The Digital Domain of Amateur Radio - Wireless TCP/IP - Amateur Radio Digital Communications (NET-AMPRNET) AMPRNET 126.96.36.199
Many hams are unaware that a Class A IP address block (16.7 Million IP addresses) has been set aside for amateur radio users worldwide to use for connecting their radio equipment to and though the internet.
This address allocation was originally obtained in the late 1970's by Dr. Hank Magnuski (KA6M) long before the public internet was in it’s infancy.
At the time the internet was called ARPA (Advanced Research Projects Agency) Net later renamed The Defense Advanced Research Projects Agency (or DARPA).
It wasn't till about 1973 development began on the protocol later to be called TCP/IP, it was developed by a group headed by Vinton Cerf. The TCP/IP protocols were adopted as Military Standards (MIL STD) in 1983, and all hosts connected to the network were required to convert to the new protocols. But the public Internet didn't really emerge till the 1990's.
Back 1969, Norman Abramson launched Aloha Project at University of Hawaii. Motivated by the poor telephone lines in the Hawaiian Islands, funded by ARPA to investigate how to build a packet switched network based on fixed site radio links. At the University of Hawaii it was not really an option to use the PSTN or any form of cabling between Hawaii’s many islands. It opted instead to connect the seven colleges spread across four islands by the use of amateur radio. Norm performed a number of experiments around 1970 to develop methods to arbitrate access to a shared radio channel by network nodes. This system operated on UHF frequencies at 9600 baud. Abramson later developed a satellite version of ALOHAnet called PACNET.
Internet Pioneer Paul Baran, W3KAS an engineer who helped create the technical underpinnings for the Arpanet, the government-sponsored precursor to today’s Internet. In the early and mid 60's he came up with the concepts of packet switching and distributed networks. He also proposed a decentralized network, a way to create a survivable infrastructure. This is where radio links came into the picture.
Baran subsequently founded a slew of technology companies including Packet Technologies and Metricom, and still continued to be actively involved in founding new start-ups – the most recent being Ethernet over wireline outfit Plaster Networks, and the IP TV infrastructure company GoBackTV.
Amateurs all over the world have been networking wirelessly, and have been long before "Wireless Web" became a cell phone feature. Volunteer hobbyist ham radio operators are investigating the construction of an entirely radio-based world-wide network using TCP/IP. Hams are in an unusual situation compared to the rest of the Internet. We're trying to build our own radio-based network, so for the goals of the network, radio connectivity is more important than internet connectivity.
Around 1978, group of amateur radio operators, the Vancouver Amateur Digital Communication Group (VADCG) in British Columbia, Canada began experimenting with Packet radio using a Terminal node controller (TNC) developed by Doug Lockhart, VE7APU. Federal Communications Commission approved the transmission of ASCII for Amateur radio in the United States in 1980, and AX.25 was born. It was considered standardized in 1984 with the release of the v2.0 specification.
Use of the TCP/IP over radio dates back to even the DOS days, using a TSR (terminate and stay resident) application to handle the TCP/IP traffic, and command line utilities to handle ping, ftp transfers, and more. In 1985, Phil Karn, KA9Q developed and made public this TCP/IP stack software for MS-DOS. When Linux came along also in the 90's, shortly there after this became the first and only operating system to include amateur radio AX.25 protocol support built-in. Allowing direct TCP/IP over amateur radio too.
IPv4 Network 44/8 is known as the AMPRNet, named from "AMateur Packet Radio Network". Much of the AMPRnet piggy-backs over the internet very much like a VPN (Virtual Private Network) except the amateur lines are not private. This greatly improves the long-haul traffic handling and it is just the "last mile" that is the bottleneck in the system, that 10 miles or so to your radio location.. The current structure of the AMPRNet is that there are a bunch of fully- and partially-isolated ("disjoint") subnets in nearly every country and most major cities around the world. In most countries, there is a local coordinator who is responsible for assigning an address and updating the master hosts list.
The AMPRnet protocols extend the normal AX25 packet transmissions with the full support of TCP/IP. This greatly enhances the functionality of the network. Telnet, FTP, ping, even http are all supported.
And it doesn't have to be limited to just slower packet radio for TCP/IP, either. Some of the newer 802.11 wireless ethernet devices use frequencies that meet with amateur radio spectrum in the 2.4 GHz area. As a result, amateurs can modify the Part 15 compliant devices to increase the power and use better antennas, providing more gain and increasing usable range. These devices are considerably faster at up to 54 megabits per second than the 1200 and 9600 bit per second speeds of VHF and UHF packet radio.
Good connectivity enables a number of applications that were not previously practical to experiment with due to bandwidth requirements; among these could be digital voice repeater linking, digital quality facsimile picture transmission, television (D-ATV), Web-SDR, multimedia, and so on.
Since the mid 1980's the 44/8 allocation has been administered by Brian Kantor, WB6CYT. He also manages the low-bandwidth tunneling 44/8 router, (routing between various 44.x.x.x networks are typically done by IPIP encapsulation where true amateur routes do not exist). This router allows a minimal connectivity between the main Internet and some parts of the AMPRNet through gateways. It is there primarily to allow experimenters on the AMPRNet the opportunity to exchange information and to obtain access to Internet resources. Each of the gateway entities in the UCSD router determine to some extent, what is done at lower levels. Some areas subnet while others do not.
In early 2012, Heikki Hannikainen, OH7LZB (the ham behind aprs.fi) modernized the amprnet routing by writing a custom RIPv2 daemon to receive RIP updates from the 44/8 ampr.org routing service, and insert them in the Linux routing table. This has replaced the encap.txt and munge script method. In late 2012 a new unified interface was given to www.ampr.org.
Along with the new interface, came policies and an Acceptable Use Agreement.
This allows those with the capability to enable directly routed subnets via BGP advertisement. This will help shed the load on UCSD, and reduce the point of failure.
All of the BGP announcements on the internet that include 44 space can be
determined by looking at the ARIN Allocated Address Table at: http://thyme.rand.apnic.net/
Prior (1995-2012); another email robot that Jim Fuller, N7VR maintained emailed a daily encap.txt list of NOS style route commands. From there if you were on the ball you had a cron script munge these into your gateways routing tables. Most folks who ran gateways were not on the ball and would manually install route updates as time permitted.
And before that Warren Toomey VK1XWT, ran the gateways robot.
IPIP can not traverse NAT because first of all it uses proto 4 (encap) and not TCP or UDP, so contrack, which manages NAT traversal does not support it.
In a NAT situation, incoming connections not triggered from the inside don't
get translated, since the NAT doesn't know the originator. Being a stateless
point to multipoint communication, you do not have a "inside"
originated connection for all connections. It is expected that NAT knows where
to forward a data packet by using information from the original outgoing
connection (established and related packets are sent to the internal originating
ip ), which is not the case for IPIP since incoming data from another host (we
have a mesh architecture) has no corresponding outgoing connection.
On the other hand, in OpenVPN, if the server is located outside, all connections are stateful and trackable by contrack, being a single IP endpoint on port 1194 originated inside, so NAT traversal is as simple as any connection originated locally.
Part of the reason Amprnet uses IPIP tunnels instead of something like openvpn has to do with keep alive traffic and decentralization. The other reason is the RFC for IP Encapsulation within IP dates back to the 1980's when the Amprnet got started.
Openvpn and other more modern protocols all have IP tunneling as part of their ability to connect disjointed network segments and hosts. However they all use a stateful client to server approach. This is so that the server can maintain a return path though any potential NAT and firewalled paths. The keep-alive traffic from this periodic handshaking would add up to a rather large amount for a class A network, such as 44 net, even with say just 200 tunnel points.
With IPIP, decentralized peer to peer tunnels are used, which saves a very large amount of bandwidth that would otherwise be necessary when everyone goes though a central server. Also with IPIP, you don't have that single point of failure.
Back to the WAPR TCP/IP page
Back to the HSMM Wireless page