axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Dobri Kitipov (JIRA)" <>
Subject [jira] Commented: (AXIS2-4288) Support HttpState object association with a client and use it when invoking httpClient.executeMethod(...)
Date Thu, 02 Apr 2009 10:36:13 GMT


Dobri Kitipov commented on AXIS2-4288:

Hi Amila,

1) Yes it is HttpConstants. I am sorry for the confusion. We use some wrappers overhere, so
I just copied a sample which is not pure Axis2 one. Here is the pure Axis2 one:

client.getOptions().setProperty(HttpConstants.CACHED_HTTP_STATE, new HttpState()); 

2) Yes, right. If there is no httpState (it is null) then the one associated with the HttpClient
will be used. So at end the code will behave as it did before the change. Thus it is backward

3) Yes, right. This is more simpler and thus more reliable way to keep the HttpState. I originally
thought about this, but finally decided that it is more natural to keep the HttpState into
the ServiceContext. I saw that cookies are kept into the ServiceContext (see AbstractHTTPSender.obtainHTTPHeaderInformation)
and since they are part of the HttpState it seemed to me that this is the prefered place.
Please, provide me with your comments (e.g. cons and pros) about keeping the HttpState into
Anyway, I will update the patches and the HttpState will be kept into the MessageContext.


> Support HttpState object association with a client and use it when invoking httpClient.executeMethod(...)
> ---------------------------------------------------------------------------------------------------------
>                 Key: AXIS2-4288
>                 URL:
>             Project: Axis 2.0 (Axis2)
>          Issue Type: New Feature
>          Components: kernel, transports
>            Reporter: Dobri Kitipov
>         Attachments: AbstractHTTPSender_AXIS2-4288_1.patch, HTTPConstants_AXIS2-4288_1.patch,
> Hi all,
> I am opening a new JIRA related to the HttpState association with a given Axis client.
> The use case
> I want to point out a concrete use case I have. I want to reuse a HttpClient among different
instances of a clients which are executed in different threads. Every client can make several
invocations. The clients can call different Web Services (WSs) deployed at different hosts/servers.
As a consequence every client may need to provide different authentication Credentials and
may need to support transport sessions using Cookies.
> Both Credentials and Cookies are part of the HttpState. As a result the HttpState associated
with the HttpClient that is reused cannot be reused that easily in the scenario described
above. Credentials are associated with a given host, port, realm and authentication schema
(defines the AuthScope object used as a key for the credentials Map part of the HttpState).
> Following is an excerpt from the AbstractHTTPSender#setAuthenticationInfo:
> creds = new UsernamePasswordCredentials(username, password);
> agent.getState().setCredentials(new AuthScope(host, port, realm), creds);
> Credentials are kept into a Map and could be identified uniquely from client to client
(thread to thread), but this Map is exposed to all clients which reuse the HttpClient which
is not a good idea.
> The situation with the other member of the HttpState (i.e. Cookies) is similar. When
we have *different* client instances (configured to use cookies options.setManageSession(true))
calling one and the same WS's operation then the effect is that both are sharing one JSESSIONID.
> The proposal
> The proposal is based on my question posted at [2]
> The idea is to provide the capability to specify/associate a separate HttpState with
every client and still reuse one and the same HttpClient. What you just need is to pass it
as a parameter to the HttpClient#executeMethod.
> I decided that the HttpState should be kept into the ServiceContext. I did all changes
needed in Axis2 kernel (in fact they are really few) and added the possibility to use a separate
HttpState and invoke HttpClient#executeMethod passing it as a parameter. The changes keep
the kernel backward compatible. I did and several tests and it looks good.
> Please, give me your comments. Do you like this extension? If so I can provide you with
the changes and finally we can agree on committing them into the kernel. 
> I have the patches but we are using a little bit older version of axis2 .I need some
time to apply the changes to the trunk version. Then I will provide them. 
> Although I am a WS committer I can not commit anything since I am still waiting my company's
lawyers to check all aspects of this. 
> Thank you,
> Dobri
> [1]
> [2]

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message