Getting Started with TCP/IP on Packet Radio

Getting Started with TCP/IP on Packet Radio


by John Ackermann, AG9V
Miami Valley FM Association
Dayton, Ohio





8 April 1992






Copyright 1992 by John R. Ackermann, Jr.
This document may be freely copied and distributed for non-commercial
use only, provided this copyright notice is included.  You may modify
this document but you must retain this copyright notice and include an
additional notice identifying you as the creator of the modifications.
You must also send a copy of the modified version to me at the address
shown below.

(This is the generic ASCII text version.  Any volunteers to help clean
it up and make it pretty?)

Introduction

This document is intended to help hams with some experience in packet
radio get started with the TCP/IP software written by KA9Q and others. 
It is not intended to take the place of the software's reference
manual, but rather to provide a quick-and-dirty introduction to the
capabilities of TCP/IP and the mysteries of installing and using the
software.

There are several different versions of the KA9Q software floating
around.  It was originally written for MS-DOS computers, but has been
ported to Macintosh, Amiga, Atari and UNIX systems.  The original
program was called NET and its last formal version was issued in
April, 1989.  If someone talks about 890421.1 NET, that's what they're
referring to.

Since 1989, work has concentrated on a rewritten program called NOS
(for Network Operating System).  NOS offers many new features that
make using TCP/IP much more effective.  However, it is a growing and
changing creature, and keeping up with it is difficult.  We recommend
that you use NOS instead of NET, but we can't tell you precisely where
to find the latest version.  Your best bet is to check with a local
user.  If you can't find a copy locally, there are several telephone
BBS systems that carry the software, but be prepared to find a
bewildering array of versions and flavors:

N8EMR's Ham BBS (614) 895-2553
ChowdaNet (401) 331-0334
WB3FFV (301) 335-0858

This document is based on the MS-DOS version of NOS, specifically the
PA0GRI adaptation of the June 18, 1991 version of NOS as modified and
distributed by N1BEE (what a mouthful!) and available as grinos from
the ChowdaNet BBS.  The discussion tries to stay away from features
that are specific to that version, but if something we say doesn't
seem to make sense with your version of the software, that's probably
why.

A last note before we plunge in -- we said it before, and we'll say it
again:  this document barely scratches the surface of NOS.  Nearly
every command described here has options or parameters that we're
ignoring.  Our goal is to give you a feel for what TCP/IP does, and to
get you on the air with NOS; to get beyond the novice stage you need
to look at the reference manual and experiment with the software. 
Appendix A includes a list of organizations and individuals that can
provide further information about TCP/IP and amateur radio.


TCP/IP and Ham Radio

TCP/IP is a set of communication protocols that have become a standard
in the computer networking world.  It is designed to link different
kinds of computer systems together over dissimilar networks.  TCP/IP
software runs on nearly every kind of computer available, from IBM
mainframes to PCs, Macs, Amigas, and Ataris.  The KA9Q software (from
now on, we'll call it NOS) is special because it includes the features
necessary to run TCP/IP over ham packet radio.

The TCP/IP protocol suite allows different kinds of computers to talk
to one another across networks.  The services it provides include
terminal sessions, file transfer, electronic mail, and data routing
services.  Computers running TCP/IP (referred to as hosts) can run
some or all of these applications simultaneously; it's entirely
possible to sit at a PC computer running NOS and carry on a keyboard-
to-keyboard chat with one station, while another retrieves a file from
your hard disk and you send electronic mail to a third.

It's also comforting to know that when you run TCP/IP, you don't give
up the ability to carry on normal packet communications.  You can use
NOS just like a terminal program to establish connections with your
local BBS, etc.

If you've looked at the size of the NOS documentation, you're probably
asking yourself what the benefit is of mastering this fairly complex
stuff.  Well, NOS has several features that improve on regular packet
radio.  It has much more sophisticated file transfer and electronic
mail capabilities than our present PBBS systems (and it's possible to
feed PBBS messages into NOS in a way that makes it much easier to use
them).  It supports multiple simultaneous connections.  It has new and
better methods of dealing with slow and congested channels that
improve the reliability and throughput of packet traffic.

NOS also has the ability to route transmissions to distant stations
without the user needing to know every hop along the way; all you need
to do is get your data to a gateway station that knows how to move it
one hop closer to its destination.  New work being done with NOS
promises dynamic routing that automatically adjusts to changes in the
network.

And, since it is directly adapted from the de facto standard system of
interconnecting computers, it offers the possibility of sophisticated
services far beyond anything available on regular packet radio.  For
example, in some areas ham TCP/IP users can log into multi-user Unix
computer systems and run applications as if they were directly
connected to those machines.


What is TCP/IP?

As mentioned above, TCP/IP is actually a set of protocols for the
transfer of data across networks of computers.  Two of these protocols
underlie most of the others, and they give the set its popular name:

TCP Transport Control Protocol, a "reliable stream service" (which
is a fancy way of saying it makes sure that all the data sent
to a remote host actually gets there).

IP Internet Protocol, which sets the basic rules for formatting
packets of data to go out over a network.  TCP rides on top of
IP.

Now that you finally know what TCP/IP stands for, there are a few
concepts that are critical because they address the basic problem of
any communications system -- identifying the parties to the
conversation.  Simply using our ham callsigns to address TCP/IP
packets doesn't work for two reasons.  First, the protocols work
across many different networks, and have to have a consistent address
scheme.  Second, and as important, ham callsigns don't contain enough
information to allow TCP/IP's sophisticated routing mechanisms to
work.


Names and Addresses

The first important concept is the IP Address.  Since these protocols
are used on lots of different computers, it was necessary to come up
with an addressing system that works with all of them, that provides
enough routing information, and that doesn't take up a lot of space. 
The answer is to build TCP/IP addresses out of a four byte sequence,
with each byte providing information about the network an
subnetwork(s) to which a host belongs.

TCP/IP addresses are called hierarchical because the four bytes have
decreasing significance from left to right.  By looking at the
leftmost byte(s) we can learn how to route a transmission to the host
represented by the rightmost byte(s).  We usually print these
addresses using the numeric value of each byte, separated by a period,
such as [44.70.12.34].  This is known as "dotted notation."  The
square brackets aren't strictly necessary, but they are convenient to
set off ip addresses; we'll use them in all our references.

We won't go into all the semantics of hierarchical addressing here,
but as an example the address [44.70.12.34] breaks down as:

44. The network assigned to amateur radio TCP/IP.
70. The subnetwork for Ohio.
12. The Dayton/Cincinnati area.
34 A specific system address within that area.

IP addresses are assigned by coordinators who derive their authority
from a central registry.  The central coordinator for the ham radio net
(44.xx.xx.xx) is Brian Kantor, WB6CYT.  He has delegated authority to
assign addresses to various state and national coordinators.  The
folks in Appendix A can help you find your local coordinator.

The second important concept is the hostname.  Obviously, ip addresses
aren't very intuitive.  English-like hostnames make remembering
addresses much easier, and TCP/IP programs, including NOS, have means
(discussed below) to map between ip addresses and hostnames.  A host
is any computer running TCP/IP; even when you're using services from
another computer, your system is still a host.  When we talk about a
remote host, we're talking about a machine that you're communicating
with via TCP/IP.

The convention in ham radio TCP/IP is to use your callsign as your
hostname.  To help reduce confusion, we usually print hostnames in
lower case, and callsigns in capital letters -- my hostname is ag9v,
and my call is AG9V (though NOS isn't case sensitive and won't care if
you don't do it this way).

Closely related to the hostname is the domain name.  A domain is a
group of machines that are logically (though not necessarily
physically) connected together.  Domain names are like ip addresses;
periods separate parts of the name, with each part representing a
different level in the domain hierarchy.  But the domain name is
ordered in reverse -- its highest-level portion is at the right, the
opposite of ip addresses.

The ham network's domain is ampr.org; org (short for "organizations")
is the top level domain, and ampr is the second level domain,
containing all ham TCP/IP hosts.

When you combine a hostname with a domain name, you get something like
ag9v.ampr.org.  This is called a Fully Qualified Domain Name (FQDN --
knowing this acronym allows you to sound like a real expert).  If a
host has multiple users, we can add the user's login name at the
beginning of the address, separated from the FQDN by a "@" character. 
This combination is commonly known as an Internet address (the
"Internet" is the general term for all the TCP/IP hosts that are
interconnected in the commercial, educational, and military domains)
and is the address form used for most electronic mail in the real
world.  For example, if there is a user "jra" at ag9v,
jra@ag9v.ampr.org would be that user's full Internet address.

There's one last twist.  Some services (such as Domain Name Service,
discussed below) need to know whether an address they are processing
is in fact an FQDN.  To do so, they look for a trailing period at the
end of the domain name.  Some versions of NOS ignore this issue, but
the PA0GRI versions (such as grinos) insist that you "anchor" all
domain names with a period at the end of the name.  In other words,
grinos will barf if you issue the command "hostname ag9v.ampr.org" but
"hostname ag9v.ampr.org." will make it happy.

This may seem like an overly complicated scheme to simply allow two
hams to talk to each other, but we use it because the ham radio TCP/IP
network can be tied to the worldwide TCP/IP network in a number of
different ways, and using the full set of TCP/IP address conventions
makes it possible for traffic to flow between the ham network and the
real world.

Leaving aside legal issues about third-party traffic, there's no
reason, for example, why electronic mail can't be automatically routed
through a gateway (a computer that interconnects two or more networks)

between a ham TCP/IP user and a non-ham who has access to the
Internet.  In fact, this service already exists in some areas.

The good news is that for traffic within the ham network, we only need
to worry about hostnames, and NOS's domain suffix command will take
care of adding the "ampr.org" extension for us; only if we want to go
across networks do we need to worry about the full details of
addressing.


TCP/IP Services

Now that we have those boring basics out of the way, the protocols
that use TCP/IP to provide real, useful services include:

TELNET The terminal emulation program.  In "real" networks, telnet
lets a user at one host remotely access a remote host, just as
if he was on a terminal directly connected to that computer. 
In NOS, the telnet function usually connects you to a remote
host's mailbox, which acts very much like a personal PBBS. 
The NOS telnet command does allow you to remotely login to a
host that supports that function; in some areas Unix computers
connected to the ham TCP/IP network provide that service.

FTP File Transfer Protocol.  A means of transferring both ASCII
(text) and binary (program, data, or compressed) files between
hosts.

SMTP Simple Mail Transfer Protocol.  A (mostly) invisible way of
moving electronic mail from one host to another.  If you
create a message on your computer (using the BM program,
discussed below), SMTP will automatically attempt to transfer
it to the destination computer.

POP Post Office Protocol.  SMTP is neat, but it's really designed
to work with hosts that are available full time.  Most ham
TCP/IP stations aren't.  POP is designed for them; it allows
incoming mail to be stored at a host that acts as a mail
server; when you come on the air, your system automatically
asks the server to send you your mail.

PING Packet InterNet Groper.  A diagnostic that sends a packet to a
specified host; if the host is accessible to you and on the
air, it responds with another packet.  PING tells you how long
the round trip took.

FINGER A way of finding out information about the users at a host. 
The finger command can simply list all the users at a host, or
spit out information (like the "brag tape" of RTTY days) about
a specific user.

ARP Address Resolution Protocol.  IP addresses need to be matched
with the correct hardware address (in our case, ham callsign)
to allow packets to be sent to their destination -- NOS
doesn't know what callsign goes with a given ip address.  ARP
does this by sending out a broadcast message when it needs to
know the callsign that matches an address.  The remote host
(if it's on the air) will answer and provide its hardware
address.

DNS Domain Name Service.  Remembering IP addresses isn't easy. 
NOS can use a file called domain.txt to contain mappings
between hostnames and IP addresses, but that means you need to
know the hostname and address of any station you want to
contact.  Alternatively, a remote host may agree to serve as a
domain name server that NOS can query when it needs to know
the address of a host.  Not all areas have a name server
available to the ham community, but in those that do, life is
a lot easier.


Installing NOS

Frankly, there's no completely painless way to get NOS running on your
computer.  NOS is somewhat picky about the directories used for its
files, and there are a number of custom parameters that you must set
to teach the program about your environment and your network.  Those
parameters are contained in a configuration file that most versions of
NOS call autoexec.net (PA0GRI versions call it autoexec.nos; our
references to "autoexec.net" mean either variety).


Files and Directories

You should create the following directories on your disk (NOS can work
from either a hard disk or a floppy; it's getting big enough, though,
that working from a 360K floppy can be tough):

/spool (holds NOS' working files)
/spool/help (help files for the mbox)
/spool/mail (mail messages go here)
/spool/mqueue (mail workfiles)
/spool/rqueue (incoming mail workfiles)
/finger (home for finger info files)
/public (file uploads/downloads)

Three files need to go in the root directory of your default disk:

autoexec.net (the NOS configuration file))
ftpusers (user ftp/mbox access)
domain.txt (hostnames if not using DNS)
bm.rc (mail program configuration)
alias (used by smtp and bm)

NOS uses two executable files.  These can be installed anywhere on
your file path:

nos.exe (grinos.exe) (main executable file)
bm.exe (mailer program)


Setting up AUTOEXEC.NET

Once the directories are created and the files copied, you need to
edit the autoexec.net file with a text editor to customize it.  A
sample file is included as Appendix B.  Some of the things you'll have
to put in the file are:

* Your hostname (usually your callsign in lower case):
hostname ag9v.ampr.org.

* Your IP address:  ip address [44.70.12.34]

* Your callsign (optionally including an SSID; local customs
vary on this):  ax25 mycall AG9V

* "attach" commands to tell NOS how to talk to your hardware. 
These can get quite hairy; Appendix C has the details.  For a
TNC on COM 1 at 4800 baud, use:  attach asy 0x3f8 4 ax25 ax0
1024 256 4800

The "ax0" in the middle of the command is the interface name -
- you use it to identify this port to NOS when you set up
routing commands and the like.  You can use any (short) name
you'd like, but the convention for COM ports is to use ax0,
ax1, etc.

* At least one routing command.  NOS needs to know where to send
packets.  A default route that sends all packets out the ax0
interface is:  route add default ax0

* If you have a gateway for packets going outside the local
area, include a route like:  route add [44.70.13.0]/24 ax0
ag9v

This command would route packets addressed to any host with
"44.70.13" as the first three bytes of its address out the ax0
interface to ag9v, which presumably knows how to get these
packets to their destination.  The "/24" means that the first
24 bits (three bytes) of the address are significant; NOS will
ignore the last byte when making routing decisions.

* If you have a domain name server, add a command near the
beginning of your configuration file identifying its IP
address:  domain addserver [44.70.12.34]

* If you have a local mail server that knows how to route
messages outside the area (see the discussion of electronic
mail, below), add a command identifying it:  smtp gateway
[44.70.12.34]


Storing Name/Address Matches in DOMAIN.TXT

If you don't have a local domain name server (DNS), you'll need to
create domain.txt in the root directory, with one entry for every
hostname you want to communicate with.  Appendix D shows how to set up
this file.  If you don't have an entry for a host in the file (or the
name server doesn't know about it), you can use the ip address instead
of the hostname in NOS commands.

If you're using DNS, NOS will save the hostname/address matches it
gets from the server in domain.txt, so you'll find that file existing
(and growing) even if you didn't create it.

Giving the Finger

If you want users to be able to learn about your station with the
finger command, you need to create a text file in the /finger
directory called <hostname>.txt (by the way, when we use angle
brackets, it means this is a value you'll need to insert yourself
-- minus the angles -- based on your own configuration).  You can
use any text editor to create the file; it should contain basic info
about your system.  Don't go overboard... one screen of text is plenty.

You can also create additional files with information about specific
aspects of your system.  For example, you might have a list of the
files available for downloading on your system in a finger file called
"filelist.txt."  A remote host who issues the command finger
filelist@<myhost> will get that list.


Some Boring but Necessary Technical Stuff

Before we move on to the good stuff about how to make NOS do magic, we
need to talk about three related commands that you may need to tweak
depending on local custom and the quality of the rf paths you're
using.  Just as regular AX.25 uses the paclen command to limit the
size of packets, TCP/IP has parameters defining how much data is moved
in one chunk.  In theory, the larger the datagram (TCP/IP's term for a
single block of data), the higher the efficiency, because the protocol
headers add a fixed amount of overhead; in larger datagrams the
overhead is a smaller percentage of the total data sent.

However, some networks (such as NetRom) can't handle large datagrams
in one piece.  More importantly, the larger the datagram, the longer
it takes to transmit, and on a busy or flaky path, the greater the
likelihood that something will corrupt it along the way.  And, it
takes longer to resend a large packet than a small one.  Because of
these factors, a fast network with clear channels and solid paths can
get away with sending much larger datagrams than a slow, unreliable one.

NOS provides three parameters that deal with datagram sizes.  The most
important one is the mtu (the sixth value in the attach asy command
described above).  It is similar to paclen; it sets the largest
packet, including any headers, that can be sent on an interface. 
Datagrams larger than the mtu are fragmented into multiple pieces,
which seriously reduces efficiency.  Each interface has its own mtu,
set as part of its attach command.

For 1200 baud channels that are shared with other packet users, an mtu
value of 256 is reasonable; in fact, since that is the largest packet
size most non-TCP/IP ham networks (like digipeaters and NetRom) are
designed to handle, 256 is the largest mtu you should specify if any
of your packets are going to move over such a node.

Faster networks may use higher values.  For good-quality channels with
fast data rates (9600 baud or above), it may be reasonable to use an
mtu ranging from 512 to 1500 (which matches the standard mtu used by
ethernet systems).

The other two parameters that set datagram size are part of the TCP
protocol.  tcp mss is the largest segment (yet another term that means
almost, but not quite, the same thing as "packet") that TCP will
assemble for transmission.  Because the TCP and IP headers attached to
each datagram total 40 bytes, mss should be 40 bytes smaller than mtu;
216 is the correct value for an mtu of 256.

The tcp window parameter tells NOS how many datagrams it can have
outstanding at once -- if it is twice the value of mss, NOS can
receive two datagrams before sending an acknowledgement.  This
parameter is analogous to the maxframe parameter in AX.25.

Although using a large window has major benefits on full duplex
networks, on typical ham networks best performance comes from smaller
windows ranging from one to three times the mss.  A good starting
point is to set the window equal to twice the value of mss (432 for
an mss of 216).

In summary, good starting points for these values are:

1200 baud, shared channel -- mtu 256, tcp mss 216, tcp window 432
High speed, clear channel -- mtu 1500, tcp mss 1460, tcp window 2920

It's very important that local users all agree on these values --
particularly the mtu and mss.  Mismatched sizes will result in
fragmented packets and a decrease in throughput.

Even more than in other parts of this manual, this discussion glosses
over lots of subtleties.  Throughput can be drastically affected by
tuning these values, and both experimentation and local consensus are
necessary to come up with settings that both work and don't stomp on
other users of the channel.

Using NOS

To run NOS, first make sure you have your TNC configured for KISS mode
(see Appendix F for details) and turned on.  Then, type nos (or grinos
for the PA0GRI version).  In a few seconds, you should see a net>
prompt.  Any error messages that appear first probably indicate a
problem with your autoexec.net file.

When you see the prompt, NOS is in command mode.  When you are
communicating with another host, NOS is in converse mode.  To return
to command mode from converse mode, press the F10 function key
(sometimes called the escape key).  All commands typed at the NOS
prompt need to be followed by the return key.

Typing ? in command mode will display a list of commands.  Typing a
command name followed by ? will display the valid subcommands.  You
can't really call it a help system, but it's better than nothing.

Some commands can be abbreviated to save typing; the degree of
abbreviation allowed depends on the command set of the NOS version
you're using.  Experimentation is the best way to see what works and
what doesn't.  One minor annoyance in grinos is that commands are case
sensitive -- "c ax0 n8acv" is fine,  but "C ax0 n8acv" isn't.  It's
safest to do all your NOS keyboarding in lower case -- apart from case
sensitive commands, in the email world, typing in capital letters is
considered shouting!

You can issue several commands from within NOS to deal with files and
directories.  pwd displays your current working directory, and cd
allows you to change directories.  dir displays files in the current
directory.  mkdir <dirname> creates a new directory, and rmdir
<dirname> removes one.  delete <filename> erases a file. 

You can also "shell out" to DOS from within NOS by entering either an
exclamation mark (!) or the command shell.  To return to NOS, type
exit at the DOS prompt.

From command mode, you can start a number of different types of
sessions to communicate with remote hosts.  Each session has its own
display screen and you can switch between a session and command mode,
or between sessions.  The se command displays the active sessions with
identifying numbers.  To switch to a session, you can type se <session
number>.  From command mode, you can return to the current (most
recently displayed) session by entering a carriage return.

You can capture incoming data from the current session to a disk file
by using the record <filename> command, and you can read in a data
file from disk with the upload <filename> command.

The most common NOS session types are probably telnet, its cousin
ttylink, ftp, and a regular packet connect (technically called an ax25
session).  Telnet is used to "login" to a remote host, ttylink is a
kind of telnet specially designed for keyboard-to-keyboard
communications, ftp handles file transfers, and ax25 sessions allow
you to carry on normal packet activity.  We'll talk about ax25
sessions first, since they give you a chance to test your setup
without having another TCP/IP station on the air.

AX.25 Mode

The connect command simply lets you do normal packet radio stuff. 
Establishing an ax25 connect through NOS is just like using the
standard TNC commands with a few small differences.  First, since NOS
can support several interfaces, each with its own hardware, you need
to tell NOS which one to use.

So, to connect to N8ACV on interface ax0, enter connect ax0 N8ACV. 
Once you get a Connected message, you'll be able to type to the
station at the other end just as you would with normal packet.  To
disconnect, press F10 to go back to command mode and type disconnect
at the prompt.  (Just as with a TNC, these commands can be
abbreviated; just how few of the letters are necessary will depend on
each implementation of NOS and the commands it supports).

The other minor difference between the NOS connect command and a
regular TNC is that the word via is not used when specifying
digipeaters.  To connect to N8ACV through N8KZA on interface ax0, you
would enter connect ax0 N8ACV N8KZA.


Telnet

The telnet command logs you in to a remote TCP/IP host; depending on
the capabilities of that host, you might find yourself chatting
directly with the user at the other end, connecting to the NOS
mailbox, mbox (which acts very much like a sophisticated personal
PBBS), or getting a Unix "login:" prompt.  To establish a telnet
session, enter telnet <hostname> at the command prompt.  To close a
session, press F10 to return to command and enter close <session
number>.  If there's only one session open, you can just enter close. 
You can also end the session by issuing the appropriate exit or quit
command at the remote host's prompt.

Some versions of NOS offer a new type of session that improves on
telnet for real-time keyboard-to-keyboard chats.  It's called ttylink,
and works just like telnet (for example, start a session with ttylink
<hostname>) except that it connects you directly to the remote host's
chat mode, and uses a split-screen format to make things less
confusing as you type to each other.

You'll get a message like "Telnet session 1 failed: Reset/Refused
errno 9" if the remote host doesn't support ttylink.  If the operator
at the other end isn't available to chat, you'll get a message like
"The system is unattended."  You'll still be able to type, but there
won't be anyone there to reply.  You can change the status on your
machine by setting the attended command either on or off.  You might
want to put this command in your autoexec.net file to set your default
status.  You exit from ttylink just as you would from telnet.

And now a note from Miss Manners:  you should never simply exit from
NOS when you have an open session.  Doing so can cause great
unpleasantness at the remote host.  Unless you're in some sort of
software or hardware lockup, or you know that the station on the other
end has gone away, always close sessions and wait for confirmation
before exiting the program.

You should also be aware that your system may have started sessions in
the background, for example to transfer electronic mail, or someone
else may have started a session with your system.  You may not even
know these sessions are running.  Pulling the plug on them would be
very impolite.  Before exiting NOS, you should first use the se
command to make sure there are no current sessions running, and then
the tcp status command to see if there are any background connections
established.  tcp status will show you a long and confusing list of
information; the important stuff at the end is the list of sockets
(which are services your system can either offer or request on the
network).  If anything other than Listening appears in the Status
column, that means there's at least one remote host communicating with
you.


File Transfers

You initiate a file transfer (ftp) session by entering ftp <hostname>
at the command prompt.  Once the session is established, the remote
host will prompt you for a user name and a password.  If your hostname
and password have been added to the remote host's ftpusers file,
you'll have the ability to download and perhaps upload files in the
directories permitted you.

If you haven't arranged with the remote host for your own account, you
can try to login as anonymous or guest; many systems support these
user names and grant limited (usually download-only) privileges to
them.  If you login under one of these accounts, you should enter your
hostname as the password; that allows the remote host to keep track of
who's been using the system.

Once you've logged in, you'll see a new prompt: ftp>.  This will
remind you that you're actually issuing commands to the remote
computer.  From the ftp> prompt, you can list the files in a
directory, change directories, upload files, or download files.

To list files, enter dir at the ftp> prompt.  You will get a listing
that shows subdirectories (if any) and files together with their dates
and sizes.  To show the current directory name, type pwd.  To change
directories, issue the cd <directory> command.  Note that directories
are displayed with a forward slash (/) instead of the usual MS-DOS
backslash (\).  That's because the Unix operating system, which is
TCP/IP's natural home, uses forward slashes.  If the remote host is
running NOS, you can use either character, but some other systems
(particularly those running Unix) will recognize only the forward
slash.

Once you've found a file you want to upload or download, you need to
make a decision.  ftp can transfer the file either as an image file,
byte for byte, or as an ASCII file, converting the line-end character
as necessary to compensate for different operating systems (Unix uses
only a linefeed character at the end of lines; MS-DOS uses carriage
return/linefeed).  Before beginning a file transfer, enter either type
i for an image file, or type a for an ASCII file, at the ftp> prompt.

What are the consequences choosing the wrong type?  Well, transferring
a binary file as type a will almost certainly fail.  Transferring an
ASCII file as type i will work, but you may find that the line-ends

are screwed up.  ASCII transfers are also quite a bit slower than
image, because each line needs to be processed separately.

To actually start a file transfer, use the command put <local
filename> <remote filename> to send a file, or get <remote filename>
<local filename> to receive one.  The file name can include a full
path if you desire; remember to use the proper path separator
character for the remote host.  If you only specify one filename, ftp
will assume that both the local and remote hosts will use the same
name.  This can be dangerous if the remote host uses a different
operating system than you do, as it may have filenames that are
illegal on your system.

If a file transfer goes awry, you can terminate it by going to command
mode via F10 and issuing the abort command.  To end an ftp session,
you can either type quit at the ftp> prompt (the preferred way), or
you can close the session from the net> prompt.

If you want others to be able to access files on your system, you'll
need to set up an ftpusers file in your root directory.  Appendix E
describes the contents of that file.

Another message from Miss Manners:  transferring files via ftp is
reliable, but can be slooooow, particularly at 1200 baud.  Before you
start downloading a 250 kilobyte file, consider how busy the channel
is, and whether you want to tie things up for (perhaps) several hours
by your download.  NOS is polite and won't hog the channel, but don't
doubt that a large file transfer will slow things down for everyone
else.


Other Protocols

The ping protocol mentioned above is very useful to see if a remote
host is on the air.  Just enter the command ping <hostname> at the NOS
prompt.  If the host is available, you will see a response indicating
what the round-trip time was to that host.  The time may be many
seconds if you're going through gateways, so be patient.

The finger protocol lets you see information about a remote host's
users and services.  Entering finger @<hostname> (note the slightly
different syntax -- the "@" symbol must immediately precede the remote
hostname) will display a list of the users (actually, finger files as
described above) at that host.  Entering finger <user@hostname> will
display the text file for that user.

Electronic Mail

We've saved NOS's electronic mail capabilities for last because they
are a bit more involved than some other parts of the program.  You use
two programs to handle mail:  bm (a mail agent, in Unix terms) to
write and read messages, and NOS to send and receive them.  First
we'll talk about reading and writing messages, and then about using
NOS to transport them.


Using BM.EXE to Read and Write Messages

bm.exe is a program that reads and writes mail message in the format
TCP/IP systems recognize.  Contrary to popular belief, "bm" stands for
"Bdale's Mailer" in honor of its creator, Bdale Garbee.  You can run
bm from the DOS prompt just like any other program, from within NOS by
shelling to DOS with ! or shell, or (in grinos) by typing the mail
command from the net> prompt.

Before using bm, you need to create its configuration file, bm.rc,
which must live in the root directory of your disk.  An annotated
bm.rc file is included as Appendix G.  Only the first three commands
in the sample file are absolutely necessary to make bm work.

There's a bit of controversy in some areas over the proper name to
enter for user.  Some folks recommend using either your first name, or
your initials (for example, my address would be "john@ag9v.ampr.org")
while other suggest using the callsign instead ("ag9v@ag9v.ampr.org").

While using the callsign may seem more impersonal, it has major
advantages when mail is moving between TCP/IP and the packet BBS
system, or when using the pop server; we strongly recommend that you
use the callsign@hostname format unless local objection is even
stronger.  It's important to be consistent within the area, so that
everyone knows how to address mail to everyone else.

When you start bm, you'll see a prompt such as ag9v> showing the
default mailbox (based on the user entry in bm.rc).  As in NOS, you
enter commands at the prompt, following them with a carriage return. 
Most bm commands are single letters, optionally followed by a mail
addressee, or a message number (or numbers).

To send mail, use the command m <addressee>.  The addressee will
normally be a user at a remote host; for example, ag9v might send mail
to k8gkh@k8gkh.  The single biggest problem with bm is forgetting to
include the hostname -- in other words, sending mail to <user> rather
than <user>@<hostname>.   Without the hostname, bm will think the user
is on your local system, and the mail will end up being stored in a
mailbox under that user's name on your own system.  That doesn't work
too well.

One way to solve that problem, and do some other interesting things,
is to create an alias file in your root directory.  When you send a
message, bm will compare the addressee with the alias file, and if it
finds a match will replace the alias with a full address from the
file.  An alias can point to a list of addresses, so it's possible to
define an alias that will send a copy of the message to everyone in
your local group.  A sample alias file might look like:

greg k8gkh@k8gkh.ampr.org
bill n8kza@n8kza.ampr.org
club k8gkh@k8gkh.ampr.org n8kza@n8kza.ampr.org
n8acv@n8acv.ampr.org wb8gxb@wb8gxb.ampr.org

The last alias demonstrates two things:  a single alias can expand to
several addresses, and you can continue a long address list on
subsequent lines by indenting them with spaces or a tab character.

Now, if you send mail to "greg" it will automatically be expanded to
the full address, and by sending a message to "club" all four users
will get a copy.

By the way, you do not use a trailing dot after an FQDN (as discussed
above) in email addressing; things won't work right if you do.

If you use bm's built-in editor to compose messages, remember that it
doesn't wrap lines; you have to hit the carriage return at the end of
each line.  You can list outbound mail with the l command; you can
kill an outbound message with the k <msg#> command (the message number
is obtained from the output of the l command).

Several commands are used to deal with incoming mail.  h displays the
headers (summary info) about messages in your mailbox.  It is the
basic command you should use to check your incoming mail.  Each
message header displayed will have a number to use with the other
message manipulation commands.  Commands given without a message
number act on the current message (the one displayed with a > by it in
the display from the h command; if there's only one message, it is
always the current one).

bm can support multiple users at a single host; a separate mailbox is
created for each user.  Unfortunately, bm has no way of knowing if
incoming mail addressed to <someuser>@<yourhost> is valid, so it will
happily accept such mail and create a new mailbox for <someuser>.  You
may never know it's there, unless you use the n command to display the
list of mailboxes.  You can also use n to change to a different
mailbox: n <mbox>.

The commonly used commands (which may be followed by one or more
message numbers if appropriate) are:

msg#    message number by itself will display that message and set it
        as the current message.
r reply to a message.
d delete a message.
s       save a message; if a file name follows the message number(s),
        the message(s) will be saved in that file.  Otherwise, they'll
        be saved in the default mbox file.
u undelete a message previously marked for deletion.
p print a message on the local printer.
w save a message to a file without including headers.
f forward a message to another recipient.
b       bounce a message.  Like forward, but keeps the original sender
        information intact (i.e., the message will not appear to have
        been sent by you).
$       update the mailbox.  This deletes messages marked for deletion
        and reads in any new mail that may have arrived since you started bm.

There are two commands that exit from bm.  x will exit without
updating the mailbox.  In other words, the same messages will be there
the next time you run the program.  q updates the mailbox (like $) and
then exits.


Moving Mail With NOS

Now, to the mechanics of getting mail into and out of your system. 
All mail that you create is sent to its destination (or at least to
the next stop on the way) by the smtp server in NOS.  The smtp timer
command (set in autoexec.net) tells smtp how often to scan the
/spool/mqueue directory for outgoing mail.  When it finds some, it
attempts to open an smtp session to the remote host in the address and
send the mail there.  There's no default for the smtp timer value, so
your autoexec.net file should include something like smtp timer 600
(which scans for mail every ten minutes).  You can manually force smtp
to scan the queue by issuing the smtp kick command from the net>
prompt.  If you have a local mail server with connections to the
outside world, you can use it to route mail for hosts that aren't in
your domain file with the smtp gateway <hostid> command.

Incoming mail can arrive at your station when a remote host does this
and starts an smtp session with you.  But if you don't keep your
station up 24 hours a day, the remote host will be trying, and trying,
and trying, to connect with you until you finally show up.  A far
better approach is to use pop -- the Post Office Protocol.  If your
system runs pop, and someone in the area has agreed to be a pop
server, NOS will automatically contact that server when you come on
the air; the server will respond by sending the mail waiting in your
mailbox.  You can then read it with bm just as if it had arrived via
smtp.

To use pop, the server must establish a mailbox and password for you,
and you need to add the appropriate commands to your autoexec.net file
(see the annotated autoexec.net file in Appendix B).

Remember that smtp or pop sessions may be running in the background
without your knowing about it.  Always check for activity with the tcp
status command before pulling the plug!

Additionally, smtp creates lock files in \spool\mqueue when it tries
to send outgoing mail.  If NOS is killed before the mail transfer has
succeeded, these files (with the extension ".LCK") will be left behind
and if they are not manually removed, they will prevent smtp from
trying again to send those messages.  To prevent this, you should
always issue the command "erase \spool\mqueue\*.LCK" before starting
NOS.  It's a good idea to launch NOS using a batch file that removes
the locks before executing the program.

Conclusion

This has been a whirlwind tour of TCP/IP.  Once you have the software
installed, using it is really quite simple, and NOS truly opens the
door to using packet radio in a whole new way.

To learn the subtleties of NOS, you should do two things:  read the
reference manual for the version you're using, and experiment with the
program.  Once you know the ins and outs, please share your knowledge
with others.  The ham radio TCP/IP community is still small, and we
need all the Elmers we can get!


John Ackermann   AG9V
2371 Stewart Road
Xenia, OH   45385

TCP/IP: ag9v.ampr.org.  [44.70.12.34]
PBBS: AG9V@N8ACV.OH.US.NA
Internet: jra@lawday.daytonOH.ncr.com
CompuServe: 72300,1160


APPENDIX A
Resources for NOS and TCP/IP

(Note:  This is a very incomplete list; please feel free to provide
additional resources to add for the next edition!)

TAPR
P.O. Box 22888
Tucson, AZ   85734

The New England TCP Association
3628 Acushnet Ave.
New Bedford, MA   02745

APPENDIX B
Sample AUTOEXEC.NOS File for GRINOS

# A U T O E X E C . N E T
# This is a sample autoexec file for GRINOS version N1BEE 0.72.
# It doesn't have all the fancy features one might hope for, but
# the basics are there, with some hopefully useful comments.
# Any line beginning with a "#" character is treated as a comment.  To
# uncomment a line, delete the # character

# These are a couple of things for NOS to use internally.
mem eff on
watchdog on
nibufs 10

# NOS needs to know three things about you:  your hostname, your
# ham callsign, and your ip address.  By convention, the hostname
# is your callsign in lower case, followed by ".ampr.org".  The
# callsign is generally used in upper case to distinguish it.  The
# ip address comes from a local area coordinator.  Note that there are
# a minimum of three places in this file where you need to insert your IP
# address -- here, in the ifconfig command, and at the end of each attach
# command.
hostname nocall.ampr.org
ax25 mycall NOCALL
ip address [44.xx.xx.xx]

# This should match your ip address
ifconfig loopback ipaddress [44.xx.xx.xx]

# This makes short forms of the hostname work.
domain suffix ampr.org.

# NOS needs to know how to convert hostnames to ip addresses.
# You can do this manually via the "DOMAIN.TXT" file, or you can
# use a nameserver if one is available.  To enable the
# nameserver, uncomment this line and plug in its correct
# address.
#domain addserver [44.xx.xx.xx]

# Some additional commands for the domain service.  Don't turn translate
# on unless you have a small domain file and/or a fast machine.
domain verbose off
domain cache size 40
domain translate off

# To use POP, uncomment these lines.  Fill in "pop mailhost" with the
# ip address of the station serving as your pop server.  Fill in the
# "pop mailbox" name with your hostname, i.e., your call.  The "pop
# userdata" line needs to have your hostname, followed by a password
# (as negotiated with your mail server).  "pop timer" sets
# how often, in seconds, to query for mail.
#pop mailhost [44.xx.xx.xx]
#pop mailbox hostname
#pop userdata hostname password
#pop timer 1800

# Attach commands are complex; these are samples for COM 1
# and 2.  See Appendix C for details.  Uncomment the
# appropriate line(s) for your hardware.
# COM1 -- 256 byte MTU, 4800 baud serial link as ax0
attach asy 0x3f8 4 ax25 ax0 2048 256 4800
# COM2 -- 256 byte MTU, 4800 baud serial link as ax1
#attach asy 0x2f8 3 ax25 ax1 2048 256 4800

# This is the basic route, sending everything out ax0
route add default ax0

# These are tcp parameters you shouldn't need to mess with.
ip ttl 16
ip rtimer 240
tcp irtt 3000

# On a shared channel, you may want to change timertype to
# exponential; that's more courteous, but will slow your
# retries down significantly.  mss and window should ordinarily be the
# same value, equal to the largest mtu set in the attach command(s) above
# minus 40.  With the common mtu for 1200 baud channels of 256, that means
# both mss and window should be 216.
tcp timertype linear
tcp bblimit 16
tcp mss 216
tcp window 216

# These set up AX.25 parameters
ax25 digipeat off
ax25 maxframe 1
ax25 paclen 256
ax25 retry 20
ax25 window 4096
ax25 blimit 15
ax25 version 2

# as with tcp timertype, you may want to set this to
# exponential on a shared channel.
ax25 timertype linear

# These are netrom setup commands.  Don't turn them on
# unless you need them, and you know what you're doing.  You
# can really screw up the network by putting out netrom
# broadcasts that don't fit with the configuration of the
# "real" netrom nodes that can hear you.
#attach netrom
#netrom interface ax0 MYALIAS 192
#netrom obsotimer 1800
#netrom nodetimer 10800
#netrom verbose yes
#netrom bcnodes ax0
#netrom ttl 8

# These start the servers.
start smtp
start ftp
start echo
start discard
start telnet
start finger

# Uncomment this line to enable logging.
#log \spool\net.log

# Default file type for ftp transfers.  Type image is for binary
# files; type ascii is for text; it's safest to set the default to image.
ftype image

# This makes telnet sessions to Unix systems work
# line-by-line, rather than character-by-character.
echo refuse

# Tell smtp how often to scan for outgoing mail
smtp timer 600
smtp batch on

# grinos can send a string of commands to the TNC on startup.  You
# could use this to force the TNC into KISS mode.  Note that you need to
# specify which interface to use.  This must be done <after> defining the
# interface and <before>any data is sent to the TNC (for example, by the
# smtp and pop kick commands below) These commands will do that for a TNC2:
#comm ax0 "kiss on"
#comm ax0 "reset"

# kick the smtp and pop servers at startup.  Only uncomment the "pop
# kick" line if you've defined a pop server above.
smtp kick
#pop kick

# GRINOS (but not other versions) can define the function
# keys with macros to make things a bit easier.  Here are a
# couple of examples.  Note that each command must end with
# a "\n" to signify a carriage return.  The numbers
# represent the keys; 59 - 68 for F1- F10 (though F10 can't
# be redefined; it's always the escape key), 84 - 93 for
# shiftF1 - shift F10, 94 - 103 for ctrlF1 - ctrlF10, 104 -
# 113 for altF1 - altF10.
fkey 59 "tcp status\n"
fkey 60 "mem status\n"
fkey 61 "status\n"

# THE END

APPENDIX C
Designing ATTACH Commands

NOS supports a number of versions of the attach command to deal with
different hardware.  We'll discuss three of them here:  asy, used for
serial port connections; pi, used to connect to the Ottawa PI card;
and packet, used to interface to hardware supporting the FTP, Inc.,
packet driver protocol.  As usual, this discussion covers the basics;
see the NOS reference manual for details on all the many options.

Hosts normally have a separate ip address for each interface.  If you
are running more than one interface, you can include that interface's
ip address (in [xx.xx.xx.xx] form) at the end of the attach command.

The asy version provides an interface to a standard PC serial port. 
The syntax is:

attach asy <ioaddr> <vector> <mode> <if> <bufsize> <mtu> <speed>

In English, these parameters are:

ioaddr -- the address of the COM port being used.  COM1 is
usually 0x3f8 and COM2 is usually 0x2f8.  COM3 and COM4 aren't
standardized; using them will require looking at the documentation
for your serial card, and probably some experimentation.

vector -- the IRQ used by the hardware.  COM1 is usually 4,
and COM2 is usually 3.  Again, COM3 and COM4 vary.

mode -- this specifies the nature of the interface.  ax25 is
for a connection to a KISS TNC, slip for a hardwired
connection to another host, ppp for a dial-up connection, and
nrs is for attaching a NOS station to a NetRom node.

if -- the interface name.  The convention is to use ax0, ax1,
etc., for KISS interfaces.

bufsize -- the buffer for incoming data, in bytes.  Usually a
value of 1024 is more than sufficient for a 1200 baud channel.

mtu -- the maximum transmission unit size, in bytes.  See the
discussion in the main text on this subject.

speed -- the speed of the serial (not radio) link, in baud. 
The best setting for this will depend on the speed of your
computer, but generally two to four times the radio speed is
adequate.

Some sample attach asy commands are:

# COM1, KISS TNC as ax0, MTU 256, 4800 BAUD
attach asy 0x3f8 4 ax25 ax0 1024 256 4800

# COM2, KISS TNC as ax1, MTU 256, 2400 BAUD
attach asy 0x2f8 3 ax25 ax1 1024 256 2400

# SLIP link, COM1 as sl0, MTU 256, 9600 BAUD
attach asy 0x3f8 4 slip sl0 1024 256 9600

The Ottawa PI card is a plug-in board for PCs designed for high-speed
performance.  It has two ports, one DMA driven for high speed and the
other interrupt driven.  The attach syntax is:

attach pi <ioaddr> <vector> <DMA chn> <mode> <name> <bufsize> <mtu>
<speed a> <speed b>

A sample attach command (using the PI's default jumper settings) is:

attach pi 380 7 1 ax25 pi0 1750 1024 0 1200

In this example, the interface name for the DMA port is "pi0a" and the
second port is "pi0b".  Because the port a speed is 0, the PI card
expects the modem to provide its own clocking.  The PI attach syntax
is explained in the manual provided with the card.

Finally, the packet interface is used to connect to ethernet cards and
other hardware that supports the FTP, Inc. "packet driver" standard. 
There's a packet driver for the PI card.  The syntax is:

attach packet <ioaddr> <vector> <if> <bufsize> <mtu>

In this case, ioaddr and vector need to match those used for the
packet TSR that supports the hardware.  bufsize is the number of
packets (not bytes) that may be outstanding.  For ethernet, the
standard mtu is 1500.

Although most of the attach commands support the inclusion of an ip address
unique to that interface, we've noticed that some (particularly the packet
driver) don't work.  If you need to assign an address to an interface, you can
use the following command <after> the attach statement:

ifconfig <name> ipaddr [xx.xx.xx.xx]

APPENDIX D
The DOMAIN.TXT File


# The domain.txt file contains mappings between hostnames
# and ip addresses.  The file can be quite complex, but
# basic entries usually resemble this.

# Fields are separated by tabs or spaces.

# These are normal address records.  The first field is the
# hostname.  The second field is a "time to  live" value
# returned by the name server.  If you manually create an
# entry, you can leave this field  blank.  The third field
# is always "IN" to signify these are internet addresses. 
# The fourth field is "A" to signify an address record.  The
# last field is the address.

k8gkh.ampr.org. 3599886 IN A 44.70.12.31
ag9v.ampr.org. 3585524 IN A 44.70.12.34

# This is a "canonical name" (CNAME) record that maps an
# alias to an official hostname.

server.ampr.org. 3599886 IN CNAME ag9v.ampr.org.

APPENDIX E
Sample FTPUSERS File

# This file establishes ftp user permissions.  Fields are
# separated by exactly one space.  The privileges value is a
# bitmask.  The only values significant for ftp are:

# 1 - read only
# 3 - read/write
# 7 - read/write/overwrite/delete


anonymous * /pub 1 # no password, read only in /pub
friend foobar /pub 3 # read/write privileges in /pub
spouse snoogums / 7 # read/write/delete everywhere

APPENDIX F
Making Your TNC Talk in KISS MODE

Once NOS is installed and your configuration files set, you need to do
one more thing:  get your TNC talking to your computer in KISS (Keep
It Simple, Stupid) mode.  KISS is a special protocol that lets your
computer do the work of processing packets; the TNC does only the very
low-level packet assembly and disassembly functions.  Nearly all TNCs
support KISS in one way or another.

Typically, you'll need to issue commands to the TNC to set the serial
line baud rate to the same speed as you've specified in the attach
command, to 8 bit data, and to no parity.  Then, issue the KISS
command (on a TNC2, kiss on), and the TNC's software reset command. 
After that, you won't be able to talk to your TNC via the terminal
program, but NOS will be able to.  (And don't worry, you can easily
return the TNC to normal mode if you want to.)  Once you've done this,
you're set to run NOS.

One trick that grinos supports is the ability to send commands to the
TNC during startup.  The comm command will send a string of text to
the named interface.  For example, to force a Kantronics DataEngine or
KAM into KISS mode every time you start NOS, include the following
commands in autoexec.nos (after you've defined the interface with the
attach command):

comm ax0 "interface kiss"
comm ax0 "reset"

Note that surrounding the text with double quote characters will
preserve spaces in the command.

Appendix G
A Sample BM.RC File


# bm.rc
# your hostname -- note that for mail we <don't> put a trailing period
# at the end of the FQDN.
host ag9v.ampr.org

# the user name (one host can receive mail for several users); usually
# your callsign
user ag9v

# your full name, for the message "From:" line
fullname John Ackermann

# if you want to have replies sent to another host, because, for
# example, you are using a pop server, this line specifies where replies
# should go
reply ag9v@ag9v.ampr.org

# for faster screen writes on the pc, use direct video, not bios
screen direct

# if you want to use an editor different than bm's built-in one
edit ed

# put saved messages here; note "/" instead of "\"
mbox c:/folder/mbox

# save a copy of outbound mail here
record c:/folder/outmail

# folder for your mail
folder c:/folder

# maximum number of messages that can be pending
maxlet 200