Re: New patch for server sockets and datagram (UDP) support.

[Kim F. Storm]

Richard Stallman <address@hidden> writes:

> 
> Given that datagram streams are handled so differently at the user
> level, would it be better to have a different function to open one?
> 
I think my latest proposal for `make-network-process' with wrappers
functions like open-datagram-client addresses this concern.

> 
>     +   DEFVAR_LISP ("network-server-log-function", 
> &Vnetwork_server_log_function,
>     +                doc: /* Function called when accepting a network 
> connecting.
>     + Arguments are SERVER, PROCESS, and MESSAGE, where SERVER is the server 
> process,
>     + PROCESS is the new process for the connection, and MESSAGE is a string. 
>  */);
> 
> Could you please explain more about this?  I am wondering whether
> having a single function to be called for all connections is the right
> interface for the job.
> 
Helmut suggested that I do this on a per-server basis; I'll look into that.

> 
> For consistency, please have two separate functions
> process-datagram-address and set-process-datagram-address.
> The former should just return the current status; the latter
> should set it, like the other set-process-... functions.
> 
Ok, I will do that.

The reason I made just one function is that I was concerned whether
doing
        (setq old-address (process-datagram-address P))
        (set-process-datagram-address P new-address)
in two steps would be atomic.  But since we don't
accept input between those two, they will be atomic.

-- 
Kim F. Storm <address@hidden> http://www.cua.dk