On Wed, Dec 28, 2005 at 05:27:53AM +0800, Syan Tan wrote: > also , transactions aren't just for atomicity, there > designed for equivalence of a serial schedule of > transactions when interleaving the input of concurrent > transactions You are correct. I should have said that the way we use them currently they are not useful for detecting concurrent updates.
Whether we should use them differently is another matter. It does seem attractive to not have to fiddle with xmin. OTOH it would mean we would need *quite* a few concurrently open connections. It would also mean we can only take a lot less advantage of read-only connections. Changing this should be left to very careful consideration. 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
