While I don't agree with you that version control doesn't work well for Word, Excel, etc, I'd say the problem there is the format isn't built for version control.
VC works great for source code because it's all line based with (either by convention or required by the grammar) one statement per line. Diffing is pretty trivial in these cases and meaningful.
Binary formats are a whole different ballgame. Even XML-based formats are problematic. My view is that XML documents require a VCS built specifically for diffing that kind of data as line based will run a lot of false positives with white-space.
As another commentor said: you can always get your stuff back with Dropbox. I'm not sure how much better you can do than this without actually changing the program that created the data.
From what I've been told, Office docs are made up of objects accessible via an API. So theoretically one could build a version control tool that works effectively for Office docs.
There are .NET interop assemblies[1], for working with Office documents, and they require Office to be installed to work. Outside of that, you're stuck rolling your own Office-document-parser.
There is a built-in "track changes" mechanism, though I don't know if all Office types handle it. Word does, at least. Though that's pretty basic.
True, perfect merging doesn't exist. I just think that replacing one file with two files that have different names isn't the best solution to letting me deal with the problem.
Well, obviously we want some copy (hopefully, the best copy) of the file to be available at its original name without having to use a Dropbox client or visit the Dropbox website. So I guess what you're looking for is an additional interface (website or client) which would allow easy choosing or even diffing following a conflict? I'd certainly support that, as I've had to diff myself in the past.
VC works great for source code because it's all line based with (either by convention or required by the grammar) one statement per line. Diffing is pretty trivial in these cases and meaningful.
Binary formats are a whole different ballgame. Even XML-based formats are problematic. My view is that XML documents require a VCS built specifically for diffing that kind of data as line based will run a lot of false positives with white-space.
As another commentor said: you can always get your stuff back with Dropbox. I'm not sure how much better you can do than this without actually changing the program that created the data.