glitches, pushed the envelope on technology, and had a decent storyline. But my colleague summed it up this way: "There's a big hole where the fun should be".
You see, users don't expect their word processor, or their spreadsheet, or the software that dispenses their soda at the local mini-mart to be fun. It's sufficient for it to be "fun-neutral": so long as it isn't so badly designed, defective, or poorly-performing as to be "un-fun", it's acceptable.
You can create a game that completely satisfies the spec, is totally free of bugs, finishes on-time, stays within its budget, and has a flawless marketing program--and it can still be a miserable failure if it isn't fun.
Yes, games ARE different.