>how does the connector set/use jvmroute? i assume there is >some sort of >cookie magic going on, but after a brief tour of the code, >it's not obvious >to me how this jvmroute stuff works.
the jvmroute is computed when sessionid is created : sessionid = f(entropy) + jvmroute >it seems to me that tomcat (any version) shouldn't have to >have anything to >do with setting jvmroute. if tomcat didn't set the jvmroute, you'll have to determine on connector : 1) that a session has been created and get the sessionid, which is hidden into http reply headers (at least in ajp12/ajp13) 2) Associate that session id with the ajp worker and keep it. 3) Each time you get a request, you have to get the worker for that session and route the request to it 4) You also have to handle the session timeout on connector side to remove then when outdated, and there you could be stucked (may be not today with Costin trick of sending http request to web-server ;) >couldn't the connector be solely >responsible for >this? the connector could add a cookie to the response, then read this >cookie on subsequent requests, and then you wouldn't have to >add stuff to >servlet containers to handle this. or, i could be totally off base :) Could we add another cookie in http reply ? Spec compliant ? What to be used when you have to use URL encoded instead since cookies are not available ? -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
