[Twisted-Python] Re: Mixmaster Architecture

Mark Evans 2s7l3way02 at sneakemail.com
Thu Nov 20 15:29:47 MST 2003


Thank you Moshe Zadka.  What you're saying is that TCP
connections can trigger ephemeral UDP pseudo-connections.

The server communicates over just one UDP port, and data
flow is bidirectional at all points in the chain.  I'm not
sure whether these facts affect the suggestion here.
The mixmaster opens its own assigned UDP port to receive
data from the server, as well as sending data to the
server over another UDP port.

> Make up your mind, telnet or raw? a Line-based protocol?

You're right, I should.  Here is the scoop.  For manual
device tests I'm using PuTTY in its "raw" mode.  That is
the origin of "raw."  Basically the telnet sessions are
line-based command/response activity.  The devices don't
utilize the full telnet protocol, just a small subset.

The terms "client" and "server" may be mixed up, from a
networking standpoint.  These are just conventions in the
legacy software, right or wrong.  The original authors
called the Java app a server, and the devices clients. 
However the devices supply information that the Java app
requests of them.  So the actual roles are reversed.

Normally, when talking to devices that know its protocol,
the system is very simple.  Here's how it works without a
mixmaster script.  The server initiates communications. 
The server broadcats a public query packet, and devices
respond with their names and technical features.  The
server constructs a list of known devices.  From there, the
server invokes various device capabilities.  When finished,
the server shuts down the devices.

Why the mixmaster?  Many devices offer identical
capabilites, but don't know the server's protocol.  The
mixmaster allows the server to use them.

My apologies for confusing terminology and my thanks for
your answers.

Mark




More information about the Twisted-Python mailing list