Linux-Hams archive - January 1997: Re: Request for Comments
Re: Request for Comments
tyy@tnet.ro
Fri, 3 Jan 1997 21:53:17 +1100 (EST)
Matthias Welwarsky typed:
>
> Jonathan Naylor wrote:
> > How about removing digi-peating from the kernel and putting into user
> > space ?
>
> I definitely like this idea. But the main question that arises with this
> is, how to interface the userspace-digipeater with the kernel-ax25. What
> API should be used? Hm, I'd like to have something like this:
>
> Get a socket(), bind() it to an address that will be the address to
> digipeat via, then signal to the kernel that we would like to digipeat
> ourselves, maybe by setsockopt(). Then listen() on that socket.
axdigi already does this sort of thing. What happens is you make a
SOCK_PACKET type of socket. You get all the bytes that the device driver
sees and the interface is is from address.
Now this is a bit of a hassle because the actual user level program has to
go through the AX.25 packet itself to see if the packet is supposed to go
through it.
A nicer thing would be something that you describe. At the moment when we
bind a port, we are asking for the kernel to give us a handle on any
connections that have the destination adress we specify. Perhaps we need
something that says give me any stream where we are the next system in line.
This would be great for ROSE operation too.
> Hm, seems to be difficult realizing this via BSD-Sockets ;-)
No really, just a change on what test you use to say "this is mine".
- Craig vk2xlz
--
// /\ | | | Craig Small VK2XLZ @home: yteg.efbyig@telepar.com.br
||==||===|==|=| [44.136.13.17] @play: fqki.dvqa@benjerry.co.uk
\\ \/ | | | @work: pxxomck@unmc.edu