home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!usc!cs.utexas.edu!sun-barr!olivea!sgigate!odin!twilight!florida.wpd.sgi.com!skibo
- From: skibo@florida.wpd.sgi.com (Thomas Skibo)
- Newsgroups: comp.protocols.tcp-ip
- Subject: Re: Sequence Numbers in TCP/IP
- Message-ID: <v7evfu0@twilight.wpd.sgi.com>
- Date: 23 Jan 93 02:45:51 GMT
- References: <andreas-220193121931@i2pc5.informatik.rwth-aachen.de> <1993Jan22.142637.26110@aston.ac.uk> <16594@auspex-gw.auspex.com>
- Sender: news@twilight.wpd.sgi.com ( CNews Account at twilight.wpd.sgi.com )
- Organization: Silicon Graphics, Inc.
- Lines: 41
-
- In article <16594@auspex-gw.auspex.com>, guy@Auspex.COM (Guy Harris) writes:
- |> >The window is how much the other end is prepared to accept (65535 implies
- |> >infinite). (this is bigger than the maximum packet size on every network
- |> >I have ever seen)
- |>
- |> Yes, but, as far as I know, it's not required that the sending TCP stuff
- |> everything into one packet, and there's a "TCP Big Window" option
- |> mentioned in RFC1106, to increase the maximum window size beyond 64K.
- |>
- |> There's also an RFC, RFC1110, entitled "A Problem with the TCP Big
- |> Window Option"; both date back to mid-1989, and not being a full-time
- |> networking weenie, I don't know what the state of Big Windows in TCP is
- |> (except that I seem to remember Vernon Schryver talking about cranking
- |> the window size up when using FDDI, in order to light the fibre up, so I
- |> suspect there's at least a semi-stable standard for Big Windows).
-
- RFC 1323 is the latest RFC describing big windows. It is basically an
- update to RFC's 1072 (window scaling, time-stamps for rtt measurement,
- and selective acknowledgements) and 1185 (using the same time-stamps for
- rejecting old, duplicate segments). The selective acknowledgement was
- dropped from 1323 because it was felt that more research needed to be
- done.
-
- RFC 1106 was another method of getting >64K windows. I don't know much
- about it. The problem with big windows described in RFC 1110 (that a 32-bit
- sequence number isn't good enough with 30-bit windows) was addressed in RFC
- 1185 and refined in RFC 1323.
-
- RFC 1323 is pretty solid. SGI's next major OS release will include a
- TCP that can do RFC 1323 (both window scaling, and time-stamps).
-
- Vernon nearly saturates an FDDI ring with a couple of Indigos and a single
- TCP connection with a 60K window (no window scaling needed). But, he likes
- to crank up the window to 128K+ to squeeze the last 0.2 MBytes/s out.
- Since the "big windows" will be available soon, you'll be able to try this
- at home.
-
-
- --
- ---
- Thomas Skibo Advanced Networking, Silicon Graphics skibo@sgi.com
-