> > Really?! Once again: WAL records give you *physical* address of tuples
> > (both heap and index ones!) to be removed and size of log to read
> > records from is not comparable with size of data files.
>
> You sure? With our current approach of dumping data pages into the WAL
> on first change since checkpoint (and doing so again after each
> checkpoint) it's not too difficult to devise scenarios where the WAL log
> is *larger* than the affected datafiles ... and can't be truncated until
> someone commits.
Yes, but note mine "size of log to read records from" - each log record
has pointer to previous record made by same transaction: rollback must
not read entire log file to get all records of specific transaction.
> >> Why not? Seems to me that establishing implicit savepoints is just a
> >> user-interface issue; you can do it, or not do it, regardless of the
> >> underlying mechanism.
>
> > Implicit savepoints are setted by server automatically before each
> > query execution - you wouldn't use transaction IDs for this.
>
> If the user asked you to, I don't see why not.
Example of one of implicit savepoint usage: skipping duplicate key insertion.
Using transaction IDs when someone want to insert a few thousand records?
Vadim