[Coco] ka9q was Re: Coco Contiki

[Gene Heskett]

On Friday 29 February 2008, Willard Goosey wrote:
>On Thu, Feb 28, 2008 at 12:34:48AM -0500, Gene Heskett wrote:
>>Keep us posted please Willard.  I have an old copy it someplace in
>>that image of my old maxtor that's currently sitting in /dd/maxtor of
>>the 1GB scsi drive
>
>Well, I did hack on it a little last night, but just to get all the
>details out of the way first:
>
>The two computers in question are a 512K CoCo3, MPI (AFAIK does not
>have strapped *CART), deluxe rs-232, 6809, NitrOS LII 3.2.6; and a
>PS/2 model 70, 386-20, Slackware 7.0 (kernel 2.2.13, sliplogin 2.x)
>
>Terminal emulation on the CoCo to the linux box works fine.  Under
>OS-9, YMODEM is as solid as a rock at 9600, ZMODEM has to drop down to
>4800 to be successful.
>
>Linux has all the proper networking goodies compiled in, it speaks
>tcp/ip over ethernet just fine. The kernel-level slip driver is
>compiled in.

I don't have that ATM, but as I build my own, its only about 15 minutes and a 
reboot away.

>For a "dialer" under OS-9 I'm using a small terminal-emulator called
>"terminal", from Rainbow.  I use it because it doesn't try to hang up
>the line when it exits.  I can use terminal as a regular
>terminal-emulator to log into the Linux box.  Since it doesn't alter
>the device descriptor (baud rate, etc) I'm pretty sure I have the
>regular serial configured correctly.

I'd been using Supercom-2.3 for that, I used it years ago to connect to my 
amiga and conduct speed tests while building rzsz-3.36 for us.  The 
connection then was bulletproof.

>So, I log in with terminal, sliplogin starts, I exit terminal and
>start "net".
>
>As a good first test, net (the ka9q binary) can indeed telnet to
>itself and ftp to itself.  It is, however, very very slow.

You are telnetting?  To an FQDN, or an arbitrary IP address?

>There is currently a problem on the linux side:  I get a weird error
>in the syslog from the slip.login script.  I didn't chase down the
>exact command last night.

ISTR I was using mgetty on this box.  But that's a very hazy memory. And I do 
not now see any such thing in my kmenu's.
>
>So, from net, I try to telnet or ftp the linux box.  It doesn't really
>work, but some packets obviously get through.
>
>Using net.trace
This is on the coco?
>(the version of the executable with the 
>debugging commands, but no ftp) shows that nearly 50% of the ip
>packets have checksum errors.  And some of the packets (presumably
>from the ones that made it through the IP layer error-free) have
>checksum errors at the TCP level!
>
>Oddly enough, ifconfig on the linux side shows no errors, but a
>smaller number of packets transmitted & received.  Hummm.

Ditto.

>I make no claims of being SLIP Master X, but my memory is saying
>something about there being two different checksum algorithms?  Or am
>I thinking of XMODEM?

We have crc16, and crc32, and either needs to be built with the same seed 
number on both ends.  I don't recall the length of it that was used for the 
*modem commands, but the coco's os9 actually used a 24 bit variant in the OS 
itself.  Zmodem as exemplified by rzsz, was 16 bit, $0000 seeded, and by 
subbing a table lookup function for that, I was able to get another 240 cps 
out of a zmodem transfer on the coco.

>Anyway, tonight I'm going to start by debugging slip.login.

What package is that in, I do not have it here (F8 install)

>Willard



-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
"Boy, life takes a long time to live."
		-- Steven Wright