On Sat, Dec 24, 2005 at 09:41:24AM +0800, Syan Tan wrote:

> I was wondering if xmin checking was necessary if you had fine grained commit
> intervals;
That won't help any.

> so commit at every  update
> when an item is changed or added in the gui view of the medical record,
Richard et al do not want that.

> then reopen a transaction for the next commit. this would mean
> a user who has a too-late update due to another user earlier updating the same
> item and committng, will be notified with a transaction 
> abort.
No, why ? Transactions only ensure all-or-nothing, eg the
atomicity of an action. They are not at all concerned about
the *content*. To detect intermittent content changes we
need xmin. Another possibility would be to lock the tables
right from the beginning but we don't want that.

Karsten
-- 
GPG key ID E4071346 @ wwwkeys.pgp.net
E167 67FD A291 2BEA 73BD  4537 78B9 A9F9 E407 1346


_______________________________________________
Gnumed-devel mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/gnumed-devel

Reply via email to