Uwe Ohse

available software

lrzsz: free x/y/zmodem implementation

Features of lrzsz

• very portable, automagically configured with GNU autoconf.
• crash recovery.
• up to 8KB block sizes (ZMODEM8K).
• internationalized (using GNU gettext). German translation of the programs output exists.
• far more secure than the original sources.
• high performance. say `make vcheck-z' and have a look at the BPS rate - i recently saw 1.4 MB per second transfering a large file through pipes (on a I586/133 system. Beat that!).
• good blocksize calculation (tries to compute an optimal blocksize based on the number of errors occured).
• It's free software.

Downloading lrzsz

Recent changes

• Version 0.12.20 - December 1998, Uwe Ohse
  
  • works on BeOS and stone-aged SCO (sco-3.2v4.2)
  • pubdir-"feature" works again.
  • "make rpm" creates a rpm file.
  • "optimal blklen calculation" was too aggressive, it now does nothing if the user demands fixed blklens.
  • various smaller and medium bug fixes.
  • a more or less important security bug is fixed (stupid use of /tmp in a piece of code which is rarely used).
  • lrz uses umask to make files unreadable which receiving them.
  • "sh systype | mail uwe-generic-counter@ohse.de" sends a success report with a description of the system type.
  • --enable-syslog is now default
• Version 0.12.19 - January 1998, Uwe Ohse
  (sorry, i forgot to announce that version here. -- uwe, 9.5.1998)
  • 0.12.18 was broken, lsz crashed if receiver found an CRC error.
  • lrz options "--rename" and "--escape" didn't work.
  • lrz didn't implement senders "overwrite-or-skip" option.
  • added dejagnu testsuite. Maybe you need a dejagnu snapshot to use it.
• Version 0.12.18 - November 1997 keep off - this version has a major bug
  • syslog output now includes user name.
  • new script lrzszbug, to be used for bugreports (untested)
  • lots of compiler warnings (egcs -Wparanoia [many -W]) removed.
  • new options --tcp-server and --tcp-client ADDRESS:PORT for both programs. (this is not too much code bloat, and somewhat useful for me - but i suspect nobody else will need it. Oh well)
  • updated to gettext-0.10.32
  • updated to automake-1.2
• Version 0.12.17 - August 1997
  • internal cleanup.
  • portability enhancements by (Philippe De Muyter <phdm@info.ucl.ac.be>)
  • lsz has a new option "--tcp" (no shortopt implemented). lsz transmits one file over normal stdin/stdout (a control file), then opens a tcp connection to transmit all other files.
    That might be useful if your telnetd is really stupid.
  This version was _not_ put on the ftp/http servers.
• Version 0.12.16 - March 1997
  • major performance improvement (less CPU time needed - don't expect faster transfers over slow lines). make vcheck now show about 50% more throughput.
  • updated to gettext-0.12.27 and automake-1.1l (automake-1.1l bug: AC_SUBST in AM_PATH_PROG_WITH_TEST leads to a "$1=@$1@" line in Makefile.in. I hacked around it in /usr/share/aclocal/gettext.m4).
• Version 0.12.15 - February 1997
  • lrzsz now compiles even with pre-ANSI-compilers (tested with the HPUX 9.00 bundled compiler - what a bad program. Shame on HP).
  • lrz supports a new option --o-sync, which forces it to open the output file in synchronous write mode. Don't use it unless you are using interrupts if update writes to the disk.
• Version 0.12.14 - January 1997
  • lrzsz now compiles cleaner on SCO, HPUX.
  • improved error reporting (i think there are still possibilities for further improvements, if anybody case spare time :-)).
• Version 0.12.13 - January 1997
  • no user visible changes. just a maintainance release.
  • updated to gettext-0.12.26.
• Version 0.12.12 - January 1997
  • lrx and lrb (aka lsz --x/ymodem) now default to 128 byte block length (to fix interoperatability problems with some Xmodems [USR courier flash upload]).
  • lrz didn't recognize every short option.
  • minor performance tweaks.
  • replace mktime() if needed.
  • updated to autoconf 2.12.
• Version 0.12.11 - October 1996
  • lsz/lrz recognize "rshell" as another name for the restricted shell.
  • lrz/lsz recognize a new option --stop-at HH:MM (stop transmission at HH:MM), and --stop-at +N (stop in N seconds).
  • lrz/lsz recognize a new option --delay-startup N: wait N seconds before doing anything.
  • doesn't hang anymore on BSD machines after getting a timeout (BSD restarts system calls after an interrupt - this made timeout code useless).
  • SIGINT handling turned on unter linux too (i still don't know why it was turned off under linux (and only there)).
  • sz - now prints better transfer statistics after if reading from a pipe.
• Version 0.12.10 - September 1996
  • lsz resends init string if it doesn't receive rz's init.
  • improved "make check". By Philippe De Muyter <phdm@info.ulb.ac.be>
  • "sz -" should work again. This will not work if sz cannot read from stdout.
  • portability enhancements by Philippe De Muyter <phdm@info.ulb.ac.be>
• Version 0.12.9 - September 1996
  • New options --min-bps N and --min-bps-time M: If BPS rate falls under N for at least M seconds (default: 120) transmission will be stopped.
  • added some missing error messages.
  • updated manual pages.
• Version 0.12.8 - August 1996
  • sz and rz now know about a new option: -E, --rename: change name if target exists.
  • new option -T, --turbo for sz: sz doesn't escape 4 special characters if this option is given (this should not make problems with any rz, but could be problematic on certain links where this characters have to be escaped).
  • debugged blocksize calculation.
  • -+, --append option fixed.

Future developement

• enhance portability.
• use VMIN code some day (it's not hard to implement. it's hard to automagically check if VMIN works. It doesn't on older BSD versions).
• make argv[0]-recognition of X/Y/Z-mode work with program_name_transform.
• write a decent documentation.

Comments

• i will not say anything about the original unix zmodem sources. Just look at them and you'll know (even lrzsz is still worse than i like it to be).
• the zmodem protocol is a pretty good example of a proprietary protocol:
  • There are some proprietary, undocumented protocol extensions made by the original author. Furthermore they are implemented in such a way that nobody else can extend the protocol in a safe way.
  • There is no way to get extensions included into the standard.
  • The protocol is not too good designed. It's impossible to extend it in a safe way to transfer files in both directions.
  • Security was not an issue to the author. The ability to let the receiver execute any command the sender tells him is nothing more then a really large security hole (as large as running older sendmail versions). Hackers dreams came true.
• On the other hand it has some really good points - especially it is high efficient, and is, if implemented correctly, usable over a number of transmission channels nobody thought of at the time the protocol was designed.

Send comments to uwe@ohse.de.