Re: Row locking and serializability

Thanks for including the response from inside Oracle. My thanks also to everyone who participated. I think I learned a lot, too.

I managed to get some feedback from Joe Celko (incompuserve's caseforum). Joe was on the SQL92 committee.

He says the three problems (dirty read, fuzzy read, and phantom read) were intended
to illustrate non-serializability, not to define it. He also said that the SQL group decided
that, because there was another group defining transaction standards beyween systems,
it would leave the final standard as to what contstitutes serializability in the other group's
purview.

The idea is that, if both groups defined a standard for serializability, and they differed,
that would just weaken the standards as a whole.

I apogize in advance if I've made any mistakes in repeating what Joe said. This business of "correct"
standards is a whole lot subtler than writing correct code.

Regards,

    Dave Cressey mailto:david_at_dcressey.com

iggy_fernandez_at_my-deja.com wrote in message <7mbtro$dv2$1_at_nnrp1.deja.com>...
>Thanks to all who participated in this debate.
Received on Mon Jul 12 1999 - 05:38:09 CDT