…
I missed a fair amount of content
…

Pipelining

• Go back in
  • Local waits for an eventual response from server containing an adjusted window size, and adjusts to it
  • If local receives a timeout, local attempts resending the packets
    • Timer is based upon oldest packet sent
  • If server receives an out of sequence packet, it won't be buffered
    • This is because the server knows local will send all of the packets again
  • Ideal for channels with large, chunked packet losses
• Selective repeat
  • Data is still sent off, and a response is waited for for a packet
  • When there's a timeout on a packet response, only that one packet is resent
  • Ideal for channels with small, individual losses
  • Window size must be less than or equal to half the size of seq# space