Hi Dave, Take a look at this thread - http://old.nabble.com/Best-Practices-for-Multithreaded-use-of-HttpClient-(with-Cookies)--td27345690.html - I think that may help.
I think, in your case, you're going to need to grab a new HttpClient instance from the ThreadSafeClientConnectionManager if you want truly unique cookie stores for the same target host:port. We have a similar use case, and have gone down this route. Thanks, Sam On 17 February 2010 18:51, Dave Irving <dave.irv...@bankofamerica.com> wrote: > > Hi, > > I'm using HttpClient (4) to talk to some web services. I'd like lots of > requests to be sent in parallel, so I'm using > ThreadSafeClientConnectionManager along with a high per-host total > connection thresholds. > > All requests thru this client will be to the same host / port (and most of > the time to the same URL). > So the result (without playing with anything) is that cookies get shared > across all connections. > > The quick question is: > > - Is there any way to tie cookies to the _connection_ that they originated > in (and not be shared outside this). I.e. supplying a HttpContext from the > "outside" wont work here. Transparent to the caller, I just want cookies to > only be used on the connection they originated on. > > > If you're wondering 'why' - the longer question is: > > The server is doing NTLM (spring security) and uses an HttpSession to manage > the state of the authentication process. So, we get a session id sent back > as a cookie. Thats the only cookie ever needed, and ever cared about. > However, if we have N connections, they all end up sharing the sessionid > cookie - and obviously things blow up (out of step authentication sequences > etc). > I dont want to have to supply an HttpContext from the "outside" (caller), as > there isn't (and shouldn't be) any way to tell one call apart from another. > (All connections will be authenticated with the same credentials - its just > that each connection needs to be authenticated exactly once.) > > If I could somehow make cookies only apply (be scoped to) a _connection_ > this would all work like a treat. > > (NTLM is the real problem here I think, but unfortunately no way out of > that... For now) > -- > View this message in context: > http://old.nabble.com/Using-HttpClient-%284%29-with-%27per-connection%27-state-tp27628555p27628555.html > Sent from the HttpClient-User mailing list archive at Nabble.com. > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: httpclient-users-unsubscr...@hc.apache.org > For additional commands, e-mail: httpclient-users-h...@hc.apache.org > > --------------------------------------------------------------------- To unsubscribe, e-mail: httpclient-users-unsubscr...@hc.apache.org For additional commands, e-mail: httpclient-users-h...@hc.apache.org