Columbia Kermit

home *** CD-ROM | disk | FTP | other *** search

/ Columbia Kermit / kermit.zip / archives / ckc190.zip / ckcker.2do < prev next >

Wrap

Text File | 1994-11-12 | 25KB | 511 lines

CKCKER.2DO -*-text-*- C-KERMIT TO-DO LIST Most recent update: Mon Sep 5 14:49:20 1994 Ideas for future releases of C-Kermit (in no special order): ***** Let GETC also store into long variable (macro) names, like MSK. Put OPEN READ (etc?) files on a stack -- hyperlinks and all that... Remove calls to ttopen() from ckuusy.c. See comments in ckuusy.c. Don't use [...] in default prompt, since it is an NRC character. Or maybe add IF NRC and underlying support (isnrc(charset))... Allow arithmetic expressions (\feval() syntax) within array subscript brackets. Maybe also in FOR-loop controls? Add SET SERVER LOGIN and supporting code. In the UNIX version, record the make command line or at least the target name ($@). Difficult because of nested targets and quoting rules... Add PRESERVE <list of global variable names or macro names>. Used in TAKE files or macros. The preserved items are automatically restored by popclvl(). Add SAVE command to save things: ROLLBACK-BUFFER, SETTINGS { COMM, TERM, FILE, PROTOCOL, ... } Add a function that fully evaluates a macro definition, unlike \m() or \fdefinition(), which only return the literal definition, unevaluated. SET DEBUG blah -- to select debugging of different kinds of things and/or different levels of verbosity. Give warning message when WHATAMI changes file or naming mode? Add a post-transfer-check feature to the protocol, involving file checksum/CRC and/or size? Add \Kmouse to toggle mouse on/off. SET SPEED help message is very misleading if not in local mode. Just error out right away. . If a packet transmission fails, we have to retransmit it. But that is likely to fail for the same reason (too long, too fast). We can't chop into subpackets for well-known reasons, but we can insert delays when retransmitting... . ASSIGN/DEFINE \%a { foo } does not strip the braces! Oops. . Make IF EXIST also work for wildcards ? . Add REPLAY command to OS/2 C-K. . Then add slide-show capability (e.g. via <OS>PAUSE<ST>). . When receiving, if transfer fails, give message about partial file kept. . No way to have a literal comma in a \Function() argument -- no kind of quoting works. . Make hitting CR during file transfer work; make retries get updated on f.t. screen. . Allow \Kverbs in OUTPUT commands ??? . Pop back to command mode immediately if carrier drops during file transfer: . At beginning of transfer, if serial, check if carrier on, set flag. . Upon read timeout, if flag on & carrier off, then ttclos(). . Add commands to select/control this behavior, e.g. SET SERVER blah blah.. . Need SET TERM HEBREW KEYBOARD-MODE { ENABLED, DISABLED } If there is a real Hebrew keyboard, we don't need to use abc's. . Let user provide own keymap: SET TERM HEBREW KEYMAP { BUILTIN, FILE <filename> } (format of file?) . Adapt SHOW VAR/KVERB keyword-arranging code to command parser. . User-definable DIAL HANGUP string (hard). . SET CLOCK ... to display elapsed time of each CONNECT session or somesuch. . Add ISO 2022 charset invocation & designation to OS/2 C-K. . Add host directed keyboard-mode switching to OS/2 C-K (CSI ? 35 h/l). . Add VT320 reports to OS/2 C-K. . Add SET TERMINAL CONTROLS { 7, 8 } to OS/2 C-K. . Add SET TERMINAL TABSTOPS for OS/2 C-K, and have SHOW TERMINAL display them. . Enable DEC UDK's in OS/2 C-K. . Enable local echoing of escape sequences in OS/2 C-K. . Allow host access to bottom line in OS/2 C-Kermit, e.g. sysline. . Integrate DEC Technical charset with other charsets, handle in same way. . SET SERVER TIMEOUT <blah>, to exit server mode automatically after blah amount of time. . Add an option to pop out of server mode automatically if the connection closes (e.g. carrier drops). . Make SET CARRIER ON (and AUTO?) apply to packet mode too. . Add SHOW ARRAY <name-of-particular-array> to dump contents of whole array. Or somesuch... . Use knowledge of controlling-terminal rows & cols when formatting file transfer display screen, keyword and file menus, etc. . When reading TAKE files, keep track of the line number. Make error reports show filename and line number. Maybe allow an optional label in the TAKE command, telling Kermit to GOTO the label in the file before starting to execute commands. (Forget the following -- On connection loss, ttol() and friends should just call ttclos() and return failure; see above). . When ttol() fails, it does not stop the protocol immediately. Instead, we have a lengthy series of timeouts and retransmissions until the retry limit is reached. Solution: 1. Change ttol() to return -3 on fatal error, -1 on possibly correctable errors. Fatal errors include connection loss, configuration problems. 2. Change spack() to pass back -3 return code. 3. Change sdata() and all other packet-sending routines to pass back spack's return code. 4. Change ckcpro.w to send error packet and terminate the transfer if it ever gets a -3 return code from any of these routines. . Similarly for ttinl(), ttinc(), ttoc(), and friends. . Add Hebrew-7 to CONNECT-mode esc seq recognizer... . Add SET ALARM / IF ALARM for compatibility with MS-DOS Kermit. . Add SET COMMAND (or maybe CONSOLE) SPEED, which lets user tell C-Kermit what the console speed is, e.g. when the operating system lies about it. . Always receive into a temporary file, for which a unique and nondestructive name is computed. Then, if the transfer fails, delete the temp file. If the transfer succeeds, rename it according to the prevailing FILE COLLISION setting. Advantage: when FILE COLLISION is OVERWRITE, this prevents a failed transfer from destroying the original file. Disadvantage: disk filling up, difficulty of modifying all C-Kermit versions to manage temp files. . SELECT { NEWER, OLDER, ... } <date-time> { LONGER, SHORTER, ...} <size> etc... Selection criteria for files to send. After building up a list of criteria (should they be ANDed or OR'd ?), the SEND <filespec> or MSEND ... command will send the files that match the filespec(s) AND meet the criteria. . Make SET TERM CR-DISPLAY also apply to INPUT echoing (oops, not as easy as it looks with new input-echo-buffering). . Add REMOTE MESSAGE. . Add REMOTE RENAME. . Add a command to let CR be sent totally bare on a TELNET connection, instead of as CR-NUL or CR-LF. Maybe institute TELNET BINARY mode for this. . Add a command to turn off dynamic packet length, and/or to set limits. . Add transfer mode (text/binary) to STATISTICS display (not easy because it might have been a local mode rather than global, or might have changed from file to file). . Clean up and incorporate Fulvio Marino's I/O buffering code, and eventually remove the old I/O code. First for UNIX, then elsewhere. . Convert Sys V / POSIX versions to not need ^\ before local-mode file transfer interruption commands (do nonblocking reads of console) (now possible because of Fulvio's code?) (or just enable CK_POLL everywhere that poll() is available). . Let Kermit run as a TCP/IP server, instead of FTP, listening on its own socket, so it doesn't have to use Telnet and ptys - this will make it go MUCH FASTER over TCP/IP connections. Get TCP socket number officially assigned. . Allow various kinds of network server operation. By default (in UNIX at least), Kermit chroot's to /pub/ftp, user anonymous. Or maybe also an option for no chroot, to allow anonymous access to entire file system, based on permissions of individual files and directories (as on most non-UNIX systems). Like ftp server, client can also log in as a specific user. . Also allow UNIX (and other?) Kermit servers to run as user's login shell. . Add LOGIN / LOGOUT support in server (it's already in the client); in the client, accept LOGIN and LOGOUT as synonyms for REMOTE LOGIN and REMOTE LOGOUT. . Add a more convenient style of client/server operation, in which SET commands given at the client take effect on both the client and the server (like the BINARY command, when given to an FTP client). . Add an FTP client. . Do something about +++ appearing in non-data portions of Kermit packets. (No, can't be done without fundamentally changing the protocol.) . Add SET MODEM USER-DEFINED. Just make a blank mdminf structure & pointers for it, and then give users SET commands to fill it in. . Add SET DIAL COMPRESSION and SET DIAL ERROR-CORRECTION for generic enabling/disabling/selection of these items. Get rid of SET DIAL MNP, and incorporate it into the new commands. These commands should allow specification of various methods AND fallback schemes. (Or not...) . Add support for more types of modems: Practical Peripherals, Multitech, etc. . Add optional system-wide dialing log, for billing dialout modem users. . Need a better mechanism for handling speed-matching modems - recognition of interface speed AND connect speed, setting of timeouts, etc, accordingly. . Add a command-line option for phone number to dial. . Fix DF2xx modem support for bidirectional orientation in Ultrix 3.1 & later. . ttinl() is getting its fingers too much into the protocol. I'd like to remove it altogether and just let rpack() do all that stuff, calling ttinc() to simply read characters from the communication line, and let ttinc() handle all the buffering. Better yet, let's turn ttinc() into a macro that calls some kind of ttfillbuf() routine whenever it needs to. This would remove several layers of function calls from the packet i/o, and it would let us handle TELNET options during file transfer, etc etc. Fulvio's code could come in handy here. . Call parchk() only at beginning of transaction, not for each packet. . Handshake is done wrong in ttinl(). ttinl() should ALWAYS be length-driven. Handshake should be looked for AFTER the packet has been read based on its length. Otherwise we can't use the printable-handshake for DDK. Once this is fixed, allow SET HANDSHAKE CODE accept any number 0..255. . Change ck?fio.c to return the text-file byte stream with CRLF delimiters, rather than making the system-independent function getpkt() do it. Better, yet make the file reading modules return some kind of end-of-record indication for both text and (where applicable, as in VMS) binary files. . When uploading text files to UNIX, *all* CRs are stripped. The way it SHOULD work is: bare CRs are kept, bare LFs are kept, and only CRLFs are treated as line terminators. This will take a major reworking of decode() in ckcfns.c. It's not obvious how to do this in a clean way, because the CRLF might be broken between packets. . Add code for breaking outbound attribute string up into more than one attribute packet. . Make A-packets fit in negotiated packet length. . Add file protections/permissions to attribute packet processing. . Support for "End of attributes" attribute = Attribute "@" (K-370 has it). . Add mechanisms for sending and receiving directory trees in a system- independent manner. New protocol must be involved for telling the remote Kermit to switch to (and if necessary, create) a new directory. Could be something as simple as a new "set file names" option. Maybe add code for recognizing and converting VMS, MS-DOS, etc, directory names based on system-type field from Attribute packet. Or preferably, devise standard syntax for file specifications (device, path, etc, maybe an X.400-style notation). . Dynamic adjustments of timeouts based on transfer rate, system load, etc. . Calculate delay between S and its ACK & use it (e.g. for SLIP connections) . Design protocol for file transfer checkpointing and add it to the program. Must work for both text and binary files, and be independent of operating system, file format, etc. . Ditto for higher levels of compression. . Separation of TERMINAL and COMMAND flow control . Add RTS/CTS support for more systems. . SET COMMAND SHIFT-IN/OUT { OFF, ON } ??? (maybe not necessary) . Handle TELNET negotiations at all times, even during file transfer... . When sending a file whose name contains international characters, translate the name to ASCII because we haven't negotiated a transfer character-set yet. . Add a SHOW command for each SET command? . Add a way to redirect output of REMOTE commands to a file. . Add \v(wordsize) == sizeof(int). . Quoted comma ("\,") in macro definition should not cause break to new cmd. (or maybe it should -- meaning of "\" would depend on level of nesting) . Find out what the system's interrupt character is, and refer to it in messages, rather than always saying "Ctrl-C", ditto for suspend character. . Let user SET the command-editing characters: SET COMMAND { CHARACTER-DELETE, WORD-DELETE, LINE-DELETE, REDISPLAY } ... . REMOTE SET FILE CHARACTER-SET <string>. Needed for anonymous Kermit server. . Add Japanese & Chinese ISO 646 (same) as FCS's. . Allow multiple actions on command line, e.g. -s, -g, -f... . Add IF INPATH <filename> <command>. . Add CDPATH handling to CD command (UNIX). . System-wide transaction log -- logs all file transfers by all users in a common long, showing username, filespec, timestamp, mode, etc (forget it -- locking & contention problems). . Make log files use the file backup mechanism. . Change zltor() to strip .~nn~ rather than converting it. . Add a command like VMS PURGE for Kermit backup files. . Add a command to select alternate forms for Kermit backup file names. . Make ckufio.c zstime() more flexible (steal code from ck9fio.c). . Define a signal people can use to kill Kermit, that it can catch, call doexit(), send error packet, whatever.... . Multiple sessions (network, serial port, etc). Session info structure, session manager. . INPUT -1 (or 0?) <string> (or something like that) to wait forever. . Make SET / SHOW KEY read extended keys on NeXT, SPARCstation, Mac... . SET TRANSLATE command as in MS-DOS Kermit for user-defined character translation during CONNECT. . addhlp() should break stuff up into separate lines if too long. . Add hh:mm:ss argument for PAUSE and WAIT, as in MS-DOS Kermit. . Add \ftranslate(string,cs1,cs2) for translating character set? . Allow more than 9 macro arguments. Put args in a special array, or SHIFT. . Add hook for UNIX file input and output filters, maybe SET { SEND, RECEIVE } FILTER <command>. Use with compress, tar, tr, crypt, etc. . Add IF READABLE (like IF EXIST). . Redo input() function to be a state-table switcher? . Add cmchar() to parse a character (control or otherwise). . Learned scripts - LOG SCRIPT <filename>, CLOSE SCRIPT, record CONNECT session as a series of INPUT, PAUSE, OUTPUT, etc, commands. But what about passwords? . Add ON { ERROR, HANGUP, INTERRUPTION, EOF, ... } . SET RECEIVE MAIL-COMMAND <template>, with substitutable parameters. Receiver parses {file} and {options} in template, substitutes actual filename and options received from client program's MAIL command. . Similar idea for SET RECEIVE PRINT-COMMAND. . Make UNIX version use pipes for incoming mail and print files. . Add more SET FILE NAMES options, like literal but with pathname stripped. Allow user to specify a template for fields, capitalization, punctuation. Like SET FILE NAME TEMPLATE [[node] sep] [[dev] sep] [[dir] sep] ... Or maybe somthing simpler like SET FILE NAMES [UN]CONVERTED [UN]STRIPPED. . Add supporting code to SET HOST for SET NETWORK DECNET (VMS, Ultrix). . Add supporting code for SET HOST X.25 for VAX PSI and other X.25 products. . Add supporting code for SET HOST via ISO/OSI in BSD Networking/2. . Find a way in UNIX to check available disk space and refuse incoming files that are too big. Tricky, maybe impossible, for many kinds of UNIX. . Internationalize: . Segregate all message strings into a separate module, to allow translation into other languages - a "resource file"? Use xstr. Complicated because many different character sets can be used for each language. . Make string comparisons and "tolower"/"toupper" work with int'l character sets (inlcuding in INPUT/REINPUT). . Fix \v(date) to report date in different languages, formats. . Fix \v(day), \v(nday) to handle non-English day names. . Fix \v(ndate) to handle non-English month names. . GETOK command answers: Yes, Ja, Si, ... . etc etc. . Implement server end of REMOTE KERMIT. Means redirecting all Kermit's screen output to the packet builder -- no more printf, puts, putc. (Really not worth it!) (Or maybe use temp files? Yuk.) . Reorganize the whole program to define and isolate system depencies better, perhaps writing a "sysgen" procedure (like Larry Wall's Configure program). . Eliminate global variables (ha ha). . Terminal emulation for UNIX workstations: key definition, screen rollback, printer control, etc. This has already been done for Macintosh and OS/2. Requires separate ck?con.c for each kind of workstation. Even then, there is often no way to get at real keyboard and screen. . IBM 3270 terminal emulation: tn3270, perhaps others. . Add support for Kerberos authentication (see 4.4BSD telnet). . Write an alternate ckucmd.c that produces screen menus rather than interactive commands, given the same command calls (cmkey, cmtxt, cmnum, cmcfm, etc). For UNIX ttys, use the curses library. For NeXT, Macintosh, Amiga, OS/2 PM, SPARCstations, etc, make real pop-up menus, allow mouse input, etc. . An X-Windows server/client interface? SunView? Motif? NextStep? etc? . Before even thinking about the two items above, design a standard for what the interface should look like. How to do user-defined translations: SET FILE CHARACTER-SET USER-DEFINED SET XFER CHARACTER-SET <valid-xfer-charset> SET USER-TRANSLATION FROM <tcs> xxx yyy ; for incoming files SET USER-TRANSLATION TO <tcs> yyy xxx ; for outbound files Applies to <tcs> + USER-DEFINED FCS. Can have one pair for each TCS. Now announcements work right, etc etc. DUMP USER-TRANSLATION <tcs> [ <file> ] ; list tables (to file) For C-Kermit, we have 4 TCS's, so 8 x 2 x 256 = 2K bytes, not bad: . Add FC_USER FCS . Add 6 tables (not supported for Kanji) . Initialize each table to identity function . Add 8 functions (even for TRANSPARENT, but NULL for Kanji) . Figure out a way of telling user whether table has been defined. . Add SHOW USER-TRANSLATION <tcs> [ <file> ] (= a bunch of SET USER-TRANSL commands, with comments) . Add DUMP USER-TRANSLATION <tcs> [ <file> ] (= just the numbers, comma-separated? space-separated? one per line?) . Add LOAD USER-TRANSLATION <tcs> <file> (= read table written by DUMP, watch out for value and table-size overflow) . Add some kind of built-in test pattern? ---------- FROM USERS From: Mark Whitis <whitis@hopper.itc.virginia.edu> Subject: Re: Modem Sharing Over TCP/IP To: Frank da Cruz <fdc@watsun.cc.columbia.edu> Date: Wed, 28 Sep 1994 22:04:02 -0400 (EDT) > > Great. I take it this is a newer version than I currently have. > > Does C-Kermit yet support kermit over an arbitrary unix program via > > pipes (popen)? This would let you run kermit over telnet, rlogin, tip, > > rsh, etc. > > > I'm not sure what you want here. It already lets you make telnet > connections. > The new REDIRECT command lets you run any other program under Kermit, > redirecting its stdio to the communication channel, which can be serial > or TCP/IP. That is a neat feature, which allows people to do things like run zmodem on top of kermit. What I had in mind was a similar feature for the bottom end. Consider the following scenarios: - I am using a poorly managed unix system at someone elses facility. They do not have kermit installed. Tip is installed and /dev/cua0 is accessable only via setgid uucp programs such as tip. I don't have root. So, I copy kermit to ~/bin and issue the following hypothetical commands: $ kermit C-kermit> set line pipe C-kermit> set pipecmd "tip 555-1212" C-kermit> c tip: dialing 555-1212 Connected login: foo password: $ kermit -ix C-Kermit server starting. Return to your local... ^]c C-Kermit> get foo.bar - I wish to talk to a network dumb system connected to a DEC LAT terminal server and emulate a terminal and transfer files using kermit. I have suns decnet software (the command "dnilogin" is equivalent to the VMS "set host" command): C-kermit> set line pipe C-kermit> set pipecmd "dnilogin dumbhost" C-kermit> c dumbhost login: foo etc. - I have (hypothetically) a modem pool which is set up to only accept connections on a protected port from a trusted host. On the trusted host, there is a program "outdial" which runs setuid root to access the protected port but is executable only by members of the unix group "outdial". Or, perhaps it lets anybody run the program but only members of the group "lddial" are allowed to dial a non-local number. The program makes the connection to the modem, issues the dialing string, and disables the escape sequence on the hayes modem. C-kermit> set line pipe C-kermit> set pipecmd "outdial 555-1212" C-kermit> c outdial: connecting to modem pool outdial: initializing modem outdial: dialing...555-1212 outdial: connected. - Paranoid user wants crypto protection. They have a program called scramble which runs another program in a subprocess and scrambles stdin is connected to the programs stdin and scrambles stdout is connected to the other programs stdout but stdin and stdout are encrypted on the way. There is some mechanism, perhaps, to turn encryption on and off so you can handle logins (although that may be one of the things you want to scramble). C-kermit> set line pipe C-kermit> set pipecmd "scramble -waittrigger telnet secret.spook.gov" or set pipecmd "scramble -waittrigger tip 555-1212" C-kermit> c login: flag password: $ scramble -sendtrigger bash %$#Enable-Scramble#$% (triggers master to switch into scramble mode). bash$ (similarly, you could create a script "scramblebash" which just has "scramble -immediate bash" and then put scramblebash as your login shell; that way, your password would be scrambled too.). This is a more general solution to the problem the TCP code was added to C-Kermit to solve. Yes, some of the above scenarios arise because of people doing things the Wrong Way(TM) but in the real world you have to deal with this and part of the beauty of kermit as compared to other protocols is that it can be configured to work under adverse conditions. You could deal with most of the above situations in a kludgy fashion by using the new TCP features of kermit: C-kermit> set line tcp/ip C-kermit> set tcp host localhost C-kermit> c login: foo Password: $ scramble telnet secret.spook.org $ outdial 555-1212 $ tip 555-1212 $ dnilogin dumbost However, this requires the user to login again, adds additional intervening layers to screw things up, and creates additional unix processes (telnet, login, shell), uses up pttys, etc. If anything, implementing this new feature is easier to do than the redirect command. You create two pipes using the pipe(3). You fork. You replace your stdin and stdout in the child process with one end of each pipe and redirect stderr to stdout. Then you issue a system call using the string supplied by the user. Back in the parent, you make two (FILE *) for the pipe using fdopen (assuming you use the higher level stdio calls) in C-kermit. One file descriptor is used for reads, the other for writes. Naturally it helps if you use two separate descriptors throughout the kermit code to access the ioport. To open a regular com port or TCP connection, you just point both to the same stream in your open or close. This is something I have thought of adding and submitting as a patch but have not got arround to it yet. ------------------- (End of CKCKER.2DO)