Best Practice with node_revisions table?

Hi,

I'm writing some code that could one day make it into a module so I'd like to make sure I'm following best practices. Can anyone please help me out with the node_revisions table? Here's what (I think) I know thus far:

• There will be one node record and 1 or potentially more node_revisions records related on nid.
• There will be only be one node_revisions record for a given node when revision tracking is off.
• node_revisions.vid is the primary key and hence will be unique.
• node_revisions.timestamp contains an integer that corresponds to the the this revision was last updated.

Correct so far?

So these are my questions:

• Can I assume that the node_revisions record with the latest timestamp is the most current?
• Or should I assume that the node_revisions record with the largest vid is the most current?
• If neither of the prior two how can I tell which node_revisions record is the most current for a node?
• What is the node_revisions.log field used for? Mine are always blank?
• Where do I set revision tracking in the UI? Do I do it on a per content-type basis? Globally? Other?
• And finally, is there anything else I should know about working with the node_revisions using raw SQL?

Thanks in advance!