gupta@cullsj.UUCP (Yogesh Gupta) (01/25/89)
In article <4113@hubcap.UUCP>, billwolf@hubcap.clemson.edu (William Thomas Wolfe,2847,) writes:
< From article <670@atanasoff.cs.iastate.edu>, by hascall@atanasoff.cs.iastate.edu (John Hascall):
< [...]
< > Perhaps having a BS in engineering caused my expectations to be unfairly
< > high.
<
< Or perhaps it caused your expectations of CS to be inappropriate.
<
< Engineering does not always require intense theoretical knowledge;
< computer science problems can seem innocuous and yet be very subtle
< and sophisticated. For example, one might conjecture that in a
< database system, the rules
<
< No transaction can read any data object that was not
< written by a committed (complete, permanent) transaction.
<
< and
<
< No two transactions can write to the same data object
< if those two transactions are executing concurrently.
<
< would be sufficient to guarantee database consistency. The
< rules seem innocent enough, but intuition in this case is
< misleading. One transaction could read X and write that value
< to Y, while another transaction reads Y and writes its value to X;
< in any serializable execution of these two transactions, X and Y
< would end up with identical values. However, there is a concurrent
< execution of the two transactions which would leave X and Y with
< different values, without violating either of the above rules.
But there is at least one very successful commercial relational DBMS
product in the market today that does exactly the above, and markets
the "feature" as one of their strengths :-). If only the buyers had
an MSCS :-).
--
Yogesh Gupta | If you think my company will let me
Cullinet Software, Inc. | speak for them, you must be joking.