A surprising difference between Git and Mercurial... I have some repositories with lots of images and small videos. They're not huge individually, but the count is high, so the repositories are pretty big. That's not what Git and Mercurial were designed to handle, but whatever, I've been using Mercurial in this fashion for years, and it's been pretty smooth.

For various reasons, and notably because Mercurial does a bad job handling renaming and moving of large files, I was thinking of switching to Git. But! Git can't pull in small revision chunks! In Mercurial, if I want to clone a repository and get only the first changeset, I can do that.

hg clone -r 0 /path/to/repo

Then I can pull all the changes up to a given revision, too.

hg pull -r 5
hg pull -r 6
hg pull -r 7

That is useful when the revisions are large. If my internet signal is spotty, then pulling one revision at a time helps avoid errors. It also limits the amount of memory the server and client need. Some of my repositories are much larger in size than the system memory is on my server or client. Plus I can use a bash loop to grab the revisions one at a time. This kind of thing is rare. I only use it when I haven't done a pull for a long time, or when I'm installing on a new machine. But at those times, it's very nice.

There really isn't a good drop-in substitute in Git that I have found. You can make a bundle and SFTP that, if you like, and I think that's great. But being able to do everything via HTTP/HTTPS is simpler and easier and more versatile.

For now, I'll have to leave my large repositories on Mercurial as they are. There may be other good options, but I don't know of any. I want distributed. I don't necessarily need version control, but version control is nice because I get free backups out of it.