24 Jul
2007
24 Jul
'07
1:45 p.m.
Hi, Mark Morgan wrote:
The race condition is that a multi-threaded server may not finish destroying a resource before the client sends the next request to the resource. In our code, if our client has: send_op() destroy() send_op() The second send_op() succeeds (whereas it should fail) unless we pause between the destroy and the send_op. The reason for this (I believe) is that destroy in Genesis II is slightly heavy-weight in that it needs to (potentially) send out notifications and clean up WS-Naming resolvers.
But if I understand that right, the race condition is in your code (or hosting environment) rather in the messaging itself? Cheers, Michel