DO NOT REPLY TO THIS EMAIL, BUT PLEASE POST YOUR BUG RELATED COMMENTS THROUGH THE WEB INTERFACE AVAILABLE AT <http://nagoya.apache.org/bugzilla/show_bug.cgi?id=15368>. ANY REPLY MADE TO THIS MESSAGE WILL NOT BE COLLECTED AND INSERTED IN THE BUG DATABASE.
http://nagoya.apache.org/bugzilla/show_bug.cgi?id=15368 [PATCH] Expire Continuations ------- Additional Comments From [EMAIL PROTECTED] 2002-12-17 08:27 ------- Excellent work, Michael, thanks! I think the runtime complexity of the code can be substantially lowered if we assume that all the continuations have the same expiration period, which cannot be changed on a per-continuation basis. With this assumption in mind, we observe that in a tree of continuations the parent continuation will expire first. We can also observe that in a list of sibling continuations, the expiration time increases from left to right (assuming the sibling continuations are added always at the end of the list). The last observation is not as important as the first one, which we can use to optimize the algorithm you propose for continuation expiration. With this simplification, the new removeContinuation() method in ContinuationManagerImpl becomes: public void removeContinuation(WebContinuation wk) { idToWebCont.remove(wk.getId()); expirations.remove(wk); wk.invalidate(); } The code above requires a new method invalidate() in WebContinuation: public void invalidate() { for (int i = 0, size = children.size(); i < size; i++) { WebContinuation wkchild = (WebContinuation)children.get(i); wkchild.parentContinuation = null; } } The WebContinuation class could be even further simplified by removing the timeToLive instance variable. Another thing I'd like to ask about this patch is whether we can have the scheduler be woken up on demand. Since we use a sorted list to store continuations based on their expiration time, and no new continuations can be inserted at the front of the list, we know exactly when the next continuation is going to expire. We don't need to wait 30 minutes to have the scheduler wake up and expire the continuations. Can we tell the scheduler at runtime when to wake up? Thanks, Ovidiu --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, email: [EMAIL PROTECTED]
