[wyoung2]

I daresay there is no DVCS appropriate for many large files that change. You want something more like a classic non-distributed VCS so you're not cloning the entire history of huge files.

Worst is when they're data-compressed so you're not only pulling all historical versions but also most every byte of each historical version.

I wrote an article for the Fossil docs covering this, but the core experiment can be run against any VCS:

https://fossil-scm.org/home/doc/trunk/www/image-format-vs-re...