home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.dcom.modems
- Path: sparky!uunet!sun-barr!ames!sgi!rhyolite!vjs
- From: vjs@rhyolite.wpd.sgi.com (Vernon Schryver)
- Subject: Re: What's a good 14400 modem?? (Intel?)
- Message-ID: <pavp9v8@rhyolite.wpd.sgi.com>
- Organization: Silicon Graphics, Inc. Mountain View, CA
- References: <1992Aug24.143532.8721@lowton.rn.com> <PZR7PB2w165w@infopls.chi.il.us> <1992Sep2.014327.13420@terminator.cc.umich.edu>
- Date: Wed, 2 Sep 1992 03:41:06 GMT
- Lines: 68
-
- In article <1992Sep2.014327.13420@terminator.cc.umich.edu>, honey@citi.umich.edu (Peter Honeyman) writes:
- > ...
- > i think the common bsd slip actually uses a two byte frame in this
- > case -- from slstart() in if_sl.c:
- >
- > /*
- > * The extra FRAME_END will start up a new packet, and thus
- > * will flush any accumulated garbage. We do this whenever
- > * the line may have been idle for some time.
- > */
- > if (tp->t_outq.c_cc == 0) {
- > ++sc->sc_bytessent;
- > (void) putc(FRAME_END, &tp->t_outq);
- > }
- >
- > so my earlier caluclation was off by one.
-
-
- No, as the comment says, the extra delimiter should only be emitted
- when the line was previously more or less idle.
- Consecutive SLIP frames should have 1-byte each of SLIP "header".
-
- This is what a netsnoop on `ping -s 10` on a SLIP line terminating
- about 16 inches from my right knee says:
-
- 0002: len 38 time 21:28:35.799
- ip: v 4 hl 5 tos 0 len 38 id 19099 off 0
- ttl 255 p icmp sum 0x65d0
- src rhyolite.wpd.sgi.com dst dataserv.denver.sgi.com
- icmp: type ECHO code 0 cksum 0x1c66 id 16127
- seq 11
- 00028: 2a a4 34 e3 *.4.
- 00032: 00 0c 34 f3 08 09 ..4...
-
- 0003: len 38 time 21:28:35.929
- ip: v 4 hl 5 tos 0 len 38 id 24932 off 0
- ttl 255 p icmp sum 0x4f07
- src dataserv.denver.sgi.com dst rhyolite.wpd.sgi.com
- icmp: type ECHOREPLY code 0 cksum 0x2466 id 16127
- seq 11
- 00028: 2a a4 34 e3 *.4.
- 00032: 00 0c 34 f3 08 09 ..4...
-
- The total packet length is 36 bytes excluding SLIP header.
- However you want to divide it, that's
- (28 bytes of IP stuff) + (1 byte for SLIP) + (ping command line value)
-
-
- > |> So you're right that my calculations are a little off. I don't think
- > |> it's worth worrying about. My inferences of when the modems stop being
- > |> able to keep up are not accurate to more than 10%, the modems
- > |> themselves vary by more than 10% from moment to moment, and I don't
- > |> care about any modem where my error of 37 bytes is 10% or more of the
- > |> modem's through-put.
- >
- > if you say so. but i like experimental results to be reproducible by
- > others, and hold precision to be a virtue.
-
-
- Reproducibility is the main reason I report so called "throughput" as I
- do. The numbers are straight from the ping command line. They do not
- require any problematical adjustments. As I noted recently, I even
- ignore the fact that the real throughput value is in some sense twice
- the command line value because data is flowing at that rate in both
- directions.
-
-
- Vernon Schryver, vjs@sgi.com
-