Copying vs of cloning?

[Jens Alfke]

On 2 Oct '07, at 10:15 PM, Greg Lindahl wrote:

> Like all database-like things, it needs something more than a simple  
> disk copy for a clean backup.

And "hg push" is exactly such a something. It's already aware of the  
concurrency issues, so it works if the source repo is being modified  
at that moment. It's also incremental, of course, only copying the  
data that it needs to. (Any general-purpose backup is very unlikely to  
be truly incremental when backing up Hg repos, because they contain  
large files that get appended to. Any file-based backup mechanism is  
going to end up copying the same data over and over. But an 'hg push'  
only copies the new data.)

However, there is one reason why a normal file-copy-based backup would  
be a good idea: in case some terrible bug turns up in Hg that destroys  
repositories (or appends corrupt revisions to them). Hg seems quite  
stable, but it's certainly less stable than the filesystem itself :)

--Jens