Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!think.com!snorkelwacker.mit.edu!world!epstein
From: epstein@world.std.com (Richard W Epstein)
Newsgroups: comp.databases
Subject: Re: Oracle CURSORS useless ?
Message-ID: <1991Jun27.230131.2358@world.std.com>
Date: 27 Jun 91 23:01:31 GMT
Organization: The World Public Access UNIX, Brookline, MA
Lines: 19

In article <25731@well.sf.ca.us> gatynen@well.sf.ca.us (Gerard Tynen) writes:
>We couldn't think of any reason to force a close on all cursors just
>because you do a commit.  I think that what happened is that early 
[deleted...]

A commit by definition makes any updates,inserts, and deleted permanent
and releases locks. After this happens, where would your cursor point?
Since you released your locks, you cannot be sure the records or
even the table is still there or that you have access to them.

Ingres has a command called 'savepoint' and allows you to rollback
to the last savepoint without aborting the entire transaction.
I don't think ANSI SQL has any provision for this.

The real need for this seems to be in syncing with external devices/files
which are not under the DBMS transaction control. Transactions loose
some of their usefulness when all parts of the update cannot rollback.

--Richard W. Epstein		epstein@world.std.com