ietf-openproxy
[Top] [All Lists]

Re: Java Proxylet

2001-09-18 09:15:10

Rajnish,

responses inline
On Tuesday 18 September 2001 9:44 pm, you wrote:
Andrew,


1. In 6.7.14,  what is origin server point? please clarify.

6.7.14 ProxyletRequest.getSession

The OPES device can act on behalf of either the client or the origin server. 
If the opes device has activated a proxylet on behalf of the client, then the 
device should use the clients session with the proxy - the assumption here is 
that the opes device is being used within an Enterprise, and has session data 
that is associated with the client (authorised for use by the client).

If the OPES device is acting on behalf of the origin server, then the 
enterprise proxy session data must not be accessible to the proxylet. In 
replacement the proxylet should have access to an origin server related 
session data structure (located on the OPES device). There needs to be data 
security between the different roles that the OPES device can play.

There are issues related to maintaining a session on behalf of an origin 
server that need discussion.

2. Section 6.16.16,

     I think onus of sending error lies with OPES system depending upon the
return values of proxylet. How are we going to use the sendError() api
mentioned in this section.

The return value represents the communication between the proxylet and the 
proxylet engine. The sendError causes the response message to be rewritten as 
an error - it may be better for proxylets to explicitly do this themselves. 

If the correct behaviour of a proxylet at processing point 1 is to respond to 
a request with 404 not found, then the response should be sent as as 404 not 
found, and the proxylet return ProxyletStatus.OK (200 OK). At point 1 the 
proxylet doesnt have access to the response so currently the only option is 
to return an error like status code.

I dont think I've managed to capture quite right the way to build up an error 
response here. Perhaps there should be a method 
HTTPRequest.getResponseOutputStream() returning a suitable output stream 
allowing the proxylet to build its own error message. Thoughts? Perhaps also 
another ProxyletStatus code that indicates that a response is ready to be 
sent back.

There is an ommission from the HTTPProxyletResponse interface in that 
it is impossible to modify the status code and status message - this is 
probably what I had in mind with the method sendError. I think I'll update 
the HTTPProxyletResponse to have accessors and mutators for status code and 
status message.

Thanks for your comments,
Andy

<Prev in Thread] Current Thread [Next in Thread>