Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!sri-unix!ctnews!pyramid!decwrl!decvax!ucbvax!HUCSC.HARVARD.EDU!roger%cfa2
From: roger%cfa2@HUCSC.HARVARD.EDU
Newsgroups: comp.os.vms
Subject: Experience with TCP/IP Terminal Servers at CfA
Message-ID: <8704252252.AA23325@ucbvax.Berkeley.EDU>
Date: Fri, 24-Apr-87 11:16:54 EDT
Article-I.D.: ucbvax.8704252252.AA23325
Posted: Fri Apr 24 11:16:54 1987
Date-Received: Sun, 26-Apr-87 22:36:50 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 53

In response to Doug Bigelow's query:

 We have purchased five Bridge CS/1 terminal servers.  Each is configured
 to service 64 termiinals using TCP/IP over an ethernet.  For VMS hosts we have
 one 780 and one microVAX II operating Wollongong, one 780 and about four
 microVAX II's with Excelan boards and software.  We also have at least two
 VAXes running Ultrix, about 20 Sun workstations and a few other
 TCP/IP-speaking hosts.

 I have been responsible for the installation of the terminal servers and
 the TCP/IP on the VMS systems.  Our hope was to give service about as
 good coming into the VAXes on terminal ports (a mixture of DMA and non-DMA)
 and have the flexibility of being able to connect to both VMS and non-VMS
 TCP/IP systems.

 The results with VMS has been a step backwards in terms of human factors
 for the terminal user:
 1)  X-OFF - X-ON flow control does not work well.  When you hit X-OFF (NO
     SCROLL) on your terminal, you get lots of characters (hundreds) before
     the host stops sending.  CTRL-Y, CTRL-C, and CTRL-O have similar problems.
 2)  Programs which do single-character output to the terminal run *very*
     slowly.  The worst example we have is the 20/20 spreadsheet program;
     another is TeX.  It seems that programs written in C often do
     their terminal I/O one character at a time.
 3)  The Excelan Telnet host software does not emulate the VMS read-with-prompt
     QIO.  The read with prompt allows you to do a write followed  by a read
     with one I/O call.  Excelan implements the write by sending one character
     per packet. This is extremely slow, as in topic (2) above.  The place
     where this is most most frequently agravating is when doing command line
     editing; VMS CLI uses a read with prompt when it refreshes the command
     line.

 All in all, Wollongong seems to behave itself better as a Telnet host than
 Excelan.

 I don't implicate the Bridge terminal servers in any of this.  I have equal
 problems when I am using my PC/AT with a 3COM board and ftp Systems telnet
 software.  The Bridge terminal servers seem to be a well thought-out and
 engineered product and I have had pleasing response form both the local and
 national technical support people.  My principal, but minor, complaint with
 the Bridge terminal servers is that they only allow one session per terminal.

 I have not tried the TEK TCP as a host yet.  I expect to make a temporary
 arangement to try it out some time in the next week or two.  I'll let you
 know how it works.

 I welcome response from anyone who has some help with these problems or
 who just wants to commiserate.

 				Roger Hauck
 				Harvard-Smithsonian
 				Center for Astrophysics