Protocols spoken by APRX

There are a few protocols spoken by the APRX:

      1) APRS-IS interchange

      2) Proprietary monitoring via shared memory segment

      3) APRX-APRXC linkage


APRS-IS interchange

	This protocol is fairly simple.

	    http://wiki.ham.fi/IGate_properties



			APRS iGate Common Properties

	Communication is over TCP streams of text lines terminating
        at CR+LF pair.    No CR or LF is permitted inside the text
        line, however.

	APRS-IS user authentication:

	    user USERID pass PASSCODE vers VERS STRINGS  \
	    	  filter FILTER STRINGS

        where 'USERID' is uppercase callsign plus possible "-SSID" tail.
	The 'PASSCODE' is calculated with semi-secret algorithm out of
	the uppercase callsign characters without '-' and SSID tail.
	The 'VERS STRINGS' are free to be anything, except string 'filter',
	and the 'FILTER STRINGS' are explained in document:
	   http://www.aprs-is.net/ServerCmds.htm
	   http://www.aprs-is.net/javAPRSSrvr/javaprsfilter.htm


	Messages:

	Lines beginning with '#' character are line noise (usually
	something that the server replies..)


	After successfull login, communication carries "TNC2" format
	APRS messages.  Namely text encoding of AX.25 UI frames in
	what became known as "TNC2 monitor style":

	    SOURCE>DESTIN:payload
	    SOURCE>DESTIN,VIA,VIA:payload

	The SOURCE, DESTIN, and VIA fields are AX.25 address fields,
        and have "-SSID" value annexed if the SSID is not zero.
	Also in VIA-fields, if the "HAS BEEN DIGIPEATED" bit is set
	(AX.25 v2 protocol feature) a star ('*') character is appended.
        VIA-fields are separated by comma (',') from DESTIN, and each
        other.

	A double-colon (':') separates address data from payload.
	The payload is passed _AS_IS_ without altering any message
	content bytes, however ending at first CR or LF character
	encountered in the packet.


			APRS-RX iGate Basic Rules


	Packets with source addresses:  NOCALL*, N0CALL*, WIDE*, TRACE*,
        TCP*, are dropped and not relayed to APRS-IS.

	Packets with VIA addresses: RFONLY, NOGATE, TCPIP, TCPXX are not
        to be relayed to APRS-IS.  Sometimes the VIA fields have '*' on
	their tails.

	Packets with payload beginning with character '?' are not to be
	relayed to APRS-IS.

	Packets with payload beginning with character '}' are so called
	3rd-party frames, and they are to be re-processed starting from
	character following the '}' character.


	When packet is sent to APRS-IS, the address gets appended
	either a "q-construct", or equivalent:

                ,qAR,gatecallsign
		,gatecallsign,I

	The "qAR" et.al. are explained at: http://www.aprs-is.net/q.aspx


			APRS-TX iGate has additional rules:

	All rules:
	    http://www.aprs-is.net/IGateDetails.aspx

	Specifically forbidden to relay to RF is "qAX", maybe some others too.
	(See: http://www.aprs-is.net/q.aspx)

	If the packet VIA-path received from APRS-IS contains TCPXX,
	NOGATE, RFONLY, then the packet is to be dropped, and not relayed
	to radio network.  (Note: TCPIP is permitted.)

	Packets relayed from APRS-IS to radio must use so called 3rd-party
	format.  Signature is '}' character.

	Rules on re-sending recently heard packets are a bit more complex,
	and are covered adequately in above referenced document.





Proprietary monitoring mechanism

	There is a tool for monitoring channel activity.
	See  aprx-stat(8)



APRX-APRXC linkage

	(For a feature in planning...)



	With introduction of APRX-Cluster (APRXC) mode, multiple APRX
	instances are used as remote attachments to TNCs, and one central
	system runs more complicated business logic deciding which messages
	to pass where, what beacons to send, etc.  The central business-logic
	server runs also connection with APRS-IS, and all things that it
	implies.

	Actually the APRX-Cluster-Server is APRX program version with
	embedded Perl for business logic.  Some very flexibly configurable
	APRS digi software has practical limitations of what the config
	can do.

	Environmental pre-requisite: APRXC and APRX nodes are in NTP sync!

	The Protocol:

	Communication is of text lines over TCP stream, they are canonic
	Internet format ending with CRLF pairs.  However no embedded CR,
	nor LF is permitted.

	(detail under study: UDP frames, or SCTP SEQPACKET ?)

	The protocol is bidirectional, and is intended to be connected from
	edge systems to cluster server.  Upon receiving a connection the
	server sends a greeting containing time-varying string.  This will
	be used by the connecting party to do authentication in APOP-style.

	The timestamps in this protocol are "U" format:
	    sprintf(timestamp, "U%ld", (long)time(NULL));
	which is to say, integer presentation of UNIX internal time in whole
	seconds.


	In following ">>" mean data sent by central system and received by
	the edge, whereas "<<" means data sent by edge system to central
	system.

		<< (connection formation from edge to the server)

		>> <timestamp> Hello <++counter> some greeting string

		<< <timestamp> LOGIN <userid> <authenticator>

		>> <timestamp> OK

		<< <timestamp> SERVICE <ifname> <speed> { RX | TX <areaspecifiers> }

		>> <timestamp> OK

	The <authenticator> is formed by hex (lower case) encoding of 16 byte
	MD5() checksum over the whole greeting string (sans CRLF), plus userid,
	plus shared plaintext password.

	Replies can also be "FAIL", if "LOGIN" or "SERVICE" is somehow bad.

	The  <ifname> is local interface name in given node.  The <speed> is
	radio channel bitrate.  The <areaspecifiers> are filter statements
	in APRS-IS style.


	The edge system compares every received timestamp (aside of the
	login-sequence above) with its internal time reference, and if
	the time varies more than 3 seconds from expected, the received
	frame is discarded.

	If APRXC Server notes that the time stamp it received is more than
	3 seconds in future or in past, it rejects the message.
	(This does not apply to ERLANG verbs.)

	The APRXC server acknowledges every received message with:

		>> <timestamp> OK

	Link idle jabber is by done by both systems by sending "TIME" verbs.
	They are sent every 20 seconds since previous transmit of anything
	on the connection, and if either notes that link has not received
	anything for 120 seconds, the link is considered to be down, closed,
	and re-initiated.

		<< <timestamp> TIME
		>> <timestamp> OK

		>> <timestamp> TIME
		<< <timestamp> OK

	Either system can initiate the "heartbeat", sometimes even both
	may send the TIME verbs at the same time, but both are not required
	to do so.


	The edge system sends received APRS frames to the central system
	in "TNC2" format in a command:

		<< <timestamp> APRS <ifname> <TNC2frame>

	The <TNC2frame> is the data received from radio as is without any
	APRS-IS -type additional parameters inserted into address field.


	When central system wants that a TX capable edge system transmits
	something, it sends following message:

		>> <timestamp> APRSTX <ifname> <TNC2frame>

	the edge system will transmit the frame as is (the APRXC will format
	the message to be ready for transmit.)  If the edge system can not
	transmit on given interface, it reports:

		<< <timestamp> FAILTX <ifname>

	which results from APRXC and/or APRX software bug listing/flagging
	tx-interface to be non-tx-interface..  Otherwise it acks the packet
	with "OK" message.

		<< <timestamp> OK

	Edge systems send 1 minute ERLANG data to central system with verb:

		<< <timestamp> ERLANG <ifname>  <rxbytes> <rxpackets> \
		              <txbytes> <txpackets> [ <rxerlang> <txerlang> ]
		>> <timestamp> OK

	(Long line is folded in UNIX-style in this documentation.
	 Data is ACKed with OK.)

	The APRXC collates these to 1 minute, 10 minute and 60 minute datasets
	of given timestamp.  The measured erlang-values are optional, if edge
        system can produce truthfull ones.

	The time-series collation database is persistent, for example RRD.