> rule number 1 of software: don't lose your user's data
That doesn't scale to the scope of Gmail. The appropriate precautions to preserve old medical records differ in kind from those appropriate for old shopping lists.
I'd like my Google account to have a "Backup URL" field, that I could set to <git:tux.example.org> or <venti:glenda.example.com>. Google would undertake to push all "my" data to that URL, to pull it back as needed to recover their systems, and to publish the format it was stored in. I'd be responsible for the rest.
You can have Gmail forward all your incoming mail (or even better, just the non-spam mail). Worst case you can then import that back into Gmail. I just have a copy kept on my computer and its various backups.
Now, I can already run "webcheckout $url" and get a git repository for any of the websites I run, as well as all of the wikis and blogs of customers at my startup, Branchable.com.
That is implemented via a proto-RFC called the "rel=vcs microformat", which simply adds a little data to web pages to indicate where the underlaying data can be cloned from. <http://kitenet.net/~joey/rfc/rel-vcs/>;
That doesn't scale to the scope of Gmail. The appropriate precautions to preserve old medical records differ in kind from those appropriate for old shopping lists.
I'd like my Google account to have a "Backup URL" field, that I could set to <git:tux.example.org> or <venti:glenda.example.com>. Google would undertake to push all "my" data to that URL, to pull it back as needed to recover their systems, and to publish the format it was stored in. I'd be responsible for the rest.
Venti: http://plan9.bell-labs.com/sys/doc/venti/venti.html or http://plan9.bell-labs.com/sys/doc/venti/venti.pdf