Originally posted by DaPi
"ORA-1555" ah yes . . . . but I have had this work for me . . . . is that because it never needs anything from any RBS released by the commit?
Well, given it a second thought, I guess the probability of getting ORA-1555 with your method is actually very close to 0, provided that full table scan will be executed by your cursor (if the rows are accessed by using index then you are bound to get "snapshot too old", that's for sure)! Although you will be performing "fetch accross commit" you won't get ORA-1555 because each block will be read only once, so there will be no need for rollback data. The only exceptions to this rule would be the blocks in which the reading/deletion of the rows will be interupted by the COMMITs. So when the deletion of the rows will continue after the commit, the block's read consystent immage will be required from the rollback segment. But there will be no problem of obtaining them because that block was only recently written to the rollback, so there will probably be not enough time for it to be overwritten, hence very slight chances of getting ORA-1555.
Jurij Modic
ASCII a stupid question, get a stupid ANSI
24 hours in a day .... 24 beer in a case .... coincidence?