]> granicus.if.org Git - apache/commitdiff
Updating my vote on the worker MPM
authorBrian Pane <brianp@apache.org>
Mon, 29 Apr 2002 06:59:20 +0000 (06:59 +0000)
committerBrian Pane <brianp@apache.org>
Mon, 29 Apr 2002 06:59:20 +0000 (06:59 +0000)
git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/trunk@94853 13f79535-47bb-0310-9956-ffa450edef68

STATUS

diff --git a/STATUS b/STATUS
index 5ccb4be7ef862d28607ee3545efbeeae142803c3..740388d2f07065a9850f38533b70e51b393937ba 100644 (file)
--- a/STATUS
+++ b/STATUS
@@ -1,5 +1,5 @@
 APACHE 2.0 STATUS:                                              -*-text-*-
-Last modified at [$Date: 2002/04/28 01:56:20 $]
+Last modified at [$Date: 2002/04/29 06:59:20 $]
 
 Release:
 
@@ -83,15 +83,9 @@ CURRENT VOTES:
 
     * Make the worker MPM the default MPM for threaded Unix boxes.
       +1:   Justin, Ian, Cliff, BillS
-      -0:   Lars,
-            BrianP (according to the system call profiles, worker's main
-            bottleneck--and the reason why it underperforms compared to
-            prefork--is in mutex ops, and there's no easy way to fix
-            that without turning it into something that isn't really worker
-            any more...I think we need to either switch to leader/follower
-            (and do enough testing and debugging to make the L/F code
-            production-quality) or come up with another design that has
-            less mutex contention than worker)
+      +0:   BrianP (mutex contention is looking better with the
+            latest code, let's continue tuning and testing)
+      -0:   Lars
       -1:   Aaron (premature decision, needs more discussion and bug fixes)
 
 RELEASE NON-SHOWSTOPPERS BUT WOULD BE REAL NICE TO WRAP THESE UP: