wtm@neoucom.UUCP (Bill Mayhew) (12/07/89)
Hello group. I was just wondering if anyone out there has tried out the uucp e (e for error-free link) protocol .. especially on the 3b1. The e protocol requires that an error free machine-to-machine link exist. In the 4th quarter Unix release notes, AT&T actually mentions using the e protocol in conjuction with the CEO 2224 modem's MNP error correction. e doesn't bother ACK'ing each 64 byte packet the way the g protocol does; e transmits the entire file. Using the BNU uucp on the 3b1 with g protocol spoofing supported in the Trailblazer modem results in xferstats of around 1200-1250 characters/sec for me when the EIA port is pushed to 19.2K buad. I tried the e protocol, and got xferstats of about 1350 char/sec. In the latter case, the Trailblazer was set up with S111=0 and S110=0 for no protocol support and no compression, respectively. The problem is that uucp would usually get through about four files then lock up with something to the effect of (this is from memory): " --> 42879 1352 char/sec i-msg> 'c' got 0000 fail" with -x7 set in uucico. The temp files built in the spool directory seemed complete. The message above was as viewed from the transmitting end. I found that I didn't get anywhere at all wtih e protocol until both 3b1s had hfc enabled with /etc/hfc_ctl +/dev/tty000 and handshaking enabled on the Trailblazers (not really surprising, of course). A while back, Peter Honeyman said in comp.dcom.modems words to the effect "Use Trailblazers with g protocol, which is what they are designed for," and left the subject at that. True words, as it turns out. It'd be nice to squeeze a few more bytes though to save a couple of bucks a more each month on ye olde phone bill. Of course, 1250 char/sec with the g protocol isn't exactly chopped liver. By the way if you feel daring, the e protocol is enabled by putting a ,e after the Systems file reference to the Dialers file device reference: Sysname Any ACUFAST,e 13015557497 "" /r in:--in: uucp. You need to have BNU for e to work. Bill
jcm@mtunb.ATT.COM (John McMillan) (12/07/89)
In article <1847@neoucom.UUCP> wtm@neoucom.UUCP (Bill Mayhew) writes: >Hello group. I was just wondering if anyone out there has tried >out the uucp e (e for error-free link) protocol .. especially on >the 3b1. 'E' is used extensively in our 3B1 HDB StarLAN transfers. Very fast -- about 3 times as fast as 'g', if I recall properly. >The e protocol requires that an error free machine-to-machine link >exist. In the 4th quarter Unix release notes, AT&T actually >mentions using the e protocol in conjuction with the CEO 2224 >modem's MNP error correction. e doesn't bother ACK'ing each 64 >byte packet the way the g protocol does; e transmits the entire >file. > >Using the BNU uucp on the 3b1 with g protocol spoofing supported in >the Trailblazer modem results in xferstats of around 1200-1250 >characters/sec for me when the EIA port is pushed to 19.2K buad. I >tried the e protocol, and got xferstats of about 1350 char/sec. In Didn't you mention "an error free machine-to-machine link..."? The RS-232 ports on the 3B1 are *NOT* an error free link. It is trivial to overrun the ON-CHIP buffers during other RS-232 output operations [incl. the phone use]. In an environment where one can guarantee NO character losses between the kernel and the modem, the 'e' protocol might work -- but it's not going to happen on the 10MHz, overburdened 68010 of the 3B1. john mcmillan -- att!mtunb!jcm