Couple of show stoppers

APACHE 2.0 STATUS:						-*-text-*-

Last modified at [$Date: 2001/04/18 20:56:03 $]

Last modified at [$Date: 2001/04/19 13:27:58 $]

Release:

      to make it agree with the operation of the StartServers

      directive.

    * Doug M. reports that there is a there is a problem with both

      threaded and prefork mpms children not properly handling

      SIGTERM which results in the server taking a long time to

      shutdown.

      Status: Paul Reder is investigating...

    * A design problem with the scoreboard can cause the threaded 

      MPM to get in a state where it will no longer serve requests.

      When MaxRequestsPerChild is hit, a threaded process will begin

      allowing it's idle worker threads to exit. The child process

      may have one (or a few) threads serving really long responses

      over slow client connections, which will prevent the child

      process from exiting. The problem is that the

      perform_idle_server_maintenance code will NOT start a new

      process to replace the dying child process until the dying child 

      has exited because the new child needs to use the old childs

      space in the scoreboard. The scoreboard and

      perform_idle_server_maintenance need to be redesigned.

      Status: Several proposals discussed on new-httpd (April 16, 2001)

RELEASE NON-SHOWSTOPPERS BUT WOULD BE REAL NICE TO WRAP THESE UP:

    * Performance: Get SINGLE_LISTENER_UNSERIALIZED_ACCEPT