Perhaps I shouldn’t use the word “backup” here. I certainly do not mean to imply that using a DVCS means you don’t need to have a backup strategy.

But the central principle of any backup strategy is that your chances of losing data go down when you have more copies of it. Some folks (including me) do feel better about their version control data when using a DVCS because there are multiple instances of the repository. At SourceGear we have a central server with RAID and ECC RAM and redundant power supplies and regular backups with offsite storage and stuff like that. But we get a little extra security knowing that there are dozens more live copies of the whole repository regularly being used on other machines. Any of them could become the central repository on short notice.