Not to cause any arguments, but this is sort a standard discussion that gets brought up periodically and I was wondering if there has been any "softening" of the attitudes against an "in place" upgrade, or movement to not having to dump and restore for upgrades.
I am aware that this is a difficult problem and I understand that if there is a radical restructuring of the database then a dump/restore is justified, but wouldn't it be a laudable goal to *not* require this with each new release? Can't we use some release as a standard who's binary format "shall not be changed." I know the arguments about "predicting the future," and all, but standards and stability are important too. I'm not saying it should never ever change or never ever require a dump/restore, but make it, as policy, difficult to get past the group and the norm not to require d/r. The issue is that as disks get bigger and bigger, databases get bigger and bigger, and this process becomes more and more onerous. If you haven't noticed, data transmission speeds are not accelerating at the rate disk space is growing. I am currently building a project that will have a huge number of records, 1/2tb of data. I can't see how I would ever be able to upgrade PostgreSQL on this system. ---------------------------(end of broadcast)--------------------------- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match