On Mon, 2007-06-11 at 19:07 +0200, Olaf Sebelin wrote:
> Hi,
>
>
> Oleg Kalnichevski <[EMAIL PROTECTED]> schrieb am 04.08.2006 um 16:42:
>
> > On Fri, 2006-08-04 at 19:54 +0530, Adalbert Wysocki wrote:
> > > Thanks Oleg! It helps but...
> > >
> > > Do you recommend using a HostConfiguration configured with my
> > > particular Protocol (and the socket factory) instead of registering
> > > the protocol with Protocol.registerProtocol(...)?
> > >
> > > Or
> > >
> > > Using as parameter to the PostMethod a relative URI without any
> > > scheme and starting from the first / ?
> > >
> >
> > Aldo,
> > It very much depends upon your design consideration. If you want a
> > Protocol instance to apply globally, it should be registered using
> > Protocol#registerProtocol method. If, however, you want a Protocol
> > instance to apply to a specific host, then you should be using a
> > custom HostConfiguration parameter.
> >
> > > Is there a milestone to solve this issue?
> > > Thanks,
> > >
> >
> > This issue simply highlights deficiencies of the 3.x API. There's not
> > much that can be done about it without breaking the API compatibility.
> >
>
> Isn't it just a small patch (against 3.1-rc1) like this?
>
> +++ HttpClient.java 2007-06-11 18:53:52.000000000 +0200
> @@ -384,9 +384,14 @@ public class HttpClient {
> if (hostconfig == defaulthostconfig || uri.isAbsoluteURI()) {
> // make a deep copy of the host defaults
> hostconfig = (HostConfiguration) hostconfig.clone();
> if (uri.isAbsoluteURI()) {
> - hostconfig.setHost(uri);
> + if (hostconfig.getProtocol()!=null &&
> +
> hostconfig.getProtocol().getScheme().equals(uri.getScheme())) {
> + hostconfig.setHost(uri.getHost(), uri.getPort(),
> hostconfig.getProtocol());
> + } else {
> + hostconfig.setHost(uri);
> + }
> }
> }
>
> HttpMethodDirector methodDirector = new HttpMethodDirector(
>
>
> I think, the problem is, that an individual Protocol hanlder is
> overwritten by hostconfig.setHost(uri). So the Solution would be just
> to use another setHost()-Method of hostConfig in case an individual
> Protocol is used. Does this break compatibility?
>
Olaf,
Have you tried running the test cases against the SVN trunk with the
patch applied?
Oleg
>
> Regards,
> Olaf
>
> ---------------------------------------------------------------------
> 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]
