Hello I'm trying to understand if the v3, which my application currently uses, handles Keep-Alive properly - because our client says it does not.
On several requests, using lsof, I can see that httpclient connections in state ESTABLISHED gradually fall back to a CLOSE_WAIT state, and they also stay in this state, unless I explicitely call org.apache.commons.httpclient.MultiThreadedHttpConnectionManager#closeIdleConnections. A question : if Keep-Alive was working properly, should the connections stay in ESTABLISHED forever ? By looking at the source, org.apache.commons.httpclient.HttpMethodBase#shouldCloseConnection looks correct, we are using HTTP/1.1, it implies connections are persistent. However, org.apache.commons.httpclient.HttpConnection#closeIfStale seems to think connections are staled, hence closing them, so it's not reusing them. I don't understand the org.apache.commons.httpclient.HttpConnection#isStale part, with socket.setSoTimeout(1);/inputStream.mark(1);/inputStream.read() : why is the timeout set to 1 and how would it affect the read ? Is the stale checking code broken in v3, breaking connections reuse ? I've read the Connection keep alive strategy chapter at http://hc.apache.org/httpcomponents-client/tutorial/html/connmgmt.html, and I can't figure out if v4 fixes it or not : if v4 assumes it can keep the connection forever, will it still do a stale check ? Thanks for your input laurent -- <a href="http://blog.jdeuxe.info/";>http://blog.jdeuxe.info/ - Java entreprise tips & tricks</a> --------------------------------------------------------------------- To unsubscribe, e-mail: httpclient-users-unsubscr...@hc.apache.org For additional commands, e-mail: httpclient-users-h...@hc.apache.org
