brian@sdcsvax.UUCP (Brian Kantor) (05/31/85)
From ihnp4!cbnap!gws@mit-eddie.ARPA Sun May 19 12:26:58 1985 Subject: lsi -11 programs Does anyone have any sort of programs for packet radio or amateur radio in general that will run on a pdp 11/03 11/23 or 11/23+ under RT-11 ?? many thanks and 73 Gary W. Sanders (N8EMR) AT&T Bell Labs (Columbus) ihnp4!cbnap!gws From bcn@mit-eddie.ARPA Sun May 19 19:22:08 1985 Subject: Routing and Addressing I am interested in routing and addressing issues associated with amateur packet radio, and am wondering what people have done so far. I read Phil Karn's paper in the proceedings of the 4th Networking Conference. The emphasis of that paper was mainly on the way things are done on the ARPA Internet. I agree with his recommendations, and would like to get started on solving (or attempting to solve) the problem. Here are my present ideas on the subject. Naturally, a full address should specify a host. The first three fields of the internet address (one of which is fixed) should specify a physical subnet. What I mean by a physical subnet is a collection of machines which can talk to one another directly (i.e. not using a digipeater). With this limitation, 8 bits may seem like a lot to allocate for hosts within a subnet. It isn't though, since repeaters may be used to extend the coverage of a subnet. One (or more) machine(s) on each subnet should be on another subnet and should act as a gateway between the two. As with the ARPA internet, subnets can be tied together by a subnet one level up in the hierarchy which has at least one host per lower level subnet, each of these hosts acting as a gateway. I think that the current emphasis should be on getting local subnets set up. As I mentioned, regular repeaters (as opposed to digipeaters) should be used to extend the range of the local subnets. Use of regular repeaters should cut some of the latency since the repeater would transmit as it is receiving the packet instead of storing it, and only retransmitting it after receiving the whole packet. We are about to set up a repeater at MIT for packet work. Clifford Neuman N1DMM From rgt@LANL.ARPA Mon May 20 06:24:06 1985 Subject: Packet Radio field sizes I read the article to this mailing list by Clifford Neuman <bcn@mit-eddie.ARPA> about routing ideas. I am interested in packet radio, but have done nothing about it for lack of money. I have not read the papers to which he refers, but do have one suggestion. Here at Los Alamos, we have an Integrated Computer Network of many computers. A friend of mine was working at a site, and they chose an 8-bit field. He suggested 16 bits. They said that 8 was enough, and would never be exceeded. Fortunately, he persuaded them to go with 16 bits. Now, about 5 years later, they have about 250 machines and adding more. Please use 16 bits, at least! I know, you will never have more than 256 machines in one subnet. But if it does happen and packet radio catches on (as it is already doing), there might be such a concentration in one area in the near future. Better to have a large, unused field in the protocol than have to redo the protocol at a later date because the sizes were exceeded. Build in plenty of room for expansion! Richard Thomsen rgt@lanl From michel@bbncct Mon May 20 07:49:53 1985 Subject: Routing and addressing Clifford Neuman raises the issue of routing. Let me carry over some lessons from the Internet. Again, these are personal lessons with which other Internetters might not agree. The original Arpanet was designed to be a small network. Quite early on the address field was expanded from 8 bits to 24 bits. But the mentality remained, that all network users would know or be easily able to figure out the physical address of a correspondent. One result of this seems to be that now that the INternet is grown to very large, world-wide proportions, addressing and routing are quite out of hand. The problem is that given a 32 bit Internet address, you can't figure out from the address what to do. The reason is that the network numbers (the first three fields of an address) don't have any topological significance. Network 194.1.1 might be in Japan while 194.1.2 could be in England. What this seems to mean is that each switching node must keep a table listing all networks, and the best strategy to get to each possible destination. Even now this is pretty unwieldy, and is a principle limit on the number of network numbers which can actually be assigned (or at least supported). The Internet is evolving various clever, ad hoc schemes to cope, but none is really ideal for the ham radio packet environment. A different approach has been taken by the CCITT in their design of an internetwork. Recommendation X.121 divides the world into 10 major regions, and further subdivides these into smaller regions, mostly according to political boundaries. Given nothing but the address, you can get fairly close to your destination. Any remaining ambiguity is presumed to be small enough to be handled by a local nameserver or table lookup in the switch, or something simple. THere are some troubles with X.121, mostly due to its Telephone company orientation. It presumes there will be few networks in any one country, ideally only 1. But it allows for variable length addressing, so that "area codes" with few networks can assign short numbers within the area. At this early stage of ham packet, the choice could still be made either way. I'd recommend the CCITT way. It might be that the X.121 choices are suitable as they stand. Perhaps the amateur fraternity should re-allocate the numbers. Perhaps the world-wide telephone numbering system could be used. Perhaps some derivative of the world-wide radio call sign prefix allocation would work. Any of these is superior to no system at all. I personally favor the telco approach. Addresses are allocated to geographic areas, and coded by expected frequency of use. Actually the European telcos did it better than AT&T by adopting variable length numbers. Addresses in the big city begin...1-some number of digits. Address in a rarely addressed location in the provinces might have a long prefix. Once inside an area, there is a similar coding. Popular addresses are given short codes. SInce this might seem to "discriminate" against the residents of the provinces, an escape is available for dialing local numbers. If you want a local number, you do not attach the "long distance" prefix. More or less like our telephone system. How long is the maximum address? In the telephone system 10 or 15 digits is a LONG nnumber, outside the USA. Coded as BCD, that is very manageable. In the protocol research community some advocate much different schemes. One scheme calls for addresses to be ASCII (actually "international alphabet #5") strings of up to 255 characters. Such an address could be read directly on a datascope, presumably, as it went whizzing by. Who allocates the area codes? Hmmm.... This is the hardest problem, by far. Sounds like diplomacy. Who maintains the "directory?" The white pages? (Sorted by name) The yellow pages? (sorted by some other consideration) Once the address assignment mechanism is chosen we can grapple with routing. SHould the user be allowed to specify a route (explicit routing)? If the address assignment is not systematic/topological, you need this capability. Telephone systems once had this, and phased it out in most places because it's unwieldy. WHen the net diameter gets big, you need a lot of overhead in the message just to carry the route info. I think it's worth considerable effort at the beginning to get this right. Designs tend to get entrenched, and that which we do today as a hack may last a long time. AXM From bcn@mit-eddie.ARPA Mon May 20 18:53:12 1985 Subject: Status report from Dave Mills Here is a copy of a status report sent by Dave Mills (mills@dcn6) on the ninth of this month: You can report that initial tests using IP/TCP were carried out using an LSI-11 and experimental software used on the ARPANET and DARPA Internet systems. A Heath TAPR board was operated in transparent mode via the WB4JFI-5 and W3VD-5 digipeaters on 145.01 MHz. The tests were carried out using virtual-terminal, file-transfer and mail protocols with traffic looped via a digipeater back to W3HCF. Additional tests using special performance-measurement protocols revealed disturbingly high packet-loss statistics of one in ten via the WB4JFI-5 digipeater and one in four via the W3VD-5 digipeater, but confirmed the viability of the IP/TCP protocols and implementation over typically raunchy Amateur circuits. Further tests using an IBM Personal Computer and an IP/TCP software package from MIT over the same channel were inconclusive due to inadequate timeouts in the MIT software. The authors of the software have been advised of this and, hopefully, can be persuaded to fix the problem. The MIT software, which is publicly available, provides virtual-terminal, file-transfer and several utility functions using standard DoD protocols, including IP/TCP. Dave From bellcore!karn@Berkeley Mon May 20 23:23:51 1985 Subject: Routing issues I agree with Tony Michel that the current Arpa Internet addressing and routing schemes are not well suited to the amateur radio environment. However, I think the real issue is hierarchical vs (relatively) flat addressing, and we can solve the problem without tossing out IP. The fundamental problem with the current Internet model is the implicit assumption that all hosts within a "network" (whether class A, B or C) can communicate directly, without IP's help. In other words, as far as IP is concerned, the Internet is a hierarchy with two and only two levels, "network" and "host". About the only way under the current conventions that a collection of hosts who are NOT fully interconnected can be made into an Internet is to assign fully interconnected subsets to small Class-C nets, or to exchange routing information at the host-specific instead of network level. Both of these techniques tend to result in routing table explosion. Clearly this restriction has already become a problem with current Internet, even ignoring the extra problems of packet radio; hence the big flurry of RFCs on subnet addressing schemes. The proposals had in common the addition of at least one new level in the addressing hierarchy, but in my opinion they are all too ad-hoc for amateur packet radio, and will probably become obsolete eventually for general Internet use as it grows even more. I believe that the most general scheme would be the best. It would allow an arbitrary number of levels within the constraints of a 32-bit IP address, with the bit field boundaries being completely variable. Routing table entries would consist of variable-length bit strings representing prefixes on IP addresses; an entry is used if the leading bits of the target address match the corresponding bits in the routing table. The special cases of 0 and 32 bit long table entries would correspond to the current notions of default and host-specific entries, respectively. As I mentioned, a gateway need not concern itself with the actual "meaning" of the bits in an address; it would simply try to "collapse" all routing table entries with a common prefix and routing decision (i.e., the same next hop) into a single table entry as a means of saving table space. There is no need for a gateway to retain any more information than that necessary to make a routing decision. For this reason, I am not really concerned with the actual boundaries and procedures that are established to assign IP addresses, so long as they are done in a hierarchical manner. Only the meaning of a few leading bits need be established on a worldwide basis to denote major regions (countries or continents, etc). National entities would then establish the boundaries for the next lower level in the hierarchy and "take" enough bits to encode them. The task of assigning the remaining bits is likewise delegated recursively until they're all gone (and hopefully you've reached the end host!) All along the way, the boundary assignments should be made with an eye towards the topology of the network in the specific area, so that as many routing table entries as possible can be "collapsed" together. If the regions at any level of the hierarchy differ in population then a Huffman-like variable width subfield encoding scheme should be used to conserve bits. There is no requirement that the number of levels be the same for all hosts, nor that the number of bits needed to designate any one entity within a hierarchical level be the same. This makes the scheme as flexible as possible and highly adaptable to local needs. ARPA has already assigned Class A network number 44 to amateur radio, so that leaves us 24 bits to play these games with. This means we can address 16 million hams, so I think the job ought to be doable. I presented this scheme in more detail in my ARRL Computer Networking Conference paper "Addressing and Routing Issues in Amateur Packet Radio". I'm very interested in comments and suggestions on this scheme. I certainly doubt that this idea is completely original with me (I seem to recall a mention of Danny Cohen at ISI proposing something similar) so if others have tried this scheme and failed, I'd like to know why. About the only problem I can see is the speed of the routing table lookup. However, I suspect that caching would help greatly here since a gateway is probably handling packets to only a few destinations over any short interval. 73, Phil Karn, KA9Q From bellcore!karn@Berkeley Mon May 20 23:41:12 1985 Subject: Re: IP over X.25 - the CSNET experience Paul, thanks for your comments on CSNET. I am somewhat familiar with CSNET, as another one of the developers is here in my group at work. So I am not totally naive as to the many problems caused by X.25. In fact, I have often used his experiences, which were essentially identical to yours, as potent ammunition against those who blindly tout X.25 as the ultimate networking solution. We have a Telenet connection, and as soon as we get the CSNET software I'll be able to experience these joys first hand. Perhaps the main difference between X.25 as seen by CSNET and AX.25 as seen a ham user is that AX.25 consists ONLY of the link layer (LAPB) from X.25; there is no packet layer as such. What this means is that there is no concept of a "message" spaning more than one physical frame. If we wanted to do "implicit" IP fragmentation, we would have to add an extra mechanism to denote the beginning and end of the datagram. This could be done, for example, by encapsulating the datagram with SLIP and sending it as serial data (ignoring link frame boundaries altogether), or by defining a few more AX.25 "level 3 protocol IDs" to indicate "head", "middle" or "tail" of a datagram. Both approaches are kludgey, to say the least. Don't forget that even the "reliable" (i.e., connected) mode of AX.25 can lose packets so whatever encapsulating method you use still has to be able to resynchronize after lost physical frames. [If you want the gory details on the flaws in AX.25, (and LAPB, too, to be fair) let me know and I'll make it the subject of a separate message.] If on the other hand there is a one-to-one correspondence between IP datagrams and AX.25 frames, then other things become much easier, e.g., mixing sequenced (I) frames with datagram-like unnumbered (UI) frames depending on the setting of the IP "reliability" bit. The cost is, of course, the extra overhead of the duplicated IP headers. The problem with just getting rid of the connection-oriented link (AX.25) layer and relying totally on TCP is that our links are currently so bad that hop-by-hop acknowledgements are often essential just to get reasonable performance over a long connection. I've worked out the formulas for the expected number of transmissions across all nodes for three acknowledgment strategies (end-to-end only, explicit hop-by-hop, and implicit (echo) acknowledgments without duplicate filtering). If anyone is interested I can post the gory details here. If anyone can suggest a link-level acknowledgement scheme other than the two I just mentioned, I'd certainly be interested, particularly if it's easier to implement and more efficient when sending datagram traffic than the go-back-N scheme in the connected mode of AX.25 (LAPB). Phil Karn From karn@umd5 Wed May 22 14:27:48 1985 Subject: Conference papers available on line I have put up the two papers I wrote for the Fourth ARRL Computer Networking Conference on machine "umd5". They can be retrieved with the "anonymous" ftp convention as: packet-radio/routing - Addressing and Routing Issues in Amateur Packet Radio packet-radio/tcp_ip - TCP/IP: A Proposal for Amateur Packet Radio Levels 3 and 4 Thanks to Louie Mamakos (louie@umd5) for making the necessary arrangements. Phil Karn From bellcore!karn@umd5 Wed May 22 15:30:29 1985 Subject: Packet Radio paper in IEEE JSAC I am proud to announce the publication of the paper "Packet Radio in the Amateur Service" by Harold Price, NK6K, Bob Diersing, N5AHD and yours truly, KA9Q, in the May issue of IEEE Journal on Selected Areas in Communications, pp. 431-439. I guess this means that amateur packet radio has "arrived"! Phil From CHEPPONIS@CMU-CS-C.ARPA Thu May 23 11:25:32 1985 Subject: link quality query; book recommendation A few days ago, Dave Mills posted a message giving present amateur packet link quality data for IP/ICMP using loopback through two digipeaters. The packet loss rates varied from 1 out of 10 lost to 1 of 4 lost. I'm wondering: would using Hamming coding on the link level reduce these rates? And what sorts of interference, etc., caused the packet lossage in the first place? (QRM, hardware failure, whatever). Also, a truly delightful book by Michael A. Padlipsky titled "The Elements of Networking Style" (Prenice-Hall 1985) gives insight into networking from an "Old Network Boy's" point of view. Find out why he thinks the ARM (ARPANET Reference Model, including TCP/UDP/IP) is the way to go, and why ISORM (The ISO Reference Model, pronounced "Eye Sore mmmm") is a loser. And lots of other networking stuff. Highly recommended! (Aside to Phil: Have you sent a copy to Terry yet?) -Mike, K3MC ------- From CHEPPONIS@CMU-CS-C.ARPA Thu May 23 13:38:31 1985 Subject: Another advantage of datagrams It just occurred to me how datagrams may be a very big win over Virtual Circuits: high bandwidth point-to-point links with low bandwidth links in between. Think of it as a sort of "frequency" division multiplexing. If we assume the source produces datagrams at a high rate, these datagrams can be carried by multiple (possibly slow) links and eventually be reassembled at the (high bandwidth) destination. Virtual Circuits, however, suffer from the "weakest link" principle: The greatest bandwith through a VC is limited by the bandwidth of the slowest link. BTW, are there any VC proponents on this list? -Mike, k3mc ------- From bellcore!karn@umd5 Thu May 23 15:45:18 1985 Subject: Re: link quality My personal intuition, from operating on 2m packet radio for about 2 years, is that the poor quality of present amateur links is due to two main reasons: 1) The modems themselves are rather inefficient. Bell 202 modem tones on an FM main carrier make very poor use of bandwidth and S/N ratio. The slow transmission speed means that the channel is easily congested, leading to frequent collisions, and of course the poor S/N efficiency leads to higher bit error rates. The 9600 bps direct FSK modem designed by Steve Goode, K9NG, has achieved bit error rates 3 db better than the TAPR 202 modem at 1200 bps, despite the 8:1 speed improvement, given equal RF power levels and bandwidths. 2) The networks are highly prone to "hidden terminal" problems. Although each station waits for a clear channel before transmitting, the following happens very frequently. Consider three stations, A, B and C. A can hear B; B can hear C, but A cannot hear C. C is sending a packet to B. A wants to send a packet. Hearing nothing on the channel, it barges in on C's transmission and clobbers B's reception. This problem is aggravated by the lack of proper backoff algorithms in the present packet radio controllers. The channels fall easily into congestion collapse, and only when the controllers begin to abort connections does the traffic fall back to a productive level. Phil From @DCN6.ARPA:mills@dcn6 Fri May 24 13:32:27 1985 Subject: Phill Karn papers Folks, Copies of Phill Karn's papers from UMD5 are now on DCN1. The one called packet-radio/routing is in PROUTE.TXT and the other called packet-radio/tcp_ip is in PTCPIP.TXT, both in the RFC login with RFC password. DCN1 is three gateways closer to ARPANET than UMD5, but is a fuzzball (not Unix) system. Dave ------- From bellcore!karn@mit-eddie.ARPA Fri May 24 14:12:10 1985 Subject: Re: Another advantage of datagrams Mike, this is called "load sharing". This is a special case of adaptive routing, which can be one of the major advantages of a datagram subnet. The VC advocates counter that this is too "difficult" a thing to do, and therefore conclude that it isn't a real advantage. However, if you look through the later revisions of X.25 and X.75, you'll see them working like mad adding something called "multi-link procedures". This allows them to gang together more than one physical link between two points, although this is far more difficult and less powerful than true adaptive routing in a datagram network because packets have to be resequenced at each link point. This seems to be a favorite ploy of the VC types. While they loudly proclaim that a certain innate advantage of datagram protocols isn't worth anything, behind the scenes they're scurrying about like mad trying to kludge it into their own protocols. Phil From bellcore!karn@umd5 Sat May 25 09:07:52 1985 Subject: Bugs in AX.25 The bug I alluded to in my earlier message has to do with the possible loss of I-frames that can result from just ONE frame being lost on the transmission link. Consider the following scenario. (Knowledge of AX.25 state transitions is assumed.) TNC A, attached to a terminal, wants to initiate a connection to TNC B, attached to a bulletin board system. A sends an SABM to B and goes into the Connect Pending state. TNC B receives the SABM, responds with a UA, and goes directly from the Disconnected state to the Connected state. HOWEVER, the UA is lost in transmission. TNC B, since it is in the connected state, is free to transmit I-frames and since its local client is a bulletin board, does so with the BB's login banner. TNC A is still in the connect pending state, since it never received the expected UA response to its SABM. It therefore has to ignore B's I-frames and, eventually, retransmits its SABM. However, this is interpreted at B as a link reset: clear all state variables and flags, and go to the connected state. The frames B transmitted earlier (the I-frames containing the BB's signon messages) are lost. A number of people have reported this phenomenon with bulletin board systems, so it is not just a theoretical possibility. I suppose I could coin Karn's corollary to Murphy's Law for communications protocols: if a failure mode is POSSIBLE in the protocol, no matter how "unlikely" it seems, then it is ABSOLUTELY GUARANTEED to occur (eventually). This problem seems to be fundamental to LAPB and cannot be solved without the use of a three-way handshake. Such a handshake is provided as standard equipment in TCP, but not everybody is convinced... Phil Karn From @DCN6.ARPA:mills@dcn6.arpa Sat May 25 21:42:46 1985 Subject: TAPR considered unfriendly Folks, I tried valiantly today to twinkle unsequenced (datagram) packets using the TAPR board, but got ambushed at every turn. I used raw Internet Protocol (IP) packet formats sent by a fuzzball and looped in full-duplex mode back to itself, a trick which works in transparent mode with sequenced (virtual circuit) packets. Unfortunately, the TAPR board not work in transparent mode with unsequenced packets; therefore, today's ambushes were in converse mode. My first battle was simply to prefix all control characters (there are about eleven in the TAPR converse mode) with the literal-next (PASS) code to insure character transparency, but the Tuscon boys have laid their traps well. Converse mode is intended for chit-chat between consenting, vanilla ASCII terminals; therefore (?), only seven bits are significant, even if the UART parameters are set for eight bits and no parity. The Principle of Least Astonishment would suggest all eight bits as significant, even if control-character interpretations were masked to seven bits. This makes it easy to win transparency battles, and also allows parity to be checked on an end-end basis. I counterattacked with the time-honored trick of encoding three octets as four septets encapsulated as vanilla, upper-case ASCII characters preceeded and succeeded by a couple of non-TAPR control characters, so that the miscellaneous TAPR gratuitous commentary could be suppressed. That hideosity worked and the unsequenced IP datagrams did fly; however, the TAPR commandos decided unsequenced packets longer that 128 octets were not polite, in spite of the fact that sequenced (tranparent) packets up to 256 octets work fine. For my next raid, I carefully broke packets longer than 128 octets into fragments no longer than 128 octets, with each one terminated by the send-packet TAPR control character. But this meant that the character following a fragment terminated by the send-packet character could wander into the TAPR board hard on the heels of the send-packet character itself, unless some kind of timeout intervened. In spite of a generous amount of buffer storage and in spite of this situation working fine with sequenced packets, the TAPR folk simply amputated anything longer than 128 octets. One would assume they feared a congestion-gas counterattack. I conclude the braindamage has gone far enough and trying to trick the TAPR board into unsequenced transparency is interpreted as indecent exposure. I should have kept my clothes on, having warred this way with more operating systems than I care to admit. Until I do something dramatic, like write an AX.25 driver for the fuzzball, I will TAPR IP only in sequinned nudity. Dave ------- From Elias.PDO1@MIT-MULTICS.ARPA Mon May 27 18:05:03 1985 Subject: A network primer on networking? My congratulations to Cliff for starting this net, and to everybody else who has contributed interesting and sometimes hilarious entries on this important subject; they made me realize that there is more to networking than sending $164.95 plus postage and handling to GLB Electronics. They also made me realize that my claims to electronic knowledge and computer wizardry shrink to subparticle proportions when confronted with the world of networking, a subject about which some people even write their Senior Thesis. So. I would like to propose that you wizards out there get together and prepare a set of files explaining "Networking for Laypersons". In particular, It would be nice to see something like a set of (three?) files, each discussing the area in successive levels of detail (layers?); the first one could be a tutorial version of pages 19-21 to 19-31 of the 1985 Amateur Radio Handbook; in particular, I think you guys could do better than the Handbook in two ways: 1- You seems to know the stuff better, at least by comparing the alphabet soup you use with the glossary on page 19-30. 2- You could pattern it to the "learning packeteer", instead of attempting the double education-reference scope of the Handbook (admirable, but confusing). The next files could dwell on the relative merits of Virtual Circuits versus unsequenced datagram packets, etc., not in the flaming mode reserved for the actual mailing list transactions, but in a paused, academic way (.... .. .... ..). Now; if only these files could be created, placed, accessed, and edited in a location where a large number of mailing list readers could access them, I can see the following process happening: 1- Info-Hams aficionados, like myself, will be informed of the existence of such files and how to read them. 2- After reading them, one would get a much better appreaciation of what's going on on packet-radio (and the real world, for that matter). 3- The real wizards, meanwhile, keep these tutorial files updated while flaming on packet-radio. Who knows, we may get the first net-wide book... on the subject of networking, which seems to be appropriate... - Antonio/KA1LLM From KPETERSEN@SIMTEL20.ARPA Wed May 29 09:31:15 1985 Sender: CLEMENTS@BBNG Subject: W0RLI Packet Radio Mailbox/BBS/GateWay system Version 9.3 Now available from SIMTEL20: Filename Type Bytes CRC Directory MICRO:<CPM.PACKET> PACKET93.LBR.1 BINARY 219648 0C0BH PACKET93.LBR contains the files that make up version 9.3 of the W0RLI Packet Radio MailBox/BBS/GateWay system. This system runs on the following hardware: Computer: Xerox 820-1 computer (the ones that were available for $50, and are still around for not much more), one or more 8" single density, single or double sided disk drives, parallel keyboard, CRT monitor. Packet Radio gear: One or Two TAPR (or AEA) TNCs with version 3.1 or later software. (Two TNCs if you are going to run a crossband Gateway.) Radio gear: One or two transceivers. The W0RLI software supports sending, receiving and forwarding mail, uploading and downloading files, capturing typescripts, logging channel activity and mailbox activity, and gateway operation between two TNCs on two bands. Read the file NOTES.TNC to start working your way through the documentation. Hank would appreciate knowing of users who are running this software. A QSL to Hank or a net message to me would be appreciated. Here is Hank's update from the February 1985 NEPRA PacketEar: The MailBox/GateWay has now been sent to 25 states and 5 countries. As far as I know for sure, it is on the air at least 20 places now. In the Boston area, 4000 messages have passed through it. The local forwarding network now includes 9 nodes: W0RLI, WB2OSZ, WB1DSW, K1BC, WA2RRKN-2, K7PYK, WA4SZK, KA1T, W1AW-4. The last two run their own software, but allow for forwarding from the W0RLI systems. A message put into any one of these systems will find its way to the system nearest the intended recipient. There are several other areas of the country now using the software: Georgia, Arizona, Iowa, Washington DC, Seattle, ENY/NYC/NNJ, Dallas, Illinois, Southern New Jersey, Los Angeles have all been heard from. All have the software in daily use. Expect to see it on Oscar-10 at KL7GNG soon. Look for it from Australia, New Zealand, Japan, Hungary. Sacramento county RACES will be using it. GateWays are running at W0RLI, K7PYK, WA4SZK and WB7DCH. de Hank Oredson, W0RLI The following is an extract from the file NOTES.TNC for version 9.3 of the W0RLI MailBox and GateWay software. This extract contains the list of changes since the last distribution to the SIMTEL20 repository, which was version 8.6. W0RLI, Hank, does not have access to either ARPANET or USENET. I will be glad to try to answer questions or to relay them to Hank. I can be reached at: ARPANET: CLEMENTS@BBN USENET: {ihnp4, decvax, linus, ...}!bbncca!clements 73, Bob Clements K1BC --- excerpt from NOTES.TNC follows --- W0RLI MailBox and GateWay Version 9.3 - 5/16/85 Created and distibuted to the packet community by: Hank Oredson, W0RLI 19 North Hill Road Westford, MA 01886 These notes are rough, more release notes and tech notes than anything else. A SYSOPS Manual (Very nice, 25 pages or so) is available for 8-1/2 x 11 SASE ($1.24 postage) from: Jon Pearce, WB2MNF 109 Pine Cone Trail Medford, NJ 08055 A very nice log file analyzer was written by: Tom Hogan, WB7DCH 26911 S E 456 St. Enumclaw, WA 98022 I have included this on the release disk as LFA.COM. Release notes, Version 9.3 : In version 9.0, there was a change to the structure of the mail file. When TNC is first run, it will update the mail file to the new structure. This is done by doing an untangle. Don't panic when this happens the first time you run version 9.3! Changes and additions since version 9.2 are: Added privelege A and B, excluded on A or B ports. Removed LA and LN, was bad idea. Support for S W0RLI @ K1BC installed... Added H local command: short / long menu (=Help). Added $X, $Y, $Z : Date, time, current msg # at last login. Verify for files specified in config.tnc that the drive is on line and write enabled. L now lists new, LN same, LA lists all. Faster forward - send "S XXX" and title then eat 2 lines. Changes and additions since version 9.1 are: Much faster untangle. D, DP, DU for remote sysop. CP ON/OFF and CR ON/OFF for remote sysop. Changes and additions since version 9.0 are: "Remote sysop" feature added. Better handling of user record currency. Excluded user disconnected with no "bye" message. Fixed bug - long packets not get monitored properly. Added ki4xo changes for 5" to boot, sbios, cbios. Added GM, GU, OA, OB, C <call>. Changes and additions since version 8.9 are: Some changes to CBIOS thanks to ke1g give faster disk I/O Added user privilege check. If E, then user is excluded from use of the MailBox. N is used for normal user, and is default. Fixed connect bug - now conok on is last sent, conok off is first. Most searches go most recent first. Added LL (List Last n), and LN (List New). User file version 1, and DU - Display Users, EU - Edit Users. Mail file structure version 2, back chaining of msg headers. N menu item, rename file. Changes and additions since version 8.8 are: Fixed (again) the disconnect process. Added DP - page mode. G menu item replaces UNTANGLE, GR to renumber the messages. L and private msg, show only to owner/sender/addressee. Changes and additions since version 8.7 are: Name of CONFIG.TNC file can be specified at execution time: TNC OLDCFG.TNC or TNC NEWCFG.TNC for example. Defaults to CONFIG.TNC if not specified. The names for files CALLS.TNC, FWD.TNC, LOG.TNC specified in CONFIG.TNC Installed ke1g improved cbios. Thank you Bill. Y replaced by YC, YF, YL, switch calls, forward, or log files. Fixed (I hope) the "Can't DISCONNECT, Link state is..." bug. Added Z (delete file) to local menu. X menu item forces forward regardless of hours given in FWD.TNC. Was not passing "*** LINKED to" thru from next GateWay. Not allow GateWay connect line with "tran" at end. Changes and additions since version 8.6 are: Fixed (again!) the lack of timeout when bombarded with con req. Added UA (append) option to local menu. Fixed dayclock month rollover. Split EDMSG, FWD, MBFILE as separate routines. --- End of extract ---