we say that when any transaction is rolled-back that means it would be re-started again , so lets say if T2 reads a data item from T1 , perform some computation on it and then writes it back onto disk and after some T1 fails , so its a dirty read problem and we say that this schedule is not recoverable , so then if T1 is rolled back i.e. it is restarted again so then will T1 execute in a serial fashion from start till end or whether there would be still some concurrent execution between T1 and T2 .
After roll-back will the value of data item be set to its initial value or since T1 is restarted again it would be set to the value as computed by T1 again ?