[whichdan]

I'm curious - has anyone here had experience doing this with a few million rows of data? Even a few revisions each means the table would grow rapidly. You'd also (probably) need three operations for each insert:

1. Insert the new row

2. Find the previous row

3. Set previous row to `active`=0

Are there performance tricks to keep this reasonable? I'd also wonder how to modify the table itself (dropping/adding columns) when there would potentially be selects + inserts + updates running 100% of the time - but I realize that's borderline out of scope of the OP's question.

[blcArmadillo]

Not sure about the other questions but in regards to changing the actual table structure. If you're dropping columns you'd first make the changes to your code base so that it no longer relies on those columns. If you have long running queries you'd just have to wait till the old ones using those columns are finished and then you could change the DB. And you'd do it in just the opposite order when you wanted to add a column.