Re: Multiple update problem

Hi Mungo,

Surely blocking means waiting for the release of the lock. When you update primary key twice, then commit, only the last value will be the "real" value of the record. I believe this is a part of "transaction-based" rdbms ideology.

Hope this helps,
Igor

Mungo Henning wrote:
>
> Thanks for the reply Thomas, but I'm still curious...
>
> Thomas Kyte wrote:
> [multiple update scenario snipped]
> ?
> ? ?I'm thinking of the following (separate) scenarios: the past is to the
> ? ?left
> ? ?of the line, the future to the right:
> ? ?
> ? ? UserA:update UserB:insert UserA:commit UserB:commit
> ?
> ? If userA updates 10 to 90, UserB's insert will BLOCK on the unique index entry.
> When userA performs the update, the index will be modified. But I was
> under the impression that for read-consistency userB should not be
> aware that another user is amending the table until the amendments
> are committed? I'm quite willing to believe that it's my impression of
> Oracle that's wrong, but can you set me straight please?
>
> When you say "BLOCK" do you mean that the insert will be suspended
> pending the lock? What would happen if userA then altered the same
> record (whose value is now ninety) to eighty without committing - would
> the block be released?
>
> TIA
> Mungo Henning

-- 
Igor Sereda,
ITC, Russia