Not a real DBA, but a MySQL guy
Taking MySQL and PostgreSQL as my examples, the only way it would make a difference is on full-table scans, and only if the columns in front of it are of variable length (VARCHAR's etc). Anytime you are resorting to full-table scans, your database is going to be slow (after about 50 columns, full-table scans loose out completely to indexed data) so the itty bit of speed you would stand to gain is most likely not worth it.
But if you have an index for that column, it should be stored in a different location from the actual data, and should make it not matter one little bit where the column is located.
From a design perspective however, the most important columns typically go first (at least in the schemas I build), but that's a guideline without any real reason for it.
|