Re: Requirements for update languages?
Date: Sat, 9 Nov 2002 21:39:09 -0500
Message-ID: <aIjz9.136$WO2.74315821_at_radon.golden.net>
"Alfredo Novoa" <alfredo_at_nospam_ncs.es> wrote in message
news:3dcd0cd7.5337444_at_news.wanadoo.es...
> On Fri, 08 Nov 2002 18:02:56 +0100,
> =?windows-1252?Q?Leandro_Guimar=E3es_Faria_Corsetti_Dutra?=
> <lgcdutra_at_terra.com.br> wrote:
>
>
> > OTOH there is this new idea from Date that transactions are an
> >unnecessary complication because we should just have atomic statements >
>that make all the transaction's updates in parallel.
>
> Can you give references?
>
> In the Third Manifesto, transaction support is still one of the
> prescriptions.
I suspect Leandro overstated his point. Date does not support the idea that the dbms should relax some constraints until commit. Transactions in SQL evolved such that they deal with logical constraint issues as well as concurrency issues.
Deferring some constraints until commit worked around difficulties with cyclical references. Date now proposes statements that specify updates to multiple relations in a single statement to deal with cyclical references.
I believe Date and Darwen prescribe transactions for their concurrency and rollback properties but not as a deferral point for relaxing integrity constraints. Received on Sun Nov 10 2002 - 03:39:09 CET