Re: [liblo] how to connect to a liblo TCP OSC server?

[Stephen S. <rad...@gm...>]

Hi Roland,


On Tue, Aug 4, 2020 at 1:36 PM Aigner Roland <rol...@fh...>
wrote:

> However I struggle to figure out how I would connect a client to this
> server, i.e. what the code should look like for a client receiving the data
> sent by this server.
>

I'm not sure if your concept of client and server is mixed up here?  In OSC
it is the client sending data to the server.

Documentation and sample code leave me scratching my head, I tried to
> figure out by checking the library source code, but so far I failed to
> understand. From the fact that lo_send_message_from requires a target
> address I infer that I would have to iterate through all connected clients
> and sent to the respective sockets manually.
>

You can get a target address by creating one, if you know the destination,
or you can get the sender's address from inside a message handler:

    lo_address a = lo_message_get_source(data);

    lo_send(a, "/hi", "i", 3);


> But how do I know when a client connection gets accepted?
>

Who knows?  The server knows because a handler is being executed.  The
client only knows whether the send failed.



> So, summarizing, my two remaining questions are
>
>    1. how do I connect a client to the server?
>
>
Your code (which I snipped from the message, apologies) is correct.  You
just create a server and specify LO_TCP for its protocol, and a client
specifying LO_TCP for its protocol.  Then the client should be able to make
a connection to the server and send it a message.



>
>
>    1. how do I know what clients are currently connected that I would
>    have to send my packages to?
>
>
Hm, I don't think liblo provides a way to enumerate the current client
connections.. could be interesting, but it's not too usual to want to
broadcast to a random client.  Usually messages from the server are limited
to _responses_ that are sent during message handlers, and in that case you
can get the response destination using lo_message_get_source as I outlined
above.

Note that if your *client* wants to receive a response... actually it
cannot, it needs to set up a *server* to receive the response, and use
lo_send_from, specifying that server.  That's how the destination server
can find where to send responses back to.  So for bidirectional messaging,
each side has a server.

If you need to broadcast a message to multiple servers, you can look into
multicast or broadcast messaging, but that's for UDP.

As for sending a message to all currently connected clients, I'm not sure..
I'll have to think on how to do that and whether it's a good idea.
Theoretically it is possible.

Keep in mind that OSC is typically a UDP protocol and is designed around
one-way messaging.  Liblo incidentally supports responses, but it's common
for UDP protocols to be unidirectional.
On the TCP side, often clients will make a connection, send a message,
*maybe* wait for a response, and then close the connection.  Long-lived
connections are more rarely used; but if they are, they shouldn't *depend*
on the connection state.

Hope this helps,
Steve