[drats_users] More testing with ax.25

[Dan Smith]

> Have done some more Rf tests via 1 hop digipeating. For the most part it
> is pretty stable ,say out of 7 msgs (Emails) sent 4 were received and
> completed. The others would time out and stay in the outbox. My question
> is there error correcting retries? If the RF channel is busy will the
> system allow to resend the missed packet? The receiving end would show
> waiting for 1st block on those that would not go or a certain % would
> make it but then stop.

This is likely just because the digi changes the timing so much that 
some of them give up.  The message routing logic will stop trying to 
send a message once it fails (to a given destination station) more than 
a few times.  This is to prevent it from incessantly trying to send a 
message over a channel that clearly can't support it.

Unfortunately, this is going to require undoing some 
performance-enhancing tweaks to the session layer that have been in 
place for quite some time now.

> Both stations have the Block size set at a min of 32. next step is to
> try some of the other timing settings.. The TNC that is digipeating is
> also set at 255 Paclen to allow a larger amount of bytes to pass through
> per transmission. I wonder if that amount is still to small for the
> "Block" to go through.

A block of 32 that is compressed shouldn't get larger than 255, for 
sure.  However, there are two other things that may be factoring in 
here.  First, if you haven't set your window size down small, then it 
could be sending a bunch of 32 byte packets, thus overrunning the paclen 
of one or more of the TNCs in the path.  The other is that the code 
tries to scale its packet sizes up aggressively if the channel is 
"behaving".  On D-STAR, there are no maximum packet sizes or boundaries, 
so I can scale up to 2, 4, or even 8KB packets if things are "going 
well".  Obviously that won't work on a TNC.

Might be worth sending me your debug log immediately after such a 
transfer failed, just so I can (maybe) get an idea of what's going on.

Obviously more TNC-focused tweaking will be required regardless :D

Thanks very much for the data!

-- 
Dan Smith
www.danplanet.com
KK7DS