On Mon, 2013-01-07 at 15:00 +0100, Bronislav Gabrhelik wrote: > Oleg, > > thanks for pointing this out. I have realized that it is my fault after > that I verified that the proxy is transparent. > > The problem is missing slash at the end of URI while constructing the > HttpGet(). > > "http://auth-digest2.brno.local"; > vs. > "http://auth-digest2.brno.local/"; > > The question is if user shouldn't be forced to use the ending slash in all > cases even if proxy is not used. I understand that HttpClient tries to be > tolerant, but the use of API would be more consistent. > > Thanks for your help, > Bronislav >
Yes, the ending slash should be enforced. If it is not the case, this is definitely a bug. Please raise a JIRA for this defect. Oleg > > > 2013/1/7 Oleg Kalnichevski <[email protected]> > > > On Sun, 2013-01-06 at 23:17 +0100, Bronislav Gabrhelik wrote: > > > Hello, > > > > > > > > > I came across the following issue. If I configure the HttpClient with > > > proxy(no-auth) then the digest authentication is failed by the server > > > with 400 Bad request. After short analysis I have found out that the > > > digest authentication request contains an absolute URI instead of a > > > relative URI.The absolute URI is set in the > > > DefaultRequestDirector.execute() by calling > > > of DefaultRequestDirector.rewriteRequestURI() in case the proxy is > > > configured. The parsed log from netmon is in the attachment. > > > > > > > > > Originally I have found the problem with my own Digest Authentication > > > Scheme (using Windows SSPI). After that I tried the build-in Digest > > > auth scheme and I was able to reproduce the issue. I noticed that the > > > request arg in the Schema.Authenticate is RequestWrapper, so i tried > > > to use RequestWrapper.getOriginal() for retrieving of URI, but I had n > > > o success. It returns also an absolute URI. > > > > > > > > > I believe it is a bug in the HttpClient. Besides that I would like to > > > address this problem in our DigestAuthScheme before it will be fixed > > > in the HttpClient. > > > > Bronislav > > > > Similar issue was reported in the past. I think HttpClient's behavior is > > correct and is compliant with the requirements of the HTTP > > specification. The problem is on the proxy side. For details please see: > > > > https://issues.apache.org/jira/browse/HTTPCLIENT-1071 > > > > Feel free to report this problem to developers of the proxy server. > > > > Oleg > > > > > > > > > Is there some consistent way how how to get relative URI from the > > > HttpRequest or HttpContext passed into authenticate() method? I know I > > > can use the URI class to get a relative path, but it seems just like a > > > workaround. > > > > > > > > > Environment: > > > JRE 1.6.0_23, Windows 7 SP1 > > > dependencies: > > > commons-codec-1.6.jar > > > commons-logging-1.1.1.jar > > > fluent-hc-4.2.2.jar > > > httpclient-4.2.2.jar > > > httpclient-cache-4.2.2.jar > > > httpcore-4.2.3.jar > > > httpmime-4.2.2.jar > > > > > > > > > Thanks, > > > Bronislav Gabrhelik > > > > > > --------------------------------------------------------------------- > > > To unsubscribe, e-mail: [email protected] > > > For additional commands, e-mail: [email protected] > > > > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [email protected] > > For additional commands, e-mail: [email protected] > > > > --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
