Re: Informix versus oracle

yeah, which makes it really annoying to have to issue a "commit" after every update.... and from what i understood, the implementaion of 9i's "Undo Tablespace" is akin to Informix's Physical Log. >>> "rkusenet" <rkusenet@sympatico.ca> 05/29/03 10:30AM >>> "Brice Avila" <briceavila@hotmail.com> wrote:- > Informix was also working on eliminating the physical log (Oracle's > rollback segment). If I understand correctly, Informix physical log bares little resemblance to Oracle's rollback segement. The physical log stores the pre-updated image of a page only ONCE per checkpoint duration. Once a page is retrieved from the disk for the first time after a checkpoint, a copy of it goes to physical log and it stays there till next checkpoint. All futher updates to that page is done in memory and logical logs only. OTOH Oracle's rollback segment stores the pre image of a block for EVERY update, because it is also used to retrieve pre update image when other users try to query the same row. This is what they call Multiversion Read Consistency.

Comments interspersed Brandt Edwin wrote: > yeah, which makes it really annoying to have to issue a "commit" after > every update.... What? Committing after every update, or for that matter every insert or delete goes against the most basic architecture of Oracle and most other serious RDBMS products with the possible exception of Sybase and SQL Server where they run out of locks rather quickly. Commits in Oracle, as in Informix, are performed when they are logically important. Not for any other reason. If you were taught or told otherwise about Oracle the person doing the instruction was either quoting from somehting they learned more than a decade ago or was just clueless. > <snipped> -- Daniel Morgan http://www.outreach.washington.edu/extinfo/certprog/oad/oad_crs.asp damorgan@x.washington.edu (replace 'x' with a 'u' to reply)