Protocols

Protocols process data on behalf of an Interface. They can modify the data being written, data being read, or both. Protocols can also mark a packet as stored instead of real-time which means COSMOS will not update the current value table with the packet data. Protocols can be layered and will be processed in order. For example, if you have a low-level encryption layer that must be first removed before processing a higher level buffer length protocol.

Protocol Run Order

Read protocols execute in the order specified (First specified runs first). Write protocols execute in the reverse order (Last specified executes first).

Protocols are typically used to define the logic to delineate packets and manipulate data as it written to and read from Interfaces. COSMOS includes Interfaces for TCP/IP Client, TCP/IP Server, Udp Client / Server, and Serial connections. For 99% of use cases these Interfaces should not require any changes as they universally handle the low-level details of reading and writing from these types of connections. All unique behavior should now be defined in Protocols.

At a minimum, any byte stream based Interface will require a Protocol to delineate packets. TCP/IP and Serial are examples of byte stream based Interfaces. A byte stream is just a simple stream of bytes and thus you need some way to know where packets begin and end within the stream.

TCP/IP is a friendly byte stream. Unless you are dealing with a very poorly written system, the first byte received on a TCP/IP connection will always be the start of a packet. Also, TCP/IP is a reliable connection in that it ensures that all data is received in the correct order, that no data is lost, and that the data is not corrupted (TCP/IP is protected by a CRC32 which is pretty good for avoiding unrecognized data corruption).

Serial is a much less friendly byte stream. With serial connections, it is very likely that when you open a serial port and start receiving data you will receive the middle of a message. (This problem is only avoided when interfacing with a system that only writes to the serial port in response to a command). For this reason, sync patterns are highly beneficial for serial interfaces. Additionally, serial interfaces may use some method to protect against unrecognized data corruption (Checksums, CRCs, etc.)

UDP is an inherently packet based connection. If you read from a UDP socket, you will always receive back an entire packet. The best UDP based Protocols take advantage of this fact. Some implementations try to make UDP act like a byte stream, but this is a misuse of the protocol because it is highly likely that you will lose data and have no way to recover.

Packet Delineation Protocols

COSMOS provides the following packet delineation protocols: Burst, Fixed, Length, Template, Terminated and Preidentified. Each of these protocols has the primary purpose of separating out packets from a byte stream.

Note that all protocols take a final parameter called "Allow Empty Data". This indicates whether the protocol will allow an empty string to be passed down to later Protocols (instead of returning :STOP). Can be true, false, or nil, where nil is interpreted as true unless the Protocol is the last Protocol of the chain. End users of a protocol will almost always simply leave off this parameter. For more information read the Custom Protocols documentation.

Burst Protocol

The Burst Protocol simply reads as much data as it can from the interface before returning the data as a COSMOS Packet (It returns a packet for each burst of data read). This Protocol relies on regular bursts of data delimited by time and thus is not very robust. However, it can utilize a sync pattern which does allow it to re-sync if necessary. It can also discard bytes from the incoming data to remove the sync pattern. Finally, it can add sync patterns to data being written out of the Interface.

Parameter	Description	Required	Default
Discard Leading Bytes	The number of bytes to discard from the binary data after reading. Note that this applies to bytes starting with the sync pattern if the sync pattern is being used.	No	0 (do not discard bytes)
Sync Pattern	Hex string representing a byte pattern that will be searched for in the raw data. This pattern represents a packet delimiter and all data found including the sync pattern will be returned	No	nil (no sync pattern)
Fill Fields	Whether to fill in the sync pattern on outgoing packets	No	false

Fixed Protocol

The Fixed Protocol reads a preset minimum amount of data which is necessary to properly identify all the defined packets using the interface. It then identifies the packet and proceeds to read as much data from the interface as necessary to create the packet which it then returns. This protocol relies on all the packets on the interface being fixed in length. For example, all the packets using the interface are a fixed size and contain a simple header with a 32-bit sync pattern followed by a 16 bit ID. The Fixed Protocol would elegantly handle this case with a minimum read size of 6 bytes. The Fixed Protocol also supports a sync pattern, discarding leading bytes, and filling the sync pattern similar to the Burst Protocol.

Parameter	Description	Required	Default
Minimum ID Size	The minimum number of bytes needed to identify a packet. All the packet definitions must declare their ID_ITEM(s) within this given number of bytes.	Yes
Discard Leading Bytes	The number of bytes to discard from the binary data after reading. Note that this applies to bytes starting with the sync pattern if the sync pattern is being used.	No	0 (do not discard bytes)
Sync Pattern	Hex string representing a byte pattern that will be searched for in the raw data. This pattern represents a packet delimiter and all data found including the sync pattern will be returned.	No	nil (no sync pattern)
Telemetry	Whether the data is telemetry	No	true (false means command)
Fill Fields	Whether to fill in the sync pattern on outgoing packets	No	false
Unknown Raise	Whether to raise an exception for an unknown packet	No	false

Length Protocol

The Length Protocol depends on a length field at a fixed location in the defined packets using the interface. It then reads enough data to grab the length field, decodes it, and reads the remaining length of the packet. For example, all the packets using the interface contain a CCSDS header with a length field. The Length Protocol can be set up to handle the length field and even the length offset CCSDS uses. The Length Protocol also supports a sync pattern, discarding leading bytes, and filling the length and sync pattern similar to the Burst Protocol.

Parameter	Description	Required	Default
Length Bit Offset	The bit offset from the start of the packet to the length field. Every packet using this interface must have the same structure such that the length field is the same size at the same location. Be sure to account for the length of the Sync Pattern in this value (if present).	No	0 bits
Length Bit Size	The size in bits of the length field	No	16 bits
Length Value Offset	The offset to apply to the length field value. The actual value of the length field plus this offset should equal the exact number of bytes required to read all data for the packet (including the length field itself, sync pattern, etc). For example, if the length field indicates packet length minus one, this value should be one. Be sure to account for the length of the Sync Pattern in this value (if present).	No	0
Bytes per Count	The number of bytes per each length field 'count'. This is used if the units of the length field is something other than bytes, e.g. if the length field count is in words.	No	1 byte
Length Endianness	The endianness of the length field. Must be either 'BIG_ENDIAN' or 'LITTLE_ENDIAN'.	No	'BIG_ENDIAN'
Discard Leading Bytes	The number of bytes to discard from the binary data after reading. Note that this applies to bytes including the sync pattern if the sync pattern is being used. Discarding is one of the very last steps so any size and offsets above need to account for all the data before discarding.	No	0 (do not discard bytes)
Sync Pattern	Hex string representing a byte pattern that will be searched for in the raw data. This pattern represents a packet delimiter and all data found including the sync pattern will be returned.	No	nil (no sync pattern)
Max Length	The maximum allowed value in the length field	No	nil (no maximum length)
Fill Length and Sync Pattern	Setting this flag to true causes the length field and sync pattern (if present) to be filled automatically on outgoing packets.	No	false

The most confusing aspect of the Length Protocol is calculating the Length Value Offset. This is especially true in the commonly used CCSDS Space Packet Protocol. The best way to illustrate this is with an example. Suppose you have CCSDS Space Packets prepended with a Sync Pattern of 0x1ACFFC1D. This would look like the following:

Sync (4 bytes)	Header (4 bytes)	Length (2 bytes)	Data (4 bytes)
0x1ACFFC1D	0x0001CADB	0x0003	0xDEADBEEF

In this case the total length of the packet is 14 bytes: 4 + 4 + 2 + 4 = 14. With 4 bytes of data, the length field is 3 because in CCSDS the length field is calculated as (data length - 1). So how would we calculate the Length Value Offset? COSMOS reads all the bytes in the packet (including the Sync Pattern) so the total length is 14 bytes. The length field is 3 so the Length Value Offset (offset to apply to the length field value) should be 11 (3 + 11 = 14).

Terminated Protocol

The Terminated Protocol delineates packets using termination characters found at the end of every packet. It continuously reads data until the termination characters are found at which point it returns the packet data. For example, all the packets using the interface are followed by 0xABCD. This data can either be a part of each packet that is kept or something which is known only by the Terminated Protocol and simply thrown away.

Parameter	Description	Required	Default
Write Termination Characters	The data to write after writing a command packet. Given as a hex string such as 0xABCD.	Yes
Read Termination Characters	The characters which delineate the end of a telemetry packet. Given as a hex string such as 0xABCD.	Yes
Strip Read Termination	Whether to remove the read termination characters before returning the telemetry packet	No	true
Discard Leading Bytes	The number of bytes to discard from the binary data after reading. Note that this applies to bytes including the sync pattern if the sync pattern is being used.	No	0 (do not discard bytes)
Sync Pattern	Hex string representing a byte pattern that will be searched for in the raw data. This pattern represents a packet delimiter and all data found including the sync pattern will be returned.	No	nil (no sync pattern)
Fill Fields	Whether to fill in the sync pattern on outgoing packets	No	false

Template Protocol

The Template Protocol works much like the Terminated Protocol except it is designed for text-based command and response type interfaces such as SCPI (Standard Commands for Programmable Instruments). It delineates packets in the same way as the Terminated Protocol except each packet is referred to as a line (because each usually contains a line of text). For outgoing packets, a CMD_TEMPLATE field is expected to exist in the packet. This field contains a template string with items to be filled in delineated within HTML tag style brackets "<EXAMPLE>". The Template Protocol will read the named items from within the packet and fill in the CMD_TEMPLATE. This filled in string is then sent out rather than the originally passed in packet. Correspondingly, if a response is expected the outgoing packet should include a RSP_TEMPLATE and RSP_PACKET field. The RSP_TEMPLATE is used to extract data from the response string and build a corresponding RSP_PACKET. See the TEMPLATE target within the COSMOS Demo configuration for an example of usage.

Parameter	Description	Required	Default
Write Termination Characters	The data to write after writing a command packet. Given as a hex string such as 0xABCD.	Yes
Read Termination Characters	The characters which delineate the end of a telemetry packet. Given as a hex string such as 0xABCD.	Yes
Ignore Lines	Number of response lines to ignore (completely drop)	No	0 lines
Initial Read Delay	An initial delay after connecting after which the interface will be read till empty and data dropped. Useful for discarding connect headers and initial prompts.	No	nil (no initial read)
Response Lines	The number of lines that make up expected responses	No	1 line
Strip Read Termination	Whether to remove the read termination characters before returning the telemetry packet	No	true
Discard Leading Bytes	The number of bytes to discard from the binary data after reading. Note that this applies to bytes including the sync pattern if the sync pattern is being used.	No	0 (do not discard bytes)
Sync Pattern	Hex string representing a byte pattern that will be searched for in the raw data. This pattern represents a packet delimiter and all data found including the sync pattern will be returned.	No	nil (no sync pattern)
Fill Fields	Whether to fill in the sync pattern on outgoing packets	No	false
Response Timeout	Number of seconds to wait for a response before timing out	No	5.0
Response Polling Period	Number of seconds to wait between polling for a response	No	0.02
Raise Exceptions	Whether to raise exceptions when errors occur like timeouts or unexpected responses	No	false

Preidentified Protocol

The Preidentified Protocol delineates packets using a custom COSMOS header. This Protocol is created to allow tools to connect and receive the entire packet stream. It can also be used to chain COSMOS instances together although that should rarely be needed with the new web native implementation.

Parameter	Description	Required	Default
Sync Pattern	Hex string representing a byte pattern that will be searched for in the raw data. This pattern represents a packet delimiter and all data found AFTER the sync pattern will be returned. The sync pattern itself is discarded.	No	nil (no sync pattern)
Max Length	The maximum allowed value in the length field	No	nil (no maximum length)
Mode	The Version of the preidentified protocol to support (2 or 4).3	No	4

Helper Protocols

COSMOS provides the following helper protocols: Crc & Ignore. These protocols provide helper functionality to Interfaces.

CRC Protocol

The CRC protocol can add CRCs to outgoing commands and verify CRCs on incoming telemetry packets.

Parameter	Description	Required	Default
Write Item Name	Item to fill with calculated CRC value for outgoing packets (nil = don't fill)	No	nil
Strip CRC	Whether to remove the CRC from incoming packets	No	false
Bad Strategy	How to handle CRC errors on incoming packets. ERROR = Just log the error, DISCONNECT = Disconnect interface	No	"ERROR"
Bit Offset	Bit offset of the CRC in the data. Can be negative to indicate distance from end of packet	No	-32
Bit Size	Bit size of the CRC - Must be 16, 32, or 64	No	32
Endianness	Endianness of the CRC (BIG_ENDIAN/LITTLE_ENDIAN)	No	"BIG_ENDIAN"
Poly	Polynomial to use when calculating the CRC expressed as an integer	No	nil (use default polynomial - 16-bit=0x1021, 32-bit=0x04C11DB7, 64-bit=0x42F0E1EBA9EA3693)
Seed	Seed value to start the calculation	No	nil (use default seed - 16-bit=0xFFFF, 32-bit=0xFFFFFFFF, 64-bit=0xFFFFFFFFFFFFFFFF)
Xor	Whether to XOR the CRC result with 0xFFFF	No	nil (use default value - 16-bit=false, 32-bit=true, 64-bit=true)
Reflect	Whether to bit reverse each byte of data before calculating the CRC	No	nil (use default value - 16-bit=false, 32-bit=true, 64-bit=true)

Ignore Packet Protocol

The Ignore Packet protocol drops specified command packets sent by COSMOS or drops incoming telemetry packets.

Parameter	Description	Required	Default
Target Name	Target name of the packet to ignore	Yes	nil
Packet Name	Packet name of the packet to ignore	Yes	nil

Custom Protocols

Creating a custom protocol is easy and should be the default solution for customizing COSMOS Interfaces (rather than creating a new Interface class). However, creating custom Interfaces is still useful for defaulting parameters to values that always are fixed for your target and for including the necessary Protocols. The base COSMOS Interfaces take a lot of parameters that can be confusing to your end users. Thus you may want to create a custom Interface just to hard coded these values and cut the available parameters down to something like the hostname and port to connect to.

All custom Protocols should derive from the Protocol class found in the COSMOS gem at lib/openc3/interfaces/protocols/protocol.rb. This class defines the 9 methods that are relevant to writing your own protocol. The base class implementation for each method is included below as well as a discussion as to how the methods should be overridden and used in your own Protocols.

Protocol APIs

Protocols should not require 'openc3/script' since they are part of a COSMOS interface. They should use the COSMOS library code directly like System, Packet, Bucket, BinaryAccessor, etc. When in doubt, consult the existing COSMOS protocol classes.

To really understand how Protocols work, you first must understand the logic within the base Interface class read and write methods.

Let's first discuss the read method.

On EVERY call to read, an empty Ruby string "" is first passed down to each of the read Protocol's read_data() method BEFORE new raw data is attempted to be read using the Interface's read_interface() method. This is a signal to Protocols that have cached up more than one packet worth of data to output those cached packets before any new data is read from the Interface. Typically no data will be cached up and one of the Protocols read_data() methods will return :STOP in response to the empty string, indicating that more data is required to generate a packet. Each Protocol's read_data() method can return one of three things: data that will be passed down to any additional Protocols or turned into a Packet, :STOP which means more data is required from the Interface for the Protocol to continue, or :DISCONNECT which means that something has happened that requires disconnecting the Interface (and by default trying to reconnect). Each Protocol's read_data method is passed the data that will eventually be turned into a packet and returns a possibly modified set of data. If the data passes through all Protocol's read_data() methods it is then converted into a COSMOS packet using the Interface's convert_data_to_packet() method. This packet is then run in a similar fashion through each Read Protocol's read_packet() method. This method has essentially the same return possibilities: a Packet (instead of data as in read_data()), :STOP, or :DISCONNECT. If the Packet makes it through all read_packet() methods then the Interface packet read counter is incremented and the Packet is returned to the Interface.

The Interface write() method works very similarly to read. (It should be mentioned that by default write protocols run in the reverse order of read protocols. This makes sense because when reading you're typically stripping layers of data and when writing you're typically adding on layers in reverse order.)

First, the packet write counter is incremented. Then each write Protocol is given a chance to modify the packet by its write_packet() method being called. This method can either return a potentially modified packet, :STOP, or :DISCONNECT. If a write Protocol returns :STOP no data will be written out the Interface and it is assumed that more packets are necessary before a final packet can be output. :DISCONNECT will disconnect the Interface. If the packet makes it through all the write Protocol's write_packet() methods, then it is converted to binary data using the Interface's convert_packet_to_data() method. Next the write_data() method is called for each write Protocol giving it a chance to modify the lower level data. The same return options are available except a Ruby string of data is returned instead of a COSMOS packet. If the data makes it through all write_data() methods, then it is written out on the Interface using the write_interface() method. Afterwards, each Protocol's post_write_interface() method is called with both the final modified Packet, and the actual data written out to the Interface. This method allows follow-up such as waiting for a response after writing out a message.

Method discussions

initialize

This is the constructor for your custom Protocol. It should always call super(allow_empty_data) to initialize the base Protocol class.

Base class implementation:

# @param allow_empty_data [true/false] Whether STOP should be returned on empty data
def initialize(allow_empty_data = false)
  @interface = nil
  @allow_empty_data = ConfigParser.handle_true_false(allow_empty_data)
  reset()
end

As you can see, every Protocol maintains state on at least two items. @interface holds the Interface class instance that the protocol is associated with. This is sometimes necessary to introspect details that only the Interface knows. @allow_empty_data is a flag used by the read_data(data) method that is discussed later in this document.

reset

The reset method is used to reset internal protocol state when the Interface is connected and/or disconnected. This method should be used for common resetting logic. Connect and Disconnect specific logic are handled in the next two methods.

Base class implementation:

def reset
end

As you can see, the base class reset implementation doesn't do anything.

connect_reset

The connect_reset method is used to reset internal Protocol state each time the Interface is connected.

Base class implementation:

def connect_reset
  reset()
end

The base class connect_reset implementation just calls the reset method to ensure common reset logic is run.

disconnect_reset

The disconnect_reset method is used to reset internal Protocol state each time the Interface is disconnected.

Base class implementation:

def disconnect_reset
  reset()
end

The base class disconnect_reset implementation just calls the reset method to ensure common reset logic is run.

read_data

The read_data method is used to analyze and potentially modify any raw data read by an Interface. It takes one parameter as the current state of the data to be analyzed. It can return either a Ruby string of data, :STOP, or :DISCONNECT. If it returns a Ruby string, then it believes that data may be ready to be a full packet, and is ready for processing by any following Protocols. If :STOP is returned then the Protocol believes it needs more data to complete a full packet. If :DISCONNECT is returned then the Protocol believes the Interface should be disconnected (and typically automatically reconnected).

Base Class Implemenation:

def read_data(data)
  if (data.length <= 0)
    if @allow_empty_data.nil?
      if @interface and @interface.read_protocols[-1] == self # Last read interface in chain with auto @allow_empty_data
        return :STOP
      end
    elsif !@allow_empty_data # Don't @allow_empty_data means STOP
      return :STOP
    end
  end
  data
end

The base class implementation does nothing except return the data it was given. The only exception to this is when handling an empty string. If the allow_empty_data flag is false or if it nil and the Protocol is the last in the chain, then the base implementation will return :STOP data to indicate that it is time to call the Interface read_interface() method to get more data. Blank strings are used to signal Protocols that they have an opportunity to return a cached packet.

read_packet

The read_packet method is used to analyze and potentially modify a COSMOS packet before it is returned by the Interface. It takes one parameter as the current state of the packet to be analyzed. It can return either a COSMOS packet, :STOP, or :DISCONNECT. If it returns a COSMOS packet, then it believes that the packet is valid, should be returned, and is ready for processing by any following Protocols. If :STOP is returned then the Protocol believes the packet should be silently dropped. If :DISCONNECT is returned then the Protocol believes the Interface should be disconnected (and typically automatically reconnected). This method is where a Protocol would set the stored flag on a packet if it determines that the packet is stored telemetry instead of real-time telemetry.

Base Class Implementation:

def read_packet(packet)
  return packet
end

The base class always just returns the packet given.

write_packet

The write_packet method is used to analyze and potentially modify a COSMOS packet before it is output by the Interface. It takes one parameter as the current state of the packet to be analyzed. It can return either a COSMOS packet, :STOP, or :DISCONNECT. If it returns a COSMOS packet, then it believes that the packet is valid, should be written out the Interface, and is ready for processing by any following Protocols. If :STOP is returned then the Protocol believes the packet should be silently dropped. If :DISCONNECT is returned then the Protocol believes the Interface should be disconnected (and typically automatically reconnected).

Base Class Implementation:

def write_packet(packet)
  return packet
end

The base class always just returns the packet given.

write_data

The write_data method is used to analyze and potentially modify data before it is written out by the Interface. It takes one parameter as the current state of the data to be analyzed and sent. It can return either a Ruby String of data, :STOP, or :DISCONNECT. If it returns a Ruby string of data, then it believes that the data is valid, should be written out the Interface, and is ready for processing by any following Protocols. If :STOP is returned then the Protocol believes the data should be silently dropped. If :DISCONNECT is returned then the Protocol believes the Interface should be disconnected (and typically automatically reconnected).

Base Class Implementation:

def write_data(data)
  return data
end

The base class always just returns the data given.

post_write_interface

The post_write_interface method is called after data has been written out the Interface. The typical use of this method is to provide a hook to implement command/response type interfaces where a response is always immediately expected in response to a command. It takes two parameters, the packet after all modifications by write_packet() and the data that was actually written out the Interface. It can return either the same pair of packet/data, :STOP, or :DISCONNECT. If it returns a packet/data pair then they are passed on to any other Protocols. If :STOP is returned then the Interface write() call completes and no further Protocols post_write_interface() methods are called. If :DISCONNECT is returned then the Protocol believes the Interface should be disconnected (and typically automatically reconnected). Note that only the first parameter "packet", is checked to be :STOP, or :DISCONNECT on the return.

Base Class Implementation:

def post_write_interface(packet, data)
  return packet, data
end

The base class always just returns the packet/data given.

Examples

Please see the included COSMOS protocol code for examples of the above methods in action.

lib/openc3/interfaces/protocols/protocol.rb lib/openc3/interfaces/protocols/burst_protocol.rb lib/openc3/interfaces/protocols/fixed_protocol.rb lib/openc3/interfaces/protocols/length_protocol.rb lib/openc3/interfaces/protocols/terminated_protocol.rb lib/openc3/interfaces/protocols/template_protocol.rb lib/openc3/interfaces/protocols/crc_protocol.rb lib/openc3/interfaces/protocols/preidentified_protocol.rb