Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.misc -> Re: changing the isolation level
On Sun, 04 Feb 2007, lehmannmapson_at_cnm.de wrote:
>>> Think of a script, that generates invoices montly. Lets say >>> you have a average volume of $1000 dollar. So if a run of the >>> script in the next month just returns a volume of $200 you can >>> be pretty sure, that something has gone wrong. But if you >>> cannot see this until you confirmed all the stuff, then it is >>> too late to revoke it. >> Not at all. Surely you don't think Oracle eBusiness Suite >> financials and >> the many banks using Oracle are overwhelmingly choosing Oracle if the >> architecture can't handle this. There are many possible solutions.
When?
> Don't try to think all people have just that one simple business case
> where everything can be managed within on transaction by one process!
> Think of heterogenous systems where one specialised system can change
> data but it cannot commit until another specialiced system reads the
> data and performs whatever with it.
What? Then stage is somewhere, have the first session read and give the data to the 2nd process. There is never a reason to prematurely commit data.
>>> Sybase supports all levels of isolation, so when I read that >>> Oracle doesn't allow a certain isolation level it just sounds >>> like a lame excuse for a missing feature. >> You are incorrect. They do not exist because they are >> unnecessary. You need to stop trying to force Oracle into your >> current mental model (Sybase-centric) and learn how Oracle >> works. Start with Tom Kyte's books and the docs at tahiti.oracle.com.
No, READ_UNCOMMITTED is around because of inferior architectures. There is never a business case to read some other process's uncommitted data.
-- Galen BoyerReceived on Sun Feb 04 2007 - 10:54:22 CST