The fact that ACS/Tcl (the original from Ars Digita) is effectively dead has bearing on this too. Just because aD has chosen to move ACS to Java doesn't mean I necessarily want to.
There's also the matter of breadth. I'm unaware of any proprietary solutions that provide the sheer range of functionality that ACS/OpenACS does.
Database-wise, Oracle still offers more functionality than PostgreSQL. (This is not a dig against PostgreSQL. In time, I expect that PG will become a serious competitor to Oracle. PG already is a serious competitor on the low end.) At list prices (who pays retail?), Oracle Standard Edition is $15,000 per CPU. That's not much money for a company that really needs Oracle's features. For a company that already has an Oracle infrastructure, it's really not much money, compared to incorporating an "alien" database into the administration/backup/recovery scheme.
