NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/comp/protocol/tcpip/6124

Internet Message Format | 1993-01-28 | 1.8 KB

Path: sparky!uunet!ferkel.ucsb.edu!ucsbcsl!spectrum.CMC.COM!lars
From: [e-mail] (Lars Poulsen)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Sequence Numbers in TCP/IP
Message-ID: <[e-mail]>
Date: 26 Jan 93 07:19:09 GMT
References: <[e-mail]>
Organization: CMC Network Systems (Rockwell DCD), Santa Barbara, CA, USA
Lines: 28

In article <[e-mail]>
   [e-mail] (Andreas Fasbender) writes:
>What is the reason for the difference in the lenght of the window-field
>(16) and the sequence-number-field (32 bit).
>Using 16-bit-numbers for sequencing
>and acknowledging could spare 4 byte in each packet.

First, let me acknowledge that I understand the question. Since these
quantities are of the same dimension, they could be expected to be in
the same units.

(1) When the protocol was specified, the typical longhaul link speed
    was 56 kbps, and the maximum bytes-in-flight tended to work out to
    around 2000, so 64KB of window seemed adequate.
    Today, we are trying to run this across the country (300 ms round
    trip delay) on Gigabit pipes, so 64 KB of window is too little.
    An addendum to TCP allows larger windows to be specified.
(2) On a gigabit link, even the sequence numbers might wrap around
    during the packet lifetime. The "large windows" TCP extension
    includes PAWS - protection against wrapped sequence.

It would not be unreasonable to specify 64 bits for each of these fields
in a new protocol suite (or count both in frames instead of bytes - the
ISO way).
--
/ Lars Poulsen, SMTS Software Engineer    Internet E-mail: [e-mail]
  CMC Network Products / Rockwell Int'l    Telephone: [phone]
  Santa Barbara, CA 93117-3083        TeleFAX:   [phone]