SolidWallOfCode commented on issue #7603: URL: https://github.com/apache/trafficserver/issues/7603#issuecomment-809425928
There is a more subtle issue here, which also involves #7629. If the destruction of the `HttpSM` is delayed because of the cache write, so is the next transaction on that inbound connection. Currently I/O completion for the cache write is decoupled from transaction finish, which means a new transaction can start before that I/O completes. Clearly this isn't an issue for a client abort, but I don't think that information is retained in a way useful in this situation. This is related to the original reason for the response data sink support, to be able to keep a tunnel up even if the user agent drops. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
