Topics

TCP/IP Over Packet

"Curt, WE7U" <curt.we7u@...>
 

On Thu, 2 May 2013, avignonmimi@... wrote:

Back in the late 90's we had a pretty good TCP/IP link from OKC to North Texas.

Actually, we used a TheNet IP node to bang the connected mode packets against to get an ACK, and then the rest of the link was ROSE. ROSE worked 100% better than pure IP nodes.

Here locally we also experimented with larger packet lengths between PC's where we were running JNOS. You definitely want to go with larger packets at speeds 9600 and up.

The obsolete AX.25 was designed for Z-80 boxes with little memory. PC's have plenty of memory for buffers. It was also designed for only 6-char callsigns, and I think some countries are issuing more letters than that. The repeater feature of AX.25 is also highly deprecated.
I seem to remember that we were using 1024-byte packets over 9600 baud in the Seattle area on TCP/IP back in the early 90's, and UI frames.

As soon as I get a box in my hand I'll be trying to do the same or better on 56k.

--
Curt, WE7U. http://wetnet.net/~we7u
Windows ate my homework!

"avignonmimi@..." <avignonmimi@...>
 

I'm scratching my head for memory, but I seem to recall setting the MTU to 4 times the PACLEN so that only one TCP and one AX.25 overhead packet was sent, and the other 3 packets were just data. This was for JNOS to JNOS. That was a real advantage of VC (connected mode on JNOS).

Course that feature would gag a netrom or rose switch, ha.

--- In UniversalDigitalRadio@..., "Curt, WE7U" <curt.we7u@...> wrote:

I seem to remember that we were using 1024-byte packets over 9600 baud in the Seattle area on TCP/IP back in the early 90's, and UI frames.

As soon as I get a box in my hand I'll be trying to do the same or better on 56k.