Re: tracking changes on a table through ODBC application

Wrong. The reason dirty reads exist in those databases is to make up for short comings. (The argument 10 million Frenchman can't be wrong is a rather species and not logical argument.) Dirty reads are not a feature they are a failing. Oracle itself does do dirty reads; it just does not allow you the user to do them. (eg trying to insert a row with a repeated primary key where the repeat was inserted in an uncommitted transaction that you cannot see. Error will be raised because the primary key violation. Case of Oracle "looking behind the scenes - doing a dirty read".) Dirty reads are part of ISAM and flat file technology when transactions did not exist.

Of course, the are "part of the standard", the standard is a compromise by a bunch of companies and not a dictate by any particular company.

But being a "db dude" I would have thought you would have been more informed.
Jim

"db dude" <db_guy_2002_at_yahoo.com> wrote in message news:f4a8da28.0307261313.2c6b65da_at_posting.google.com...
> >
> >
> > We are talking Oracle here. Oracle doesn't have 'dirty reads' like the
> > Mickeysoft toys do.
> >
> >
>
> Both DB2 and SQL Server support all four *standard* isolation levels.
> Even open source databases like PostgreSQL are also getting there, so
> its quite suprising that oracle is lagging behind in not supporting
> these isolation levels. Even going upto the extent of claiming that
> "dirty reads" are undesireable when they are pretty much standardized
> and exist in other databases.
>
> Here is what Oracle documentation says about changing isolation level
> to "uncommitted read":
>
> "READ UNCOMMITTED
> Oracle never permits "dirty reads." Although some other database
> products use this undesirable technique to improve thoughput, it is
> not required for high throughput with Oracle."
>
> It should actually read:
>
> "READ UNCOMMITTED
> Oracle never permits "dirty reads." since one of our developers way
> back in prehistoric times hardcoded certain attributes that prevents
> us from supporting uncommitted reads in a straight forward manner
> and/or without breaking existing functionality."
>
> Sucks just like Oracle's JDBC driver. Its definitely *undesireable*
Received on Sat Jul 26 2003 - 16:31:02 CDT