Re: MVCC implementation

David Cressey wrote:
> "Cimode" <cimode_at_hotmail.com> wrote in message
> news:1153498280.454510.326480_at_h48g2000cwc.googlegroups.com...
> Quote:
>
> MVCC is just another buzz word for a set of principle applied to
> guarantee a limited integrity during transactions server level
> interactions. One example is row versioning. Implemented in ORACLE as
> Read Consistency and SQL2005 as READ_COMMITED_SNA¨PSHOT. The
> principle is to make sure that all transactions always have access to
> the last version of commited data so that the SELECT's always read
> commited versions of the data, no matter what transaction are currently
> on hold.
>
> It is light years from being a decent relational implementation even if
> it claims a noble goal which is integrity preservation. I do not see
> how an implementation could compensate for loss of integrity at
> transactionnal level when it ignores the most important : integrity at
> definition level
>
> End quote.
>
> What is the connection, if any, between SQL2005 READ_COMMITTED_SNAPSHOT
> and Snapshot transactions as implemented in DEC Rdb version 1 in 1984? Are
> these different concepts?
> Are they similar concept independently conceived? Or is there a track from
> the DEC RDb implementation to the SQL 2005 definition?
I do not know about DEC RDb implementation but I know about Read Consistency implemented by ORACLE. All I know is that on RM standpoint, it's an illusion to hope implement and guarantee integrity at operation time (in this case transaction timing).

> Can someone who knows both SQL 2005 and DEC Rdb's history answer this?
Received on Sat Jul 22 2006 - 20:18:04 CEST