Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!rutgers!ucsd!ucsdhub!esosun!seismo!uunet!wiley!david
From: david@wiley.UUCP (David Hull)
Newsgroups: comp.dcom.modems
Subject: Re: Which is best?
Summary: TrailBlazer turnaround time
Message-ID: <2875@wiley.UUCP>
Date: 28 Oct 88 07:07:59 GMT
References: <9515@conexch.UUCP> <1125@vsi1.UUCP> <299@telebit.UUCP>
Reply-To: david@wiley.UUCP (David Hull)
Organization: TRW Inc., Redondo Beach, CA
Lines: 30

In article <299@telebit.UUCP> rls@telebit.UUCP (Richard Siegel) writes:
>And yes, he is right, our spec for turnaround time is 120 msec + or - 30 ms
>plus the actual line delay (about 60 msec for a cross country call).
>
>I think that the original poster probably had a very old TrailBlazer that only
>had one packet size (the long packets of 136 ms).

Another possibility is that he was using some curses-based program with
the "wrong" termcap entry.  When I first started using my TrailBlazer,
I found that the delay between typing a character in 'vi' and having it
echoed was terrible: around half a second.  I found that if I changed
the termcap entry from "ic=7\E1@" to "im=\E[4h:ei=\E4l" it improved the
interactive response tremendously.  Curses sends the "ic" (insert
character) string for each character inserted, while it only sends "im"
(insert mode) at the beginning and "ei" (end insert mode) at the end of
a series of characters.  The extra traffic with "ic" must have interacted
badly with the TrailBlazer's turnaround algorithm to make for very slow
interactive performance.

So if your terminal supports an insert mode, use it!

-David

P.S.  Before anyone suggests it, removing the 7 ms of padding from
"ic" didn't make any noticable difference.
-- 
					---------------------------------------
					David Hull  TRW Inc.  Redondo Beach, CA
					  ...!{uunet,cit-vax,trwrb}!wiley!david
					     david%wiley.uucp@csvax.caltech.edu