Re: Newbie question about db normalization theory: redundant keys OK?

From: David Cressey <>
Date: Fri, 28 Dec 2007 13:41:12 GMT
Message-ID: <Y17dj.80$oh5.3_at_trndny08>

"paul c" <> wrote in message news:wVScj.21602$vd4.14144_at_pd7urf1no...

> Until sometime in the 1990's, all the big-name databases used quite
> primitive locking methods based on physical coordinates, none of them
> could lock, say, a Customer row where Custid = 2532. (Note that as
> above these mechanisms were quite separate from physical memory or disk
> synchronization ones).

What I worked with in that time frame was DEC Rdb/VMS. I guess you would have to say that Rdb/VMS was not a "big name database". However I think you could say that Digital was a "big name computer company" in 1985.

As I recall Rdb, you are right. It didn't have "content based" in the sense that you described above. It did, however, have a type of "insert and delete" locking on a table, that would allow you to detect a deadlock condition where, at the time, Oracle would charge ahead and yield a phantom update.

I used to have an SQL script for just such a comparison. It used the "Free seats" case I outlined above, with one exception: the free seats were computed by the total seats assigned to the flight, minus the count of seat reservations committed (a row count). A trigger or constraint could detect and prevent overbooking. That script would run differently on Rdb/VMS and on Oracle RDBMS, which didn't have insert locking.

With regard to concurrency being a myth, I have to say this: not at the physical level. Overlapping disk seeks, disk transfers, as CPU cycles makes a big difference in throughput.
"Serializable" and "Serialized" are not quite the same thing, even though they are logically equivalent.

> Gray called this a "predicate lock" but it's not
> clear that he had relational predicates in mind. He was right when he
> said that a logical lock manager could be a black hole. I knew one
> developer, former professor at that, who may have been going mad before
> he was told to enhance the predicate lock manager of an obscure product
> to support ranges. This may have tipped him over the edge. The last
> time I bumped into him, he couldn't tarry for long as the secret police
> were hot on his tail.

He's not a victim of MI5 persecution, is he? Received on Fri Dec 28 2007 - 14:41:12 CET

Original text of this message