Everybody is giving some variation of "there is no excuse" or "well maybe if you're stupid!", but this seems entirely inconsistent with what I see other people actually do.
For example, how many people here have all your email, digital photos, web browser bookmarks, and system configuration in a version control repository? I know some people do (hi Joey!), but most people I know, even most programmers, don't.
The comeback is probably going to be "well, I have backups, and I don't need to track changes to email, or branch my photos, or ...". Great! So let's try to extrapolate from that to when a reasonable person might think that version control is unnecessary:
- a bunch of small, independent pieces of data
- that are inherently temporal, and are rarely edited in-place, in the sense that changes consist primarily of adding new data
- which might be large binary files, a weakness of VCSs
- which might live in a program-specific format (like an SQLite db), for which it would take an extra step (or twelve) to get the data in a format that would actually be useful to track in your VCS
- which are already backed up
That seems like a good starting point.
Of course, I want to track everything in VCS, but due to the above issues, it would take more effort than it's worth. I wish programs I used would store data in formats that made it easy to use a VCS, and I wish VCSs were better about tracking data which isn't just "all text files below a certain folder". (Text files were something Unix got right! Sadly we've got some kind of AT&T-MIT-Microsoft-Xerox hybrid architecture going on, which doesn't let the best of any philosophy really shine.) That's just not the world we live in, for better or worse.