hc-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Erik Bunn (JIRA)" <j...@apache.org>
Subject [jira] [Created] (HTTPCORE-346) NIO: HttpAsyncService should couple IOSession and HttpContext set/getAttribute()
Date Wed, 24 Jul 2013 17:11:49 GMT
Erik Bunn created HTTPCORE-346:
----------------------------------

             Summary: NIO: HttpAsyncService should couple IOSession and HttpContext set/getAttribute()
                 Key: HTTPCORE-346
                 URL: https://issues.apache.org/jira/browse/HTTPCORE-346
             Project: HttpComponents HttpCore
          Issue Type: Wish
          Components: HttpCore NIO
    Affects Versions: 4.3-beta2, 4.2.4
            Reporter: Erik Bunn
            Priority: Minor


Under httpcore-4.1, I wrote code that attached identifying information (unique certificate
id) to the IOSession attribute context in SSL verification (where we have access to IOSession
only, ), and utilized that information at the request processing stage (where we have access
to HttpContext). HttpContext.getAttribute() proxied to its parent context, eventually reaching
the IOSession attributes.

The refactoring for 4.2.x async processing has decoupled the HttpContext object available
in the HttpAsyncRequestHandler implementation methods from the IOSession. Looking at HttpAsyncService.processRequest(),
you can see that it provides HttpAsyncRequestHandler.handle() and .processRequest() with the
(always empty) BasicHttpContext contained in its state.context. 

At both locations, HttpAsyncService does have access to the NHttpServerConnection containing
the previously established SessionHttpContext seen by IOSession. 
1) Could this context be sent to the HttpAsyncRequestHandler, instead of state.context? 
2) Alternatively, would it be possible to initialize state.context with the contents of the

3) Alternatively, would it be possible to modularize HttpAsyncService.state creation to allow
extending and customizing this behaviour?


There may be security or nio object reuse implications that I am not immediately seeing, and
which might invalidate connecting the attribute storage in this way. I would appreciate your
input if that is the case. If not, I would claim this is a bug in the recent nio HttpContext
refactoring.  (Marked as minor under the assumption few people need to work with paired TLS
certs.)






--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org


Mime
View raw message