[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 7.2 terminal woes



Timing tests to a SUN of "cat 10k.text" where 10K.TEXT is a file that is
200 lines of 48 characters (which if a CR and LF are transmitted comes to
10000 characters -- I originally wrote this file out for a machine that
did use CRLF).  Timings were by hand from the time the return was hit at the
end of a return.  We do have the ECOs loaded.  (The mouse was off the
main screen to the left during these timings.)

VT100:	11.5 seconds   (about 8700 baud)
SUPDUP:	6 seconds	(about 16.7 Kbaud)

To our Vax running VMS/Eunice:

DM2500 (homegrown interface):  27 seconds  (about 3700 baud)
VT100:	22 seconds (about 4500 baud)
9600 baud hardwired line:  12 seconds (about 8300 baud)


Both hosts are on our local net.  Both connections were TCP TELNET (except
the TCP SUPDUP connection).

This shows several things.  Your transmission speed may be due to your
foreign host.  Compare the two speeds of the VT100 on the two different
hosts.  Since the VT100 simulator could run in 11.5 seconds on the Sun
but took 22 seconds on the Vax, that indicates that the slowdown on the
Vax was due to the Vax not the Lisp machine.  Since the Vax could pump
out the text in 12 seconds to a hardwired line, I presume the extra 10
seconds had to do with network issues.

Secondly it shows that the slowdown to recording the 200 lines was probably
on the order of (11.5 - 6 =) 5.5 seconds.  (I haven't tried to analyze what the 
the SUN does in trying to send the output on SUPDUP or TELNET/VT100
nor in what the output is.  One may be more efficient at the SUN end or in the
networking.)


We generally use the DM2500 simulator to talk to our Vax and SUPDUP to
talk to the SUN.  I don't use the SUN much but when I use the Vax (quite a bit),
I often use the ability to scroll back.  I wouldn't go back to a non-DW
simulator for talking to the Vax.


(A wish for the terminal simulators:  Locally bind
SI:*KBD-REPETITION-INTERVAL* and make it a parameter of the terminal
window.  I find a value of 5 (ie, 12 chars/sec) works fine with my Vax
and Sun.  Part of the complaint about the terminal emulator not keeping
up with repeated characters may be that the a batch of typed characters
are going out in a single packet and the editor may skip showing
individual updates per character as it has more characters already
waiting.)


P.S.  A slight grump to Symbolics:  The ECOs didn't come with source so I
can't see what was done to the VT100 simulator to make it faster.  I don't
know whether my DM2500 simulator could make use of a similar patch.