home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
kermit.columbia.edu
/
kermit.columbia.edu.tar
/
kermit.columbia.edu
/
archives
/
ckc190.zip
/
ckvker.bwr
< prev
next >
Wrap
Text File
|
1996-03-21
|
35KB
|
755 lines
CKVKER.BWR "Beware File" for C-Kermit Version 5A -*- text -*-
(Open)VMS VERSION FOR DEC VAX AND AXP
Applies to 5A(190)
Last update: Wed Mar 29 18:58:22 1995
Authors: Frank da Cruz and Christine M. Gianone (Columbia University, NYC);
Terry Kennedy (Saint Peters College, Jersey City, NJ).
Copyright (C) 1985, 1994, Trustees of Columbia University in the City of New
York. The C-Kermit software may not be, in whole or in part, licensed or
sold for profit as a software product itself, nor may it be included in or
distributed with commercial products or otherwise distributed by commercial
concerns to their clients or customers without written permission of the
Office of Kermit Development and Distribution, Columbia University. This
copyright notice must not be removed, altered, or obscured.
Report problems, suggestions, fixes, etc, to:
Frank da Cruz
Columbia University
Academic Information Systems (AcIS)
612 West 115th Street
New York, NY 10025 USA
Internet: fdc@columbia.edu
Fax: +1 212 662-6442
VMS C-Kermit installation instructions are in the file CKVKER.INS. Please
be sure you have read that file before concluding that C-Kermit isn't working
right on VMS.
Note: "VMS" as used in this document refers to both VMS and OpenVMS on both
VAX and AXP (Alpha) processors. Most of the words in the previous sentence
are trademarks (TM) of Digital Equipment Corporation.
DOCUMENTATION
C-Kermit 5A is documented in the book "Using C-Kermit" by Frank da Cruz and
Christine M. Gianone, Digital Press, Burlington, MA, USA, ISBN 1-55558-108-0.
Price: US $39.95. To order, call Columbia University, New York City,
at +1 212 854-3703, or Digital Press / Butterworth-Heinemann at:
+1 800 366-2665 (Woburn, Massachusetts office for USA & Canada)
+44 1865 314627 (Oxford, England distribution centre for UK & Europe)
+61 03 9245 7111 (Melbourne, Vic, office for Australia & NZ)
+65 356-1968 (Singapore office for Asia)
+27 (31) 2683111 (Durban office for South Africa)
COMMAND PARSER
VMS-style command-line editing (arrow keys, etc) is not supported. Kermit
does not use the VMS F$PARSE facility -- it has its own command parser that
lacks certain features of F$PARSE (arrow-key editing, etc) but has many other
features that F$PARSE lacks: "?"-help, keyword and filename completion,
filename menus, variables, macros, etc. As of edit 190, C-Kermit does support
command recall (via Ctrl-B and Ctrl-N, not via arrow keys).
If you write a DCL command file that starts Kermit with a command-file name
as its first command-line argument, e.g.:
$ kermit oofa.scr
and then SUBMIT this DCL command file as a batch job, be aware that the
batch job is executed out of your login directory, so if the command file
(OOFA.SCR in this case) is not in your login directory, you must either SET
DEFAULT to the directory it is in, or else give a fully qualified filename:
$ set default [mydir.mysubdir]
$ kermit oofa.scr
or:
$ kermit [mydir.mysubdir]oofa.scr
Contrary to expectations of VMS users, the MSEND command does NOT use
commas to separate file specifications. E.g. say this:
C-Kermit>msend ckc*.% cku*.% ckv*.%
not this:
C-Kermit>msend ckc*.%, cku*.%, ckv*.%
CD (Change Directory) to a DECnet node does not work in VMS C-Kermit.
OPEN READ does not work if FILE TYPE is LABELED. Use SET FILE TYPE TEXT
before OPEN READ, then change the file type, if necessary, before file
transfer operations. (This bug will be fixed in the next release.)
The OPEN !READ process needs to be closed explicitly. If it is not, then
subsequent SEND commands will erroneously try to read from the !READ process.
OPEN !WRITE does not work in VMS C-Kermit.
VMS C-Kermit does NOT provide program status codes in the normal VMS manner.
Rather, it returns the codes described on pp. 323-324 of "Using C-Kermit", by
assigning them to the symbol CKERMIT_STATUS. For example, if a RECEIVE
operation failed:
$ show symbol ckermit_status
CKERMIT_STATUS == "4"
$
Arguments supplied to the EXIT (or QUIT) commands take precedence:
C-Kermit>exit 1234
$ show symbol ckermit_status
CKERMIT_STATUS == "1234"
$
If C-Kermit encounters no execution errors, and EXIT (QUIT) is given without
an operand, then:
C-Kermit>exit
$ show symbol ckermit_status
CKERMIT_STATUS == "0"
$
You can use the CKERMIT_STATUS symbol as in this DCL example:
$ kermit -s oofa.txt
$ if ckermit_status .eq. 0 then goto ok
RUNNING C-KERMIT IN DCL COMMAND PROCEDURES
It is often desirable to wrap C-Kermit in a DCL command procedure. Such
a procedure, for example OOFA.COM, can be run either directly on your job's
controlling terminal by:
$ @OOFA
or as a batch job via:
$ SUBMIT OOFA
When you are writing a DCL command procedure that runs C-Kermit, you
must make a choice:
1. If you want to be able to include Kermit commands in the DCL procedure as
"image data" (i.e. lines that don't start with $), then you can NOT
include any Kermit commands that would require access to the real console
terminal's keyboard and screen, such as CONNECT. That is, the person who
runs the DCL procedure can NOT interact directly with a remote computer.
This type of DCL command procedure can be run either on a terminal via @,
or as a batch job via SUBMIT. If you include a CONNECT command in this
type of batch job, the CONNECT command will fail with the following
message:
Sorry, Kermit's CONNECT command can be used only on a real terminal.
If this is not a batch a job, then you must:
$ DEFINE SYS$INPUT SYS$COMMAND
in your DCL command procedure before starting Kermit.
2. If you want the user to be able to interact directly with the remote
computer through Kermit's CONNECT command, then:
(a) The DCL procedure can be run only with @, not with SUBMIT. That is,
it cannot be a batch job; it must have access to the console terminal.
(b) You must include the following DCL command in the DCL procedure
immediately before starting Kermit:
$ DEFINE SYS$INPUT SYS$COMMAND
(/USER, /NOLOG, etc, switches may be used).
(c) You can not include Kermit commands as "image data" in the DCL command
procedure. Instead, you must create a separate Kermit command file,
and use command-line arguments to instruct Kermit to execute it; for
example:
$ define /user/nolog sys$input sys$command
$! Execute oofa.scr instead of normal initialization file.
$ kermit -y oofa.scr
or:
$ define /user sys$input sys$command
$! Execute oofa.scr after executing normal initialization file.
$ kermit "-C" "take oofa.scr"
Here is a sample DCL command procedure of the first type, which can be run
either on the controlling terminal or as a batch procedure, and requires no
interaction from the user. Lines beginning with dollar sign ($) are DCL
commands, other lines are fed to the application program (Kermit).
1. $ write sys$output "Hello from DCL"
2. $ set default [myuserid.mysubdirectory]
3. $ kermit
4. set prompt {}
5. echo Hello from C-Kermit
6. @ write sys$output "Hello from DCL from inside C-Kermit"
7. take oofa.scr
8. exit
9. $ write sys$output "All done."
(The numbers are not part of the file.) Lines 1-3 are DCL commands. Line 3
starts C-Kermit. Lines 4-8 are C-Kermit commands. Line 4 shows how to set
C-Kermit's prompt to nothing to reduce clutter in the batch log, should you
desire. Line 5 shows how to enter messages in the batch log. Line 6 shows
how to run DCL commands from within Kermit (you can use @ (at-sign), !
(exclamation mark), or the word RUN -- all of them are synonyms, followed by
a DCL command). Line 8 exits from C-Kermit back to DCL.
In line 7, C-Kermit is told to execute a script program from another file,
OOFA.SCR. Script programs to be run during the batch session are best kept in
separate C-Kermit command files because certain commands, notably GOTO, FOR,
WHILE, and XIF, do not work when entered in the interactive command stream.
Here is a sample command file:
set take echo on ; Make Kermit commands appear in the batch log
set take error on ; This stops execution automatically upon error
set input echo on ; This makes INPUT material appear in the batch log
set host blah ; Make a network connection to host "blah"
set file display serial ; Use SERIAL or NONE for the batch log, not FULL or CRT
input 5 login: ; Wait for a login prompt
output myuserid\13 ; Send my user ID and a carriage return
input 5 Password: ; Wait for password prompt
output \$(P1)\13 ; Send my password (see below) and a carriage return
input 20 \13\10$\32 ; Wait for system prompt
output kermit\13 ; Start Kermit on host "blah"
input 5 Kermit> ; Wait for Kermit> prompt
output server\13 ; Put remote Kermit in server mode
in 5 READY TO SERVE... ; Wait for READY message
get oofa.txt ; Get a file from the remote server
bye ; Terminate the remote session
end ; Return to local C-Kermit prompt
VERY IMPORTANT: Batched login scripts are inherently insecure because the
passwords are visible in plaintext, either in a file or else in the batch
queue entry. VMS presently offers no secure way (known to the writers of this
document) to enter a password into a batch job.
Two very insecure methods can be used:
1. Put the password in the Kermit script file. The risk here is that anybody
who gains access to the file, or to the system backup tapes, can learn your
password on the remote system.
2. Give the password as a parameter to the SUBMIT command when starting the
batch job, for example:
$ SUBMIT OOFA /NOTIFY /PARAM=("mypassword")
(This sets the DCL parameter P1 to your password on the remote host (for
further information, give the DCL command "help submit /param"). Quotation
marks are necessary to preserve lowercase letters (important when logging
in to UNIX hosts). DCL parameters may be referenced in Kermit commands as
\$(P1), \$(P2), etc.) The disadvantage here is that the VMS SHOW
ENTRY/FULL command displays the parameters from your SUBMIT command, making
the password visible to (at least) the system operator, and (most likely)
also to other users, such as members of your group (batch queues are, by
default, read-accessible by all members of their group).
Both methods can be made somewhat safer by adjusting the protections on the
files and/or batch queues that will contain sensitive information, but there
can be no guarantees. Therefore: EXERCISE EXTREME CAUTION with passwords in
login scripts and batch jobs.
And please note further that passwords passed in plain text -- as they still
must be in most cases, particularly those involving dialup access -- are
subject to discovery by various other means, including, but not limited to,
wire tapping.
RUNNING C-KERMIT FROM ALL-IN-1
Dr. David Kelly, Australian Environmental Protection Authority
kellyd@airmoon.epa.nsw.gov.au
ALL-IN-1 uses mailboxes (MBX) devices, rather than terminals. TT: is
reassigned from the user's controlling terminal to a mailbox device. C-Kermit
uses TT: as its default line device and so doesn't work straight off under
ALL-IN-1. SYS$INPUT is reassigned to something else again. SYS$OUTPUT
remains assigned to the user's original terminal line so it can be used to
specify the line device for C-Kermit when called from within ALL-IN-1. Below
is a script which can be run from ALL-IN-1 which calls C-Kermit to receive a
file. SYS$OUTPUT is temporarily redefined to stop some guff showing on the
screen.
$! RECEIVE_FROM_PC.COM
$!
$! Transfer file from PC into ALL-IN-1 using KERMIT
$! Invoked by TRANSFER_PC_TO_A1.SCP, which is in turn called by the RF
$! option on DT menu.
$!
$ set noon
$ on control_y then goto exit
$
$ tt1=f$trnlnm("sys$output")
$ kermit :== $epa__system:Ckermit
$ define/user sys$input sys$command
$ define sys$output sys$login:del.txt
$ kermit -l 'tt1' -b 9600 -r -a a1file.a1f -q -i
$ deassign sys$output
$ del sys$login:del.txt;
$exit:
$ exit
Similarly a file can be sent :
$! SEND_TO_PC.COM
$! Transfer document from ALL-IN-1 to the PC
$! invoked by TRANSFER_A1_TO_PC.SCP which is, in turn, called by the
$! SF option on the DT menu
$!
$ set noon
$ on control_y then goto exit
$!
$ write oamailbox "OA GET #CURDOC_FILENAM"
$ @dclmailbox:
$ a1file = "''result'"
$ vmsfile = "A1FILE.A1F"
$ copy/nolog/noconfirm 'a1file' 'vmsfile'
$ kermit :== $epa__system:Ckermit
$ define/user sys$input sys$command
$ tt1=f$trnlnm("sys$output")
$ define sys$output sys$login:del.txt
$ kermit -l 'tt1' -b 9600 -s A1FILE.A1F -q -i
$ deassign sys$output
$ del sys$login:del.txt;
$ if $severity .le. 1 then goto exit
$! if an error occurs, tell ALL-IN-1
$ write oamailbox "OA GET $PC_KERMIT_STATUS=0"
$ @dclmailbox:
$exit:
$ deletex/nolog a1file.a1f;*
$ exit
EXTERNAL PROTOCOLS
You can use the ZMODEM SZ and RZ commands as "external protocols" over a
connection you have established with C-Kermit, to a host or service that
does not support Kermit protocol. Start the file transfer on the remote
end, escape back to C-Kermit, give the SPAWN command, and then (for example):
$ define tt xxx:
$ rz
where xxx is the designation of the terminal device (TT or LTA) that you have
dialed out on. When the transfer is complete, LOGOUT from the SPAWN'd
subprocess and you'll be back at the C-Kermit prompt.
GENERAL FAILURES
...can occur for many reasons beyond Kermit's control, many of them related to
VMS system parameters or limits on the user or process: disk quotas, user
pagefile quotas (AUTHORIZE parameter PGFLQUO), system pagefile space filling
up, etc. See CKVINS.DOC (installation instructions) for details.
To increase a user's pagefile quota, tell AUTHORIZE to MODIFY
username/PGFLQUO=number. The system itself might be running out of pagefile
space, which would cause the system to grind to a halt and eventually crash.
You can check the system pagefiles with SHOW MEMORY/FILE: add up the "Free"
numbers for the [*]*PAGEFILE.SYS files and see if the total is big enough
(there should normally be at least 100K free pages on an active system). If
not, the system manager would use the procedure @SYS$UPDATE:SWAPFILES to
resize the files.
VMS C-Kermit can hang or crash with an "access violation" under certain
conditions when trying to hang up a modem-controlled device that is already
hung up; investigation shows that the hang or crash happens in VMS kernel
space, not in Kermit.
"Zombie" process can be left behind under certain conditions when a VMS
Kermit server has been sent a BYE command, particularly over a TCP/IP
connection (connection disappears before VMS has a chance to print its
"logged out" message, and now there is nothing to print it on).
One user reported "massive failures" when transferring files with VMS C-Kermit
through a particular kind of terminal server. She had followed all the
directions in the manual, the CKVINS.DOC file, and the CKVKER.BWR files (as it
was before this item was added). The terminal server uses TELNET protocol to
an DECstation 3000 Model 600 running VMS 6.1 and TGV MultiNet 3.3, using an
Equinox ELG48 terminal server. Later, she reported: "It turned out that
upgrading the software on our terminal server has fixed the problem. It's so
odd that the problem only occured after we upgraded from VMS V5.5-2 to V6.1,
since the terminal server worked fine before the upgrade. It's also weird
that this terminal server has always worked fine for our Suns, also. Here's
the details of the terminal server if you want to keep these details on file:
Equinox PBX 20 with ELG 48 board. The ELG 48 rev was 2.30. After upgrading
it to V2.33, everything works fine."
FILE OPERATIONS
As of edit 190, VMS C-Kermit supports append operations: the various logs
(packet, debug, transaction, etc) can be opened in append mode by including
the APPEND keyword after the filename, e.g.:
LOG TRANSACTIONS TRANSACT.LOG APPEND
An arbitrary file can be opened for output in append mode:
OPEN APPEND OOFA.TXT
and the SET FILE COLLISION APPEND option now works during file transfer.
When using append operations:
. Be careful not to append files of different types together, such as a text
file to an indexed file, or a fixed-record binary file to a text file, etc.
The result will generally be unusable.
. SET FILE COLLISION APPEND does not work when the FILE TYPE is LABELED.
This is deliberate: labeled transfers are designed to give you an exact
copy of the file, including attributes.
There is no facility in C-Kermit to distinguish between "overwriting" and
"versioning". SET FILE COLLISION OVERWRITE always creates a new version on
VMS.
BUG: As of this writing, APPEND operations do not use the RMS "first free
byte", and so start on a new block boundary.
FILE TRANSFER
File transfer modes (TEXT vs BINARY) are set automatically for each file when
sending. The SET FILE TYPE BINARY and SET FILE TYPE TEXT commands are ignored
when sending files. To force binary-mode transmission, use SET FILE TYPE
IMAGE. See the VMS appendix of "Using C-Kermit".
When sending binary files that have an odd record length, please note that
these files are actually stored with an even record length on disk. For
example, suppose DIR/FULL X.VDM says "fixed-length records, record length 17".
On disk, the file really has 18-byte records; each 17-byte record is padded
with a NUL (0) byte to make its length even; this is revealed by DUMP.
C-Kermit sends the raw records, INCLUDING THE PADDING. Thus, if you send such
a file to (say) DOS or UNIX for actual use, your DOS or UNIX application must
be coded to account for this -- if the record length is odd, add one to it.
If you send the file back to VMS, just tell VMS C-Kermit to SET FILE RECORD
to the original odd length, and the resulting file will be identical to the
original one.
Incoming files are rejected if the available space on the disk device is less
than the size of the file. However, the user's disk quota is not checked.
Quota checking could erroneously report that a user couldn't store a file for
a number of reasons: for example, the user has the EXQUOTA privilege, C-Kermit
is installed with EXQUOTA privilege (not recommended!), overdraft, etc.
Because of the large potential for denying a transfer that would fit, the file
is accepted regardless of the disk quota. This is consistent with the way
other VMS utilities work.
The file size shown in the file transfer display when sending a file might
be incorrect under certain conditions (but the file is still transferred
correctly).
Incoming files, if accepted, are always stored as a new file with the next
highest version number, even when FILE COLLISION is set to OVERWRITE or
or RENAME.
When you send a BYE command to a VMS C-Kermit server, it does not guarantee
that the VMS job will be logged out. If C-Kermit was SPAWN'd from another
process, only C-Kermit itself disappears in this case. Even if the whole VMS
session ends, if the user came in through a LAT terminal server, they will be
back at the "Local>" prompt and the phone line won't be disconnected.
Transfer of VFC (Variable with Fixed Control) files, such as those created
by DCL, is problematic, since the meaning of the control bytes is defined by
the application.
VMS MAIL messages: If you want to download mail messages to a PC (or other
non-VMS system), select the message of interest using the SELECT and DIRECTORY
commands within VMS MAIL, then EXTRACT/ALL to extract all the selected
messages to a normal text file, then use Kermit to SEND this file. Don't even
think about trying to transfer your mail file as-is to a non-VMS system; it is
a complicated indexed file, possibly containing pointers to other files, etc.
ZIP files: If you have trouble transferring ZIP files into or out of VMS
using BINARY mode, use IMAGE mode instead (SET FILE TYPE IMAGE). The same
applies to binary files created by VMS UNZIP.
When transferring files in LABELED mode, the file transfer display will show
the name the file was sent as, not the "true" name within the labeled file.
Also, note that a transfer may fail with an obscure error (can't create output
file) if there is something incorrect with the label information (for example,
if you specified that the file should be restored to the original directory
and you don't have privilege to write to that directory on this system).
DEC PATHWORKS file services normally create files in stream mode, but
this can be overridden when the file service is created:
$ ADMIN/PCSA
PCSA> SET FILE_SERVER SERVICE service-name/ATTRIBUTES=SEQUENTIAL_FIXED
The normal stream files will be treated as TEXT by Kermit. To transfer
PATHWORKS files that are really binary, such as executables, use IMAGE mode.
Reportedly, when VMS C-Kermit is in local mode and transferring a file (file
transfer display is showing) over a MultiNet TCP/IP connection and a broadcast
from a completing batch job (SUBMIT/NOTIFY) arrives, it crashes C-Kermit with
%SYSTEM-F-ACCVIO, access violation. The stack dump shows this occurs in the
netinc() routine while reading a packet (rpack).
Reportedly, when transferring files TO a VMS system over a LAT connection (for
example, from a PC equipped with PATHWORKS and MS-DOS Kermit), packet sizes
greater than 255 (some reports say 70!) cannot be used, irrespective of the
VMS SYSGEN parameters regarding MAXBUF, etc. The problem seems to lay in the
LAT protocol itself, or the particular implementation of it, whereby
applications are not informed of -- and cannot find out -- limits on
transmission. (And yet, others say they have no problems with file transfers
over LAT connections, even with packet sizes greater than 1000.)
SERIAL AND LAT COMMUNICATIONS
There is no way to select a serial communications speed higher than 38400 bps.
In many cases, speeds above 19200 -- or even 9600, depending on the actual
serial device -- might not be supported. Furthermore, hardware flow control
is not supported (see below). These are limitations of VMS, not of Kermit,
and they make it very difficult to use VMS with high-speed data-compressing
modems.
If you CONNECT to a modem or other device, and see a neverending stream of
messages, the terminal device probably has the /LOCAL_ECHO characteristic.
As of edit 189, C-Kermit attempts to turn off this characteristic
automatically as part of the SET LINE procedure.
The SET CARRIER command is not supported in the VMS version of C-Kermit.
Certain operations driven by RS-232 modem signals do not work on VAXstations
or other DEC platforms whose serial interfaces use MMP connectors (DEC version
of RJ45 telephone jack with with offset tab). These connectors convey only
the DSR and DTR modem signals, but not carrier (CD), RTS, CTS, or RI.
When used on a serial communication device, the HANGUP command (as well as
the CONNECT-mode escape command, H, and the HANGUP done by the DIAL command
when DIAL HANGUP is ON) takes at least 3 seconds. This is a FEATURE of VMS.
If a DIAL or SET SPEED command gives the error:
?ttbin: sys$qiow: %SYSTEM-F-NOLOG_IO, operation requires LOG_IO privilege
then either the user must be given LOG_IO privilege or else the device must be
given the SET_SPEED attribute. However, note that under certain versions of
VMS the TT2$M_SETSPEED bit in TTY_DEFCHAR2 is not properly propogated to LAT
devices. It is best to issue the command SET TERM/PERM/SET_SPEED LTA31: at
startup when the LTA31 device is initially created (which, of course, would be
done by a sufficiently privileged account).
During terminal connection (SET LINE) and file transfer over a serial device,
buffer-overrun or BYTLM-quota-exceeded messages might appear. It is essential
that any VMS system that needs to use Kermit or any other program to transfer
files over serial devices, especially when long packets or sliding windows are
to be used, be SYSGEN'd with large typehead buffers, and that user accounts
be given large BYTLM quotas. See CKVINS.DOC.
Note that LATmaster software (optional as of VMS V5.4-1, mandatory as of
VMS V5.5) requires a minimum Alt-Typeahead buffer of 2064 bytes. Thus, you
may already have increased the size. Kermit needs packet-size times
window-size plus a small amount of overhead.
To get around problems on systems where users have small BYTLM quotas, the
txbufr() routine in CKVTIO.C has been limited to reading 512-byte chunks at
a time from the communication device. This does not appear to have an adverse
affect on performance, but time will tell. If it does, a quick fix is to
recompile CKVTIO.C, defining CKV_IO_SIZE to be something bigger, e.g.
/define=("CKV_IO_SIZE=8192")
or whatever. A better fix might be to have txbufr() check the user's
remaining BYTLM quota before doing each read. But the overhead in doing this
might cancel out the advantage of doing it.
It is possible to SET LINE to an LTA (LAT) device, but correct operation
is reportedly dependent on the version of DECserver code and the VMS
version, and which patches have been applied. Correct operation has been
verified for DECserver 200 software V3.1-37 and VMS V5.4-3, unpatched.
If you use C-Kermit to SET LINE to an LTA device and receive a hangup message
immediately:
contti: ttiosb.status: %SYSTEM-F-HANGUP, data set hang-up
then:
. Make sure you've created an LTA port on your VMS system which is
mapped to the DECserver port that the modem is connected to.
. Can you use the VMS SET HOST/DTE command to connect to that line? If you
get the same error (which you should) there's a configuration problem in
the DECserver setup for that port.
. Are you trying to use that modem for both dial-in and dial-out
configurations? Try configuring different ports for dial-in and dial-out.
. In order for VMS to connect to the dial-out modem, it needs to see the
carrier detect signal asserted. If that signal isn't asserted, the server
will return a "hangup" error on the first character sent to the port.
C-Kermit's SET CARRIER command has no effect in VMS.
. Additionally, some modems want to see various settings on RTS/CTS and
DSR/DTR before they will accept input. If you have a breakout box and
someone who is skilled at using it, you can usually resolve these problems.
C-Kermit puts LAT terminal servers into PASSTHRU mode, which disables their
forward/backward session switch characters.
Reportedly, if you have CONNECTed out through a LAT device, the CONNECT-mode
escape command to hang up (<esc-char>H) does not work. Reason: unknown.
Cure: unknown (The LAT programming interface is very poorly documented).
Workaround: SET LINE <cr> to close the SET LINE device.
Reportedly, although Kermit can SET LINE to a LAT device and work OK, the
same can't be said for a "LAT group" (whatever that is). The user who
submitted this report said that this problem could be worked around by telling
VMS to SET TERM <blah> /NOALTYPEAHD before starting Kermit (take this one
with a grain of salt).
Reportedly, to use C-Kermit with a LAT device under LATmaster, the associated
terminal device must be set /NOREADSYNC.
FLOW CONTROL
The SET FLOW RTS/CTS command is not supported in the VMS version of C-Kermit.
VMS does not support RTS/CTS (hardware) flow control.
In VMS, flow control is governed by two SET TERMINAL parameters: /TTSYNC and
/HOSTSYNC. TTSYNC lets the terminal control the flow of data from the host
and HOSTSYNC lets the host control the flow of data from the terminal. In
general, these are implemented as Xon/Xoff flow control in each direction, but
on LAT and TCP/IP connections, they can also affect the internal networking
protocol.
When C-Kermit is in "remote mode", i.e. it is on the far end of a connection,
and is not establishing a connection itself, it uses your current VMS SET
TERMINAL parameters for flow control during command processing. During packet
mode, however, it obeys your C-Kermit SET FLOW-CONTROL setting, which, to
ensure the chances of lost data are minimal, is XON/XOFF by default; XON/XOFF
corresponds to VMS SET TERM /TTSYNC /HOSTSYNC.
When C-Kermit is in "local mode", i.e. it is being used to establish a
connection with SET LINE or TELNET, there are two components to your
connection: the part between your terminal and C-Kermit (call this "Part A"),
and the part between C-Kermit and the remote computer or service that you have
connected to ("Part B"). At all times, the flow control used on Part A is
governed by your VMS SET TERMINAL parameters, and the flow control used on
Part B is always governed by C-Kermit's SET FLOW-CONTROL command.
If you are using C-Kermit in local mode to access a remote host to use the
EMACS editor, you might find that the Ctrl-S (Search) and Ctrl-Q (Quote)
commands don't work -- your screen and keyboard "freeze" when you type Ctrl-S,
and Ctrl-Q seems to be ignored. This means that your VMS command terminal has
the /TTSYNC characteristic; Ctrl-S and Ctrl-Q are being used for flow control
between your terminal and the VMS system -- the remote system and EMACS never
see them. There are two ways around this problem:
1. Tell VMS to SET TERM /NOTTSYNC before starting C-Kermit. In this case,
you are in danger of losing data on the connection, particularly if your
connection to VMS is through a LAT device.
2. Leave the /TTSYNC charactistic in force and use the long forms for the
EMACS commands: ESC-X Search-Forward and ESC-X Quoted-Insert. Or assign
these functions to other EMACS keys in your EMACS initialization file.
NETWORK COMMUNICATIONS
There is (as yet) no support for initiating connections over DECnet, nor for
VAX/PSI. Certain types of TCP/IP are supported (including DEC TCP/IP (UCX),
CMU-OpenVMS/IP ("CMU/Tek"), TGV MultiNet, Wollongong WIN/TCP or PathWay,
Process Software TCPware); other types: not yet (e.g. Fusion).
The TCPware version works correctly with TCPware versions 4.1-2 or later;
earlier versions, such 3.1-3, have a bug that can result in failure of
C-Kermit to make network connections, with a message like:
?contti: network sys$qio: %SYSTEM-F-IVCHAN, invalid I/O channel
Process Software recommends upgrading to the current TCPware release.
If you enter the VAX from elsewhere through a TELNET connection, and the VAX
is running CMU-OpenVMS/IP, Fusion, or DEC TCP/IP (UCX), you might notice that
file transfers into the VAX fail almost immediately. If this happens, it is
most likely the result of small VMS typeahead buffers. See CKVINS.DOC for how
to increase typeahead buffer sizes, or work around the problem by telling VMS
C-Kermit to ask for smaller packets, for example:
C-Kermit>set receive packet-length 65 ; (Use the longest length that works)
The UCX version of Kermit works on MultiNet systems too, because MultiNet
automatically goes into "UCX compatibility mode" when a UCX application is
run.
You can also use the non-network version of C-Kermit on a MultiNet system to
make TCP/IP connections as follows:
$ telnet/create foo.bar.baz.com
Trying...
TELNET session now connected to _NTY5:
%DCL-I-ALLOC, _$4$NTY5: allocated
$ kermit
C-Kermit 5A(189), 30 June 93, OpenVMS AXP
Type ? or HELP for help
C-Kermit>set line telnet_nty
C-Kermit>connect
etc...
When using the CMU-OpenVMS/IP TCP/IP transport, assign the system logical
INET$SERVICE_TELNET_TCP to the telnet port as follows:
$ DEFINE /SYSTEM INET$SERVICE_TELNET_TCP 23
This is only required if the -j option is used without specifying a port to use
(e.g. -j host). If this logical assignment is not made using `-j host' option
will fail with the error:
%CKERMIT-E-FATAL, can't open host connection
The default port, hardcoded in C-Kermit, is 23. Another port may be specified
using the -j option as `-j host:port'.
SET INPUT ECHO OFF seems to have no effect when given to VMS C-Kermit and
the INPUT command is reading from the console terminal.
PERFORMANCE
If you are experiencing very poor performance on serial connections, use the
VMS command SHOW TERMINAL to make sure that the terminal device has the DMA
(Direct Memory Access) characteristic. If it does not, try setting it (or
get your system manager to, in case privilege is required):
$ SET TERMINAL device_name /PERMANENT/DMA
On some slower VAX models with built-in serial ports, such as the VAXstation
3100 or MicroVAX-II, receiving files on serial ports at (say) 19200 bps
results in high CPU utilization, slowing down the system for other processes.
This is partially because VMS C-Kermit's serial i/o routines need to be
restructured along the same lines as the network ones (nonblocking buffered
reads), but also because on certain systems, such as the VS3100, serial
ports interrupt the CPU every time a character arrives. Most VMS systems
nowadays, however, support either DMA for serial port i/o, or have their
users coming in through terminal servers.
A similar problem is observed when C-Kermit is receiving files on a VAX/PSI
(X.25) system, attached to a certain X.25 network (Autonet), but not others,
attached via a DSW42 interface: huge numbers of I/O requests drive the load
way up. Reportedly, "this is due to a VAX PSI feature called Synchronized
Echo Protocol (SEP), which is supposed to coordinate echo by the X.25 PAD,
e.g. when typing ahead. When disabling this feature, the file transfers
proceed fast and efficient. This feature is a network-specific X.25
"facility" negotiated at call-setup time, not an X.3 parameter (standard or
otherwise) -- Facility Number 66 (decimal, or 42 hex). It could not even be
set or viewed by the user or the VAX administrator. It had to be disabled by
the network provider. I think that most X.25 networks do not even implement
this feature and thus is it not common issue. In any case, in our situation,
I asked the X.25 network provider to disable this feature, and now C-Kermit is
performing efficiently, but now, of course, echoing (e.g. of material that is
typed ahead) is no longer synchronized."
(End of CKVKER.BWR)