km@emory.uucp (Ken Mandelberg) (11/28/88)
The built in Kermit protocol support on the Trailblazer seems not to be
as useful as I had hoped. Here are the results of some Kermit file
transfers I did with a 23K test file. Both systems were hooked to
Trailblazer+'s at 9600 baud.
Packet Size Kermit Spoofing Time for Transfer
90 ON 105 secs
90 OFF 385 secs
256 OFF 169 secs
512 OFF 106 secs
1000 OFF 73 secs
This is all with C-kermit 4E(070). When I tried packets above 1000
bytes the transfer timed out, and I didn't want to trouble shoot it. On
one side the computer (a 3B1) is directly connected to the Trailblazer.
On the other side the computer (a Sun) was connected through a terminal
server. The long packet timeouts could have been trouble in any of
kermit/modems/terminal server or the two computers. However, note that
maximum transfer rate (with 1000 byte packets) is 315 chars/sec, and
the same hardware gives about 900 chars/sec with UUCP (spoofing on).
One interesting thing is that although the Trailblazer does benefit
from the spoofing at small packet sizes, at anything above 512 byte
packets I am better off just doing without spoofing and eating the line
turnaround time.
Unless there is something unusual about my experiment it would appear
that you are better off without the Trailblazers Kermit support unless
you are forced to communicate with a system (or through a communication
link) that forces small packets.
Is it time to try and get the Telebit people to allow the Kermit
spoofing to use more than 90 character packets?
--
Ken Mandelberg | km@mathcs.emory.edu PREFERRED
Emory University | {decvax,gatech}!emory!km UUCP
Dept of Math and CS | km@emory NON-DOMAIN BITNET
Atlanta, GA 30322 | Phone: (404) 727-7963PLS@cup.portal.com (Paul L Schauble) (11/29/88)
The posted results for Kermit spoofing vs. non-spoofing surprise me. I was under the impression that the Trailblazer set up some of it's 512 carriers in each direction and changed the allocation dynamically as the relative traffic in the two directions changed. If this were true, then protocol spoofing shouldn't make much, if any, difference. Can someone enlighten me as to how it really works? Also, does anyone have any data on the US Robotics Courier 9600 HST does at running standard (not large block) Kermit? ++PLS