On Mon, 19 Dec 2005, Mark Womack wrote: | - prerender the message before the lock?
I'd say that this is a rather obvious performance improvement, and will IIUC also alleviate the deadlock-situations, since any objects that might cause deadlock by being locked as a side effect of rendering would be locked without any other locks held, preserving the linear lock-order. Also, I think I saw that NDC uses HashTable keyed on Thread for its operation, which I find super-bad if I understand the usage correctly. What about ThreadLocal, in which case the NDC.remove() invocation on thread exit wouldn't be needed either?! ThreadLocal is "since 1.2", and since 1.4 and recent it is much faster than contended synchs around a Map. .. or at least use ConcurrentMap? But ThreadLocal is rather obvious here, I believe? Am I missing something? I could do such a patch.. Regards, Endre. --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
