Suppose I've got a part of schedule as:
....
T1 : Read(X)
T2 : Write(X)
T1 : Read(X)
....

We call this unrepeatable read problem as T1 does not get the same
value of X.
But I've got an objection. Why see it as a problem? Our operations may
want us to make T1 read the value of X overwritten by another
operation (ACID violation).

Then why would any transaction would read the data again, and moreover
would like to see the same data which was already with it in the
buffer.
Doesn't make sense.

On Apr 27, 11:16 am, Ravi <ra.ravi.... (AT) gmail (DOT) com> wrote:
Consider refresh of a web page? Actually refresh from any client that
does not or cannot retain all the selected data.

If you want to read dirty data, go ahead. But there is a reason why
it's called "dirty data".

Ed

"Ravi" <ra.ravi.rav (AT) gmail (DOT) com> wrote


If you want a new value for X, then you want to do it in the context of a
new transaction.

T1 : Read(X)
T2 : Write(X)
T3 : Read(X)

Now T3 can look at the new value of X without raising aconsistency problem.

As to why a transaction might read the same data twice, it's not always
obvious what data an action ends up reading.