ctabka@sri-tsc (08/22/85)
From: David Wolfe <ctabka@sri-tsc> I currently work at the integration contractor for the DoD experimental packet radio network and I thought that I might be able to extend some of the points that Phil Karn mentioned. Firstly, the IP protocol is used because it provides a bare bones capability that allows dissimilar networks to intercommunicate with low overhead. If virtual circuit capability is desired, it can be added by implementing an appropriate Host-Host protocol on top of it (TCP is an example) in a modular way. Other protocols that provide different degrees of service run on top of IP. Amateur radio would do well to consider IP for any contemplated national networks for several reasons... 1.) The design has been done 2.) It can run on small, simple gateways (i.e. it`s cheap) 3.) It works (not the least consideration) 4.) It allows the decisions about higher level services to be delayed until the patterns of network use stabilize. 5.) The ham networks will be able to internet with a wide variety of other computer networks. 6.) And as Phil mentioned, datagrams are reasonable for adapting around disasters, topology changes and system crashes. Approaches such as using X.75 or other internetwork schemes based on virtual circuit models should be reconsidered because... 1.) If one guy in the tandem chain crashes, all current calls going through the switch get lost. 2.) The phone company solves problem 1. by using highly (some of the most highly reliable ever) reliable switches. This implies the use of fault tolerant machines for the communication subnet routing function, an approach that is useful for large corporations but not hams trying to build a network on the cheap. Also it is worth remembering that the phone company is partially locked into this way of doing things because it is a natural extension of the preexisting telephone technology (an important consideration with the size of the existing phone plant). 3.) The virtual circuit way can always be built on top of datagrams anyway if that`s what people want. For these reasons, I find the talk about basing Ham networks on X.25/X.75 to be very depressing. There are currently several DoD experimental networks running around the country. There is also some military PR work being done over in the UK. The technology seems to work fairly well,but it is difficult to tell when it will be deployed operationally. There has been a packet radio testbed in the Bay area since about 1974. The commercial situation is almost totally undeveloped and may be retarded by the advent of Cellular telephone. Probably the most interesting idea for commercial packet radio is to use it to distribute computing services in third world countries where they have not been building telephone services for 100 years. Reliable PR nets can be set up in the wilds without depending on End or Mobile Telephone Offices. Finally, one of the reason that spread spectrum and packet radio go so well together is that the use of SS increases the capture properties for the packets (similar to FM capture familiar to hams) in the face of multipath. This is worth keeping in mind as SS seems about to become legal on some bands. The challenge is to build a low cost approach to SS as the radios that we use (about the size of an IC290, cost about 12K) are far too expensive. This advantage pertains to civilian nets as well. There are some additional advantages that arise out of SS that briefly are. 1.) Using different spreading codes to that different nets can operate on the same band. 2.) Low probability of intercept for snoopers 3.) If the spreading codes are combined with cryptography, you get increased communications security. 4.) If number 2. is turned around, communications that is more jam resistant (The 2 meter repeater guys would love it!) 5.) It provides the basis for precision range measurements. (The preceeding does not represent the policies of SRI International, DARPA, or in fact, anyone else but the author) Dave Wolfe WA4YSS