Re: [cpp-netlib-devel] Implemented Rudimentary Unit Test inbranchhttp_integration

[Dean M. C. B. <dmb...@fr...>]

Hi Divye,

These are great observations, thanks for taking the time. Let me try and
address your questions inline below.


> -----Original Message-----
> From: cpp...@li... 
> [mailto:cpp...@li...] On 
> Behalf Of Divye Kapoor
> Sent: Sunday, April 20, 2008 1:53 AM
> To: C++ Networking Library Developers Mailing List
> Subject: Re: [cpp-netlib-devel] Implemented Rudimentary Unit 
> Test inbranchhttp_integration
> 
> Hi,
> I was just going through the test code and I found 2 
> difficulties:- 1. The http 1.0 implementation tests don't 
> seem to fail if they are run with no network connection (or 
> didn't I wait long enough for them to time out?) 

I actually haven't tried this myself -- and admittedly, had not been
thinking about this when my focus was getting tests in to make sure that
the implementation compiles first. However, you should be getting a
timeout error but depending on your OS, the amount of time it takes may
not simply be 30 seconds -- although I think in Linux this is the OS
default.

> 2. There 
> doesn't seem to be any timeout or connection status feedback 
> available with the client object. For example, how does a 
> client come to know that a connection seems stalled and needs 
> a restart (for example, an intelligent download agent?). 

You're right -- this information should be put into the 'get' or 'post'
methods, most probably refactored into its own (private) function inside
the client. This can be a simple Asio Timeout handler using an approach
given by Chris Kohlhoff in one of our conversations on the Boost C++
developers list. I can implement the changes later.

> Where does this kind of info fit in the basic_message 
> hierarchy?  (I didn't find any appropriate functions or tests 
> for this kind of situation)
> 

It actually doesn't it in the basic_message hierarchy because timeouts
are not part of the message abstraction. This rather is a protocol
client specific thing and should be handled at that layer.

> Have I looked at the right places for these things or have I 
> missed them completely?
> 

You've looked at the tests so I'd say you've been looking at the right
place(s). :)

Thanks Divye!

--
Dean Michael Berris
Software Engineer, Friendster, Inc.
<dmb...@fr...>
+639287291459