gritz@homxa.UUCP (R.SHARPLES) (07/25/85)
Is there anyone out there, possibily from Convergent Technologies, that can help with these problems with the 7300? 1) How do you get the break key to transmit a break out through cu (from a unix window). I use "cu -l/dev/tty000" to go out the serial port to a "hardwire" to a micom port selector. The micom expects 1200baud but if you send if 2 breaks it will switch to 9600. 2 breaks is the only thing that will do it. Is there any way (assuming the break key doesn't work) to send a break? 2) Why does the unixpc write to the screen so slowly when connected to a host at 9600. I have logged into a vax (thru cu) through an lan (ungermann/bass) at 9600. Still, characters appear on the screen just as slowly as at 1200. I have checked and it is not that the vax is loaded down. The AT&T help people are aware of both problems but can not offer any help. I am looking for some sort of workaround for both. Russ Sharples AT&T Bell Labs Holmdel, NJ 201-949-0575 homxa!gritz
mjs@eagle.UUCP (M.J.Shannon) (07/26/85)
> Is there anyone out there, possibily from Convergent Technologies, that can > help with these problems with the 7300? > > 1) How do you get the break key to transmit a break out through > cu (from a unix window). I use "cu -l/dev/tty000" to go out > the serial port to a "hardwire" to a micom port selector. The > micom expects 1200baud but if you send if 2 breaks it will switch > to 9600. 2 breaks is the only thing that will do it. Is there > any way (assuming the break key doesn't work) to send a break? The way to get cu to transmit a break is "~%b" (the "~" must appear immediately after a newline). > 2) Why does the unixpc write to the screen so slowly when connected > to a host at 9600. I have logged into a vax (thru cu) through an lan > (ungermann/bass) at 9600. Still, characters appear on the screen > just as slowly as at 1200. I have checked and it is not that the > vax is loaded down. I've not seen this problem on any of the 7300's I've used, so I can't offer any suggestions. Sorry. > The AT&T help people are aware of both problems but can not offer any help. > I am looking for some sort of workaround for both. > > Russ Sharples Good luck on finding an answer to problem 2. -- Marty Shannon UUCP: ihnp4!eagle!mjs Phone: +1 201 522 6063
dopey@ihlpl.UUCP (James C Blasius) (07/27/85)
> 2) Why does the unixpc write to the screen so slowly when connected > to a host at 9600. I have logged into a vax (thru cu) through an lan > (ungermann/bass) at 9600. Still, characters appear on the screen > just as slowly as at 1200. I have checked and it is not that the > vax is loaded down. > I would suggest that if your terminal emulator on a 9600 baud line is painting screens at a 1200 baud rate, something is broken in your machine! seriously, I've just been trying my 9600 baud line to my 3b5 to see if it paints at a 1200 baud rate and I assure you it's faster. Nearly as fast as a full screen vi session locally on the machine. There is a misconception I've heard several dozen times concerning the vt100 emulator. The misconception is that the emulator runs slowly, even at 1200 baud. I would suggest that those of you who believe this try catting large files with the emulator. I think you will find that the emulator is not slow. The emulator is sporadic, not nearly as smooth in its screen painting as cu for example, but it's not slow. However, you must remember that the screen on the UNIX PC is a bitmapped screen. Bitmapped screens without special screen support hardware end up making the processor move around large numbers of bytes whenever the screen scrolls. When you approach 9600 baud you may begin to find that your overall throughput will not match the theoretical, more or less depending on the kind of data you are sending (i.e., if you are sending short lines there's going to be more bytes moved per character on average, etc). This is not limited to the UNIX PC, but is an issue with all bitmapped output devices. james blasius ihnp4!ihlpl!dopey
rrt@duke.UUCP (Russell R. Tuck) (07/29/85)
In article <1054@homxa.UUCP> gritz@homxa.UUCP (R.SHARPLES) writes: > 2) Why does the unixpc write to the screen so slowly when connected > to a host at 9600. I have logged into a vax (thru cu) through an lan > (ungermann/bass) at 9600. Still, characters appear on the screen > just as slowly as at 1200. I have checked and it is not that the > vax is loaded down. > >The AT&T help people are aware of both problems but can not offer any help. I have also experienced Problem 2, with an interesting twist: I use a 7300 connected through its RS-232 port to a VAX 11/785 running 4.2BSD. When I use the phone manager's terminal emulator to login on the VAX and cat a large file, the screen scrolls at normal 9600 baud speed. However, when I use cu over the same line to login on the same VAX and cat the same file (with the same low load average on the VAX and no background activity on the 7300), the lines appear much more slowly, at what appears to be 1200 baud. I have done stty's on all the relavant ports (both machines), and they all say the line is at 9600 baud. The people on the AT&T support hotline could not reproduce this behavior on a machine there and could give me no help. Could I have an older release of cu or some system routine which cu uses (and the phone manager does not), which performs simple I/O inefficiently? Could there be some hardware problem in my machine which affects cu, but NOT the phone manager's terminal emulator? I am interested in any solution or idea that might lead to one.
gritz@homxa.UUCP (R.SHARPLES) (07/31/85)
As a follow up to my own posting: some of the mail I have received has indicated that the slow speed of cu is inherent in its design. Cu was designed primarily for connections over phone lines at 1200 baud. It buffers all text in the program. Hence, the execution speed of the program will effect the usable baud rate. Over a phone line, the bottle neck is the 1200 baud modem. But on a 9600 baud serial port the bottle neck becomes cu's execution speed. The terminal emulator doesn't use cu and is not affected by this. Whether anything can be done about cu I haven't found out yet. Perhaps there is some way to modify the buffer to speed up operation. Any ideas? Russell P. Sharples AT&T-BL homxa!gritz
larry@kitty.UUCP (Larry Lippman) (08/03/85)
> In article <1054@homxa.UUCP> gritz@homxa.UUCP (R.SHARPLES) writes: > > 2) Why does the unixpc write to the screen so slowly when connected > > to a host at 9600. I have logged into a vax (thru cu) through an lan > > (ungermann/bass) at 9600. Still, characters appear on the screen > > just as slowly as at 1200. I have checked and it is not that the > > vax is loaded down. > > > > I use a 7300 connected through its RS-232 port to a VAX 11/785 running 4.2BSD. > When I use the phone manager's terminal emulator to login on the VAX and cat > a large file, the screen scrolls at normal 9600 baud speed. However, when I > use cu over the same line to login on the same VAX and cat the same file > (with the same low load average on the VAX and no background activity on the > 7300), the lines appear much more slowly, at what appears to be 1200 baud. > I have done stty's on all the relavant ports (both machines), and they all say > the line is at 9600 baud. Welcome to the club. I bet if you look at an stty -a you will see that ixon and ixoff are enabled. Your 7300 is sending X-ON/X-OFF characters to limit throughput. If you do an stty and disable them, you will no doubt lose characters at 9,600 baud. I also have this problem on 3B2's and systems running Unisoft Sys V. The simple truth is that ain't no way cu can keep up a throughput of greater than 2,400 baud. Even 2,400 is marginal, but 1,200 is safe. Don't feel bad; I too learned this lesson the hard way. Larry Lippman Recognition Research Corp. Clarence, New York UUCP {decvax,dual,rocksanne,rocksvax,watmath}!sunybcs!kitty!larry {rice,shell}!baylor!kitty!larry syr!buf!kitty!larry VOICE 716/741-9185 TELEX {via WUI} 69-71461 answerback: ELGECOMCLR "Have you hugged your cat today?"
olson@fortune.UUCP (Dave Olson) (08/05/85)
Saying that the throughput of 'cu' is limited by it's design is ridiculous! I've been using cu on 9600 baud lines on 11/70's (V7), VAX 11/780 (4.1BSD) , and Fortune 32:16 (V7/4.1/++ hybrid) (of course :-) ), and ALL of them will get throughput at 9600 baud that is VERY close to a hardwired line. It is true that some characters will be dropped without flow control IF the RECEIVING system is very heavily loaded, but this does not sound likely in the case described. If cu can't keep up on the 7300, it is either a bug in the 7300 implementation, or there is a bottleneck in the tty i/o kernel routines, which are circumvented by the other programs mentioned. Dave Olson, Fortune Systems