John:
It has happened here with users with faulty, deaf, modems or choked up
computers (bad memory management setup), whenever polling gets spawned
and then becomes very very wasteful of the packet channel capacity.
I think it was Julian, EA4ACL that discussed that on this list around
march or april 1999, cannot be too conclusive reminding that far back, he
sent a patch, but I believe it deactivated DAMA. I used it, and lost it as
well. Yes, there may be some protocol issue with kernel AX.25 I am not
wise enough to identify.
JNOS is far cleaner in its behavior, but I am also paranoid in avoiding
digis whenever I can. That is a bad experience with a fwd partnet on 20 m,
300 baud. He had an internal digi using 38K4, when I was using BPQ on
MSDOS. When BPQ 4.08 a sees there is a digipeater, it assumes it is a 2 x
300 baud path, and puts a longish delay between packets, which is
unnecesary, but it is not wise enough to know, maybe some RTT poll could
find that out (how to do it is something else). I don't know if LinuxNode
0.2.4 behaves like that, but I have avoided digis once again.
I don't think you have a configuration problem. Seems that somebody wise
enough to do it (and with enough time) should review the AX.25 kernel
code, once again. You are not alone on this and I am not the one capable
to fix it, either.
That was using 2.0.36. And for direct connects, I AVOID netrom as much as
I can. Some BPQ configurations (FBB, Winpack as BBS) seem to prefer netrom
on a single hop, wasting some more bandwidth by including an otherwise not
necessary netrom header in the data.
73, Jose
---
On Fri, 4 Feb 2000, John Ackermann wrote:
> Sorry if this is totally ignorant, but I've had embarassingly little
> experience tracing AX.25 protocol issues.
>
> I'm trying to track down some serious performance problems on our
> 19.2kb network. We're seeing frequent disconnects between a Linux
> AX.25 station and a G8BPQ node.
>
> The Linux setup is kernel 2.0.34 with ax25-module-14 and utils 2.1.42a.
> For the 19.2 link, we're running a T1 timer of 2.5 seconds, and T2
> timer of 200ms.
>
> At the bottom of this message is a trace of a common situation that we
> see, with lots of poll/final exchanges in a short time period. This
> seems odd, and wasteful of bandwidth. It sometimes seems like we have
> more supervisory packets like this than we have packets with payloads.
>
> Is this normal? If not, do we have a configuration problem that's
> causing this (on either the Linux or the G8BPQ side), or is there some
> other explanation?
>
> Thanks for any light anyone can shed on this.
>
> 73,
> John N8UR
> [EMAIL PROTECTED]
>
> --------
> from listen -p pi0a -a -t, running on the W8APR-3 machine:
>
> [Fri Feb 4 20:39:57 2000]
> Port pi0a: AX25: W8APR-3->N8BJQ <RR R F R0>
>
> [Fri Feb 4 20:39:59 2000]
> Port pi0a: AX25: N8BJQ->W8APR-3 <RR R F R0>
>
> [Fri Feb 4 20:40:00 2000]
> Port pi0a: AX25: N8BJQ->W8APR-3 <RR C P R0>
>
> [Fri Feb 4 20:40:00 2000]
> Port pi0a: AX25: W8APR-3->N8BJQ <RR R F R0>
>
> [Fri Feb 4 20:40:02 2000]
> Port pi0a: AX25: N8BJQ->W8APR-3 <RR C P R0>
>
> [Fri Feb 4 20:40:02 2000]
> Port pi0a: AX25: W8APR-3->N8BJQ <RR R F R0>
>
> [Fri Feb 4 20:40:04 2000]
> Port pi0a: AX25: N8BJQ->W8APR-3 <RR C P R0>
>
> [Fri Feb 4 20:40:04 2000]
> Port pi0a: AX25: W8APR-3->N8BJQ <RR R F R0>
>
> [Fri Feb 4 20:40:08 2000]
> Port pi0a: AX25: N8BJQ->W8APR-3 <RR C P R0>
>
> [Fri Feb 4 20:40:08 2000]
> Port pi0a: AX25: W8APR-3->N8BJQ <RR R F R0>
>
> [Fri Feb 4 20:40:08 2000]
> Port pi0a: AX25: N8BJQ->W8APR-3 <RR C P R0>
>
> [Fri Feb 4 20:40:08 2000]
> Port pi0a: AX25: W8APR-3->N8BJQ <RR R F R0>
>
---
Ing. Jose A. Amador Fundora | Tel : (537) 20-7814
Dept. de Telecomunicaciones | E-Mail : [EMAIL PROTECTED]
Facultad de Ing. Electrica |
ISPJAE |
