solidwallofcode / trafficserver Goto Github PK

Walt Karas notes there is a potential issue here because the RFC requires the host provided by the user agent but the header is computed after remap which may alter that host. We may need a configuration option which causes the original host to be preserved for use later. This is related to but not exactly the same as pristine host header.

The original NetHandler setup (NetHandler::startNetEvent) uses an event to do the real initialization. I presume this date from before the thread initialization cleanup work when it was the only way to guarantee the initialization would happen after the event loop started. That should be fixed, and configure_for_thread should be moved to be directly in the thread initialization continuation.

Also, there's no need at all to read the global configurations in every thread. Maybe just once, at the same place the thread spawn continuation is set.

The config update callbacks - these could be changed so that rather than having a separate callback for every thread, there is one callback that sends events to all the ET_NET threads.