Path: utzoo!utgpu!utstat!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!amdcad!rpw3 From: rpw3@amdcad.AMD.COM (Rob Warnock) Newsgroups: comp.mail.uucp Subject: Re: Hardwired "g" protocol values Message-ID: <24789@amdcad.AMD.COM> Date: 10 Mar 89 07:45:15 GMT References: <60827@pyramid.pyramid.com> <455@lakart.UUCP> <406@texsun.Sun.COM> Reply-To: rpw3@amdcad.UUCP (Rob Warnock) Organization: [Consultant] San Mateo, CA Lines: 58 In article <406@texsun.Sun.COM> jthomp@hemaneh.UUCP writes: +--------------- | What we really need, is a good 'SLIP' modem, and dial-up slip service. | Then we could use SMTP, and NNTP, and do away with uucp, (for the most part.) +--------------- I think we already have good modems. Consider the following configuration: 1. One end: PC/AT clone running Phil Karn's "KA9Q" TCP/IP code, speaking SLIP over COM1 at 9600 baud. 2. Other end: VAX-11/780 running 4.3bsd, with SLIP driver, on 9600-baud DH-11 port. (The SLIP driver was *slightly* hacked by me to add proper carrier-detect & SIGHUP support.) 3. Telebit Trailblazer+ on both ends. 4. KA9Q's "TCP Max Seg Size" set to 512, "TCP Window Size" set to 4096. I have measured large FTP file transfers (100K+) at 880 bytes/sec of user data (that is, FileSize/ElapsedTime = 880). The best I have ever seen between Unix systems, even using the TB+'s UUCP mode, was a little over 800 bytes/sec at 9600 baud. The sustained transfer rate was somewhat sensitive to the "Max Seg Size" and "Window Size" settings of the "KA9Q" code. I tried a number of variations before settling on the above settings, which seem fairly robust in the FTP environment. I don't know what throughput I would have gotten between two 4.3bsd systems running SLIP on TB+'s. (Anybody want to try it and report?) p.s. Note that you don't really want a "SLIP modem"; what you really want is SLIP header compression in the hosts, to keep the chatter off the async line in the first place. That way, you might be able to get reasonable Telnet performance. Sad to say, the KA9Q/SLIP/TB+/TB+/SLIP/4.3 combo gives *abyssmal* Telnet interactive performance even at 9600 (though full-screen refresh is just fine). That's probably because a single typed character causes 43 bytes of data on the line (SLIP Start_Frame, IP, TCP, the byte, SLIP End_Frame), which causes the TB+'s to go into "unidirectional mode" with a turn-around time of ~1 sec., instead of the 7 (9?) times/sec ping-pong they do when you're using them as normal terminal modems. With SLIP header compression in the hosts, a single typed character would yield maybe 3-4 characters on the line, which wouldn't pop the TB+ out of ping-pong mode, and Telnet would win. p.p.s. SLIP is one place the "true" full-duplex of v.32 probably wins a lot over PEP. Anybody got a pair of v.32 modems they can try a Telnet/SLIP connection over? Rob Warnock Systems Architecture Consultant UUCP: {amdcad,fortune,sun}!redwood!rpw3 ATTmail: !rpw3 DDD: (415)572-2607 USPS: 627 26th Ave, San Mateo, CA 94403