Source: The ABCs of IP Addressing by Gilbert Held (2002)
Example:
The actual entry in the Sequence Number field is based on the number of
bytes in the TCP data field, i.e., because TCP was developed as a byte-oriented
protocol, each byte in each packet is assigned a sequence number. Because
it would be most inefficient for TCP to transmit 1 byte at a time, groups of
bytes, typically 512 or 536, are placed in a segment and 1 sequence number
is assigned to the segment and placed in the sequence field. That number is
based on the number of bytes in the current segment as well as previous
segments, because the sequence field value increments its count until all 16-bit positions are used and then continues via a rollover through zero. For
example, assume the first TCP segment contains 512 bytes and a second
segment will have the sequence number 1024.
I looked it up in a dictionary and I don't think there was a meaning that even remotely matched what I've got there in that paragraph.
Best Answer
Have you ever seen a car with a mechanical odometer?
Many cars and trucks built during the 1960s had odometers that showed mileages between 0.0 miles and 99,999.9 miles. The odometers were connected via gears to the vehicles' transmissions. The last digit slowly moved as the wheels moved. Each time a 9 needed to be replaced by a 0, the next digit would also be bumped up by 1. When the vehicle reached 100,000 miles, all of the digits would move. The odometer was said to "roll over". (In this example, it would "roll over" from 99,999.9 miles to 0.0 miles.)
The computer programming use of the word "rollover" is analogous. An unsigned 8-bit character can roll over from 255 to 0. A 16-bit unsigned integer can roll over from 65,535 to 0.
In general, most programming languages either throw errors when overflow occurs, or silently roll over from a data type's maximum value to its minimum value. For example, a signed 8-bit character can roll over from 127 to -128, and a 16-bit signed integer can roll over from 32,767 to -32,768.