|
A Technical History of the ARPANET -
A Technical Tour
IMP-to-IMP
Error Detection
Note: The word "message" can mean either messages or portions
of messages (i.e. packets).
Source-to-Destination error control is comprised of three tasks:
- Detecting bit errors in the delivered messages.
- Detecting missing messages or pieces of messages.
- Detecting duplicate messages or pieces of messages.
Detecting bit errors in the delivered message is done through
the use of checksums. A checksum is appended to the message at the source
and is checked at the destination. When the checksum does not check at
the destination, the incorrect message is discarded, requiring it to be
retransmitted from the source.
To detect missing messages, the source IMP keeps track of all
messages for which an RFNM has not been received, and the destination IMP
keeps track of the replies it either has yet to receive or has already
sent. When the RFNM is not received for 30 seconds, the source IMP sends
a control message containing the message number to the destination
inquiring about the possibility of an incomplete transmission. The
destination will respond with either an indication that the message was
not received or that it is out of range, or with a correct duplicate
reply (RFNM). The source continues until it receives a response. This
insures that the source and destination IMPs keep their message number
sequences synchronized should a message become lost.
Duplicate messages can be detected with an identifying sequence
number. There is a number assigned to each message at the source, and the
destination has a window of acceptable message numbers. Messages with
out-of-range numbers are discarded, as well as duplicate messages.
|
