John Keymer, November 28, 2018 - 8:39 am UTC
Perfect, thank-you very much. I presume you give thought to pctfree only when you expect updates to the row to be significantly larger than when it was created (so as to avoid row migration in the future)?
Just for my own understanding, why did the other parameters matter "in the past"? I.e. what was it that was so slow that would make you want to allocate a load of space up front? I also could never understand maxextents - why would I want to limit the size of my table??
November 29, 2018 - 12:30 am UTC
pctfree is important for me because most tables are insert only, or insert/delete, and yet we blow 10% of the space for updates that might never occur.
We used to bludgeon some internal dictionary tables (uet$/fet$) to manage everyone's extents, so choosing how many to grab initially and how often to grab them was important. Locally managed tspaces pretty much solved that.
maxextents was a means for stopping an unexpected/undesired explosion in growth of a particular table.