Re: oracle - mysql comparison

"Daniel Morgan" <damorgan_at_x.washington.edu> wrote in message news:1090387704.737703_at_yasure...
> Dan wrote:
>
> > Where was VC wrong?
>
> > - Dan
>
> When you respond to what I wrote rather than what you wrote I wrote I
> will gladly respond.
>
> From where I am reading ... you came into a conversation in the middle
> and are not tracking on the point I have been trying to make which
> relates to point-in-time accuracy.

I think you are missing the point. There is a formal definition and criteria for "point-in-time" accuracy in the face of concurrent manipulation of data. Following from this there is a definition and a criteria for "point-in-time" accuracy that allows for interleaving of transactions, called serializability. Saying "I use multi-versioned read consistency" and a snapshot in time is not necessarily it.

If one gets the definition right, then it becomes much easier for everyone to talk at the same level understanding in terms of "point-in-time" consistency and perhaps even reach consensus on understanding the trade-offs. Your previous posts have given me the impression that you confuse the terms 'serialized' and 'serializable'. If this is indeed the case, then the argument will have a tendency to become circular in related conversations such as concurrency control mechanisms and implementations (locking, timestamp algorthings, optiimistic vs. pessimistic, 2PL, etc). , which it has to a degree, because you are using a premise as a basis of definition that is fundamentally flawed, or at the very least, different. The Oracle Concepts document does have some misleading wording and has a very narrow scope, and if that is what you use as your Bible, then it is not surprising.

> any type of cereal: Not Corn Flakes, Not Cheerios, Not Oat Meal.
> Hope you now have a sense of how I feel reading what you wrote. Hope
> you are thinking "I wrote serial not cereal."

Whatever this means...great.
>
> Daniel Morgan
>
Received on Thu Jul 22 2004 - 12:18:55 CDT