Re: tcp-www.cc

[Tom Henderson <tomh@raptor.CS.Berkeley.EDU>]

On Wed, 4 Feb 1998, Deniz Gurkan wrote:

> Dear ns-users,
> 
> I have included the tcp-www.cc into ns, and recompiled it successfully.
> But, I couldn't understand the features of tcp-www.cc exactly. 
> 
> For example, what does the callback function do? In the README file for
> the implementation written by Emile Sahouria it is written that: "The
> derived class Agent/TCP/WWW provides a callback when a TCP connection
> completes."  Is the implementation imitates a persistent connection HTTP
> with this feature? 

When this module was written, there was no callback routine in TCP to let
the Tcl source know that the transfer was complete.  tcp-www.cc simply
modified the recv() method of TCP to callback to the Tcl level.  Since
that time, I've noticed that callback (the "done" proc) has made it into
TCP, so tcp-www.cc probably isn't needed anymore, but I haven't been 
updating the code recently.
 
> 
> Secondly, I couldn't understand why the connections are handled as one-way
> in the output trace files. Since the 3-way handshake takes place between
> the server and the client, there should be a two way traffic? Maybe it is
> because of simplicity constraints. But, please let me know if so.. Because
> I will probably need the whole communication details in my simulations.
> 
> 
Again, when this was written a year ago, there was no FullTCP.  

We wrote this module to provide background traffic for our foreground TCP
(file transfer) studies; it is not meant to be used to directly study HTTP
dynamics.  If you are interested in that, you probably should consider the
code being developed by Bay Networks.

Tom