"=?ISO-8859-1?Q?H=E5kan Th=F6rngren?=" writes:
> >>>>> "David" == David Kelly <dkelly@hiwaay.net> writes:
> 
> David> FreeBSD's man says:
> 
> Hmm, I had a look in the NetBSD kernel source and as far as I can see
> both SLIP and PPP are there.  FreeBSD apparantly does it in another
> way.  I guess I should have a look at the FreeBSD PPP code as well. 

Actually FreeBSD does it both ways. "ppp" is a user-land program which 
uses the tunX interface. "pppd" is the one using traditional pppX 
devices.

> Accessing the serial port can be done via the serial I/O device, no
> need to bang any hardware.

Right. If one were to go for minimal cost at the expense of software
effort then an enhanced access to the 8530's typically used by Sun would
be good. AX.25 is HDLC synchronous. If one were to emulate the KISS TNC
interface in the kernel then apps (and network stacks) written for the
KISS interface would port easily. There was/is a SoftKISS driver for the
Macintosh which did this. Externally this "tnc" is no more than a tone
encoder/decoder for the Bell 202 tones (1270 and 2270 Hz, I think) plus
a relay, FET, or transistor for the PTT switch.

Otherwise a "mass produced" commercial TNC is in the $75 to $100 range.

> There are other problems to look into, like how to deal with ARP, how
> to make a good API for AX25 (sockets somehow?), timers and possibly
> threading.

Linux has an API for AX.25 via sockets. Lots of debate about making 
AX.25 the same as IP vs. different. If its different then traditional 
apps won't route over it without an additional gateway daemon. If they 
are the same then you can control access via routing and well known 
firewall technology. If they are different then the AX.25 apps can 
deal directly with unique AX.25 issues. If they are the same then 
everything has to be generalized into the existing IP model. Etc.

> The good points of having AX25 outside the kernel (in my opinion) is
> that it makes it easier to port to other UNIX flavours and one does
> not have to mess getting the code into the NetBSD code base.  
> The advantage of having it in the kernel would be to have a tighter
> integration with other protocols and performance (which is not really
> a big issue as outlined above).

Considering there already are interfaces to the kernel's networking then
it makes most sense to me to develop outside the kernel. Much like an
http server. Once you establish the technology then somebody can play
with folding that into kernel space. What is it, "Tux", the Linux kernel
httpd server?


--
David Kelly N4HHE, dkelly@hiwaay.net
=====================================================================
The human mind ordinarily operates at only ten percent of its
capacity -- the rest is overhead for the operating system.