[nos-bbs] KLMproxy - Windows TCP/IP without the IP over RF - anoldidea ...

[Bill Vodall WA7NWP]

> IP over AX25 might have an overhead on 1200bps, although it is only 40 bytes 
> out of 256 bytes, i.e. 15% slower than a plain AX25 packet, but it does work 
> when tweaked properly. JNOS is good for this sort of tweaking.

It does indeed work well when tweaked.  Specially if it's a long path where
basic ax25 wouldn't work or would time out.

However, the overhead is worse then then 15%/40 bytes makes it appear for a
couple reasons.

First is that the ACK also has the 40 byte overhead.  I may be off a byte or
two here, but a RR(next) in an AX25 packet essentially has none of the 40
bytes of the TCP/IP headers.

Secondly, the AX25 windowing scheme helps.  (As would no doubt a well tweaked
TCP/IP circuit..)   A recently example showed up when I used Airmail to send a 16k
image via a local (no-digi's) telpac_node.

My system sent three long data packets

braaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaap

braaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaap

braaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaap

and the node responded with a single RR-next packet.

3 data packets <RR>  3 data packets <RR>  ...

It only took a few minutes to dump the data file.  I feel it
was about twice as fast as the equivalent FTP transfer would
have been.   Someday when I'm looking for something to do,
I'll give it a try and get some real numbers.

Bottom line is that for point to point I feel we can do better
with direct packet then with TCP/IP. 

>
>

On that note, is there any Linux tool, preferably with a *NOS equivalent command
to simply and quicly copy a file from one station to another using packet?   With that,
we could exchange UUCP styple bundled and compressed email and news messages.

73
Bill - WA7NWP