Make editors slightly moist


Happiness comes out of contentment, and contentment always comes out of service.
--Harbhajan Singh Yogi

There is no easy walk to freedom anywhere, and many of us will have to pass through the valley of the shadow of death again and again before we reach the mountaintop of our desires.
--Nelson Mandela

There are no happy endings in history, only crisis points that pass.
--Isaac Asimov

From me to you

All projects and possible support mailing-lists are hosted for free by, some of them are mirrored, also for free, to and by Thank you!

S - n a i l — v14.8.5, 2015-09-05 (announcement)
Downloads: .tar.xz / .tar.gz   …   Mail: List /
Code: {*master*,next,[crawl],timeline,heirloom} git:// (browse) (browse)

S-nail is a mail processing system with a command syntax reminiscent of ed(1) with lines replaced by messages. It is intended to provide the functionality of the POSIX mailx(1) (Wikipedia) command S-nail heraldic animal and offers (mostly optional) extensions for line editing, IDNA, MIME, S/MIME, SMTP and POP3. It is usable as a mail batch language. ... read the complete manual.

S-nail ships with archlinux (Wiki) and KaOS. S-nail packages are available for Crux, Debian, OpenBSD and The Void (Linux) distribution. In the archlinux User Repository (AUR) there is s-nail-git, which provides a direct git(1)-to-package layer ([master] by default). Also i'm really happy that “Fossies” — the Fresh Open Source Software Archive selected S-nail for its archive! The goal of S-nail is to provide an efficient and standard compliant implementation in 2020, the 42nd birthday of the Berkeley Mail codebase. S-nail is based on the BSD Mail fork Heirloom mailx, formerly known as nail.

S-nail has no prerequisites but a normal Unix environment (make(1), an ISO C89 C compiler etc.) and it is also possible to work directly with a repository checkout. The repository itself consists of the [master] branch, which is immutable and only gains important bugfixes and otherwise mature changesets only, mostly in preparation of a new release; it is based upon a full clone of the [heirloom] cvs(1) repository. The [timeline] branch is an ongoing effort to give a complete picture of the Unix mail / BSD Mail program, with one commit per release. [next] is based upon [master] and includes upcoming changes which are supposed to be quite stable, so that daring users may give it a try. The [crawl] branch finally is plain chaos and developer only. E.g., to clone only the [master] and [next] branches of S-nail:

 $ mkdir s-nail.git
 $ cd s-nail.git
 $ git init
 $ git remote add origin -t master -t next git://
 $ git fetch -v
 $ git checkout master
m d o c m x — v2, 2015-05-12
Code: part of S-roff as below
groff(1) enhancement request:

The mdoc(7) manual semantic markup language doesn't support any kind of anchoring: whereas you can exactly state what x is whenever you refer to it – variable, function etc. –, you have no option to define the exact place where x is itself defined, or wether it is at all defined in a given manual page.

Also, whereas mdoc(7) does support differentiation in between anchors and references for headlines (.Sh / .Ss and .Sx, respectively), referencing a headline is only of notational interest, the reference is in no way “active”, never.

mdocmx(7) extends mdoc(7) by adding all the missing functionality and more, including (dependent on the output device) referencable external manuals, with a single new multiplexer command: .Mx (read the manual page).

And because non-multipass troff(1) implementations are not capable to generate forward references to anchors not yet defined there is a preprocessor necessary to circumvent this problem: mdocmx(1), implemented in portable sh(1) and awk(1). (A nice property of mdocmx(7) is that it “knows” wether a document has been preprocessed or not, therefore making it possible to distribute preprocessed manuals and still being backward- and forward-compatible.)

For security reasons all parts of the usual manual pipeline require the environment variable MDOCMX_ENABLE=1 to be set to a non-empty value (it must be non-empty because of groff(1) deficiencies). E.g., on a system where groff(1) as well as less(1) have been patched to support mdocmx(7), the following shell functions could be used to read mdoc(7) manuals directly (use ^A, i.e., control-A in less(1) to jump to anchors or read external manuals):

 mdoc() { (
   : ${MDOCMXFLAGS:=-dmx-toc-force=tree} "${1}" |
   groff -Tutf8 -mdoc ${MDOCMXFLAGS} |
   LESS= less --RAW-CONTROL-CHARS --ignore-case --no-init
 ) }

Question: How does it work?

For output devices like HTML or PDF mdocmx(7) will use the corresponding and well-known support troff(1) macro packages to generate anchor and reference information.

For the TTY output device (grotty(1) in groff(1)) it will inject context information (anchor?, link? and their name) directly via a new troff(1) \X'' command; The TTY output device therefore needs to be extended in order to understand this new command. The context information will then be written to the PAGER as backspace-escaped text — since on Unix backspace-escaped sequences are used to create manual page style formatting information (bold text, text underlining, etc.) since decades, most tested pagers are able to silently ignore those sequences. (Despite the fact that “backspace” is the control character to delete the character before it.)

If the Pager has been extended to actually interpret the backspace-escaped sequences as anchor and reference context information, it can use its normal search facilities to search for anchors and could also offer the possibility to startup man(1) to follow external manual page references.

Note that due to the way that mdoc(7) is implemented section and subsection headers may not contain macro recursions. Please see the referenced groff(1) enhancement request for more on this; the next mdocmx(7) iteration will generate warnings for such use cases (at least in the preprocess); just as stated in the enhancement request mdoc(7) itself has to be rewritten in order to overcome this restriction.

S - r o f f — in the process of being forked as of 2015-09-05
Code: {*master*,[next],[crawl]} git:// (browse) (browse)

I have forked groff(1) from the last commit that is still GPL2 licensed (1.19.2-574-gecbf4f1). Almost all further changes up to and including groff(1) 1.22.3 are included, however. S-roff, that is the name, is a stripped-down version of groff(1), missing any related facility, the output devices grolj4 and grolbp as well as most of contributed packages. My general idea is to have a forward looking point–of–view.

The plan is to make it UTF-8 clean all through the toolchain, to make it more user friendly, e.g., by adding automatic detection of required preprocessors. Nonetheless keeping backward compatibility and accessibility of individual subcomponents, like nroff, troff and all the preprocessors. S-roff logo -- magnolia night I would like to see the manual all in manual pages (mdoc). I really would like to have builtin support for TTF/OTF fonts. All of this is mid– to long–term.

In the short run i'll have to completely rework the build system and adapt It'll still take a while this rather big codebase to not need any autotool, but only the shell and make, just the same easy way as is used for S-nail, then perform a lot of rather invisible but desired code overhauling, like implementing consistent argument parsing etc. I hope this step can be finished in spring 2016, followed by an early initial release.

S - W e b 4 2 — v0.8.6, 2015-06-24
Code: {*master*,next} git:// (browse) (browse)

One more option to manage your website. A camel approach to website building. vim(1) and it actually generated what you are looking at. ... read the complete documentation.

TODO: i should extend it with a markdown-compatible syntax block, and restrict compression to gzip(1), in which case it would be perl(1)-only on the generator side.

S - S y m O b j — v0.8.0, 2012-12-17
Downloads: .tar.gz (via CPAN)
Code: {*master*} git:// (browse) (browse)

Throw an eye on my Symbol table and Object perl(1) module, it offers a somewhat easy symbol table and object creation/management. ... read the complete PODocumentation.

I'm using perl(1) since 1997 in many projects and for a lot of purposes. In all that time i haven't found a single bug! Hoooray and thanks, perl(1) porters!

S - C T e x t — early alpha (but based upon own UCD) as of 2013-07-06 (and 2015-09-05)
Code: {*master*,[next],[crawl]} git:// (browse) (browse)

Unix / POSIX continues to be byte-based, slowly drifting towards the byte-based UTF-8 multibyte character set. But the C and POSIX standards don't offer any option to work on multibyte character data except by first converting it into wide character strings. S-CText logo -- Spring This means that expensive and most often completely redundant round-trip conversions are necessary, even for the most simple tasks.

Furthermore, Unicode defines multi-codepoint characters — characters that are composed out of multiple codepoints, so-called graphems (see Unicode Text Segmentation), many of which don't have a precomposed single-codepoint representation. Neither ISO C nor POSIX define an interface that can deal with these sequences-of-codepoints that form a single character.

S-CText provides a programming interface that addresses these problems. It offers sets of string functions which deal with (possible) multibyte strings in the encoding of the / a LC_CTYPE locale (sct_ prefix function family, optional) as well as with UTF-8 encoded strings (sut_ prefix function family). In addition the internal representation of the Unicode character database can be accessed (sud_ prefix function family). And there is an optional character set conversion interface (scc_ prefix function family).

Expect a first usable version ... before 2018. (It has to rock!)

S - T o o l b o x
Code: {*master*} git:// (browse) (browse)

The README as of 2015-11-26:

A repo of some small tools

uushar (manual: uushar.1):
   Writes a sh(1) shell script to the standard output which will
   recreate the file hierarchy specified by the command line operands.
   Directories will be recreated and must be specified before the files
   they contain.  Compared to the well-known shar(1) program, uushar(1)
   adds optional compression and an uuencode(1)/uudecode(1) binary
   safe layer::

      $ uushar [-(Z|z|J|j)] file ...

   It is easy to insert trojan horses into uushar files.  It is thus
   strongly recommended that all shell archive files be examined before
   running them through sh(1).  Archives produced using this
   implementation of uushar may be easily examined with the command::

      $ grep '^[^X#]' shar.file

   Note that the uudecode(1) program must be capable of decoding to
   /dev/stdout, i.e., it must be POSIX compliant (Issue 6 iirc).

   2014-02-27: rewritten: short options, check and forbid filenames with
   embedded double quotes, added uushar.1 manual.

   Creates a shell archive similar to uushar as above, except that the
   archive (1) consists only of executable file members, and (2) will
   be itself executable.  Run it (the generated shell archive script)
   to invoke any of the programs contained therein.  On the first run,
   the wrapper will create a hidden directory in your $TMPDIR to unpack
   the archive members therein.  Run without arguments to show its
   contents and the creation date.
   It needs no manual, simply execute it and it will guide you through
   the archive creation process.

   Note that the uudecode(1) program must be capable of decoding to
   /dev/stdout, i.e., it must be POSIX compliant (Issue 6 iirc).

   2014-03-05: rewritten: check and forbid filenames with embedded
   double quotes, ensure the given default executable, if any, is part
   of the generated archive, reexecute after archive expansion, truly
   disallow empty archive name.
   A _real_ periodic for NetBSD, FreeBSD, DragonFly BSD and OpenBSD that
   ensures that the daily/weekly/monthly maintenance stuff is executed,
   even if your laptop is running only one hour a day.  Invoke this once
   per hour in the roots crontab and disable the usual periodic stuff of
   your system instead.  Note it doesn't deal with timezone and DST
   (daylight saving time) switches, but instead only checks the day of
   the year that is reported by date(1).  E.g., on my FreeBSD 10.0 box::

      # do daily/weekly/monthly maintenance
      15  * * * * root /usr/bin/nice -n 15 /usr/libexec/
      #1  3 * * * root periodic daily
      #15 4 * * 6 root periodic weekly
      #30 5 1 * * root periodic monthly

   2015-08-24: newer periodic(8) sleep in between jobs, so update the DB
   before we start it!
   To ease maintaining of topic branches i use a linear history, so
   that i can use ``$ git rebase -i`` for fixups and squashes, without
   having to worry about recreation of topic branches.  Instead i tag
   commit messages, and use this script to create the topics once i'm
   finished.  Read the --help output for more, and make use of verbosity
   when you use it first.

   2013-09-13: newer git(1) reverse ``rev-parse`` output: adjusted.
   Offers some automatized operations on revision control repositories,
   like updating, (fast-forward) merging, garbage-collecting.

   2016-XX-XX: complete rewrite that completely replaces the old
   ``arena-manager`` and requires only a POSIX environment (no perl(1)).
   The new ``setup`` mode must be used to initialize a managed
   (git(1)-based) repository; this results in a much more powerful
   functionality and automatization.  Please look into the script header
   for more.  The old script has been removed (see ``git(1)`` history).

Even less of widespread usability:
   Download NNTP articles from, incrementally, and store them in
   a local MBOX.
   Read and adjustthe script header for documentation and configuration.
   Synchronize a bunch of local repos with their remotes.
   Brute simple version that can only push.
   Needs awk(1).
   git(1)'s diff --check does unfortunately not offer a mode which
   detects NBSP (non-breaking space, U+00A0) characters.
   This is a real problem for keyboards (layouts) which produce combined
   sequences, i.e., the german keyboard produces NBSP for ALT+SPC, but
   it also requires ALT to get any of []|{} (ALT+[5-9]).  So NBSPs
   happen.  This script can be used as (a template for a custom)
   pre-commit hook, now that i've looked at it again (2013-05-13).

   It also handles non-space indentations, tabulators and mixed
   (space-before-tabs) indents, dependent on the environment variables
   NSPACEINDENT, TABINDENT and MIXINDENT, respectively.  Setting
   MKFILEDIG will treat [Mm]akefile and *.mk files specially.
   E.g., this is one of my pre-commit hooks::

      #!/bin/sh -
      export MIXINDENT
      exec "$HOME/usr/bin/" "${@}"

   2014-04-14: added MKFILEDIG.
   Check indentation, trailing whitespace and non-breaking spaces (see above) for all files given on the command line.
   Sufficient for a needed quick-n-dirty check of some random file;
   now that i've looked at it again (2013-05-13).

   2014-04-14: added MKFILEDIG.

   Encode to/decode from Base64 content-transfer-encoding, as specified
   in RFC 2045, section 6.8.  Rather failure-tolerant...

   2014-07-29: fixed SIGSEGV which could happen because of ignoring
   errors during base64 decoding; this tool needs a rewrite to actually
   do what it is supposed to, i.e., provide a failure-tolerant decoding
   facility: it should print something like `xy bytes remain unparsed'
   after an error occurred instead of simply stating that there was
   invalid base64.

   Simple GNU Libidn based thing that encodes and decodes IDNs.
   Rather identical to the idn utility that comes with that one, but
   i needed to get comfortable with Libidn and so i've wrote it.

   Show keyboard scancodes for FreeBSD, OpenBSD, NetBSD and Linux.
   This is not yet finished but for Linux.  But it should help to get
   you going.  (Unfortunately the BSDs are not so documented and i
   need to dig into the kernel sources to implement the missing pieces.)

   2015-10-20: make it compile via c99(1) under Linux.
   My private backup script.  Simple (uses tar(1)), but does some stuff
   which i need for my work flow (backups data which is in local repo
   clones but not yet pushed to their "real" counterparts).

   2015-09-02: no longer following symbolic links
   2015-10-09: add -b/--basename option, add $COMPRESSOR variable, start
               via shell and $PERL5OPT clear to avoid multibyte problems

   Single setuid root BSD program to invoke /sbin/shutdown(8), either
   to reboot or to halt the system.
   You've been warned.

   FreeBSD 5.3 x86 prog to open/close /dev/cdrom tray.  May work on
   newer ones, but haven't tried it in a while.  It had 416 object
   file bytes when everything was placed in ``.text`` (no .EH frames
   back then).,
   Simple scripts to ease maintaining of OpenBSD plus.html entries.

# s-it-mode
S - M u s i c b o x
Code: {*master*} git:// (browse) (browse)

A complete rewrite of our internal rather shitty jukebox, that managed to encapsulate HTML interface, CGI server side (TCP server using MPG321 remote control for playback), ripping (split on CDROM size boundaries with best fit, and that was the best feature of it) in a single script. It only supported MP3 (lame(1), that is). It ran (and still runs) on Linux and FreeBSD. It's database (using a lout-style trigger followed by a content line) did not store much information about recordings:


 # [PLAYTIME(0000:00.00 string..)] TAB [COMMENT] NL

 # [PLAYTIME(0000:00.00 string)] TAB [GENRE] TAB [COMMENT] NL



Yes, it's really, really odd, but remember it was indeed FILE based and splitted on CDROM size boundaries, so that each CDROM's database had to contain all the information of all files on it. The final database could be gained by cat(1)ing all those together, redundant entries would be removed upon load ...

So it becomes obvious why i actually wanted to have a new musicbox with a more sophisticated database, one that is capable to deal with the odds of classical music. Well, i did and re-ripped all my CDs with the script, which is yet the only ready component of S-Musicbox. It's CDInfo:: class, the one that is responsible for actual CDROM detection and ripping, supports only Mac OS X at the moment, however. It is pretty easy to extend, though. And the is a rather mature program, that i like.

Stale or stalled stuff

S - P o s t m a n — v0.4.0-beta7, 2013-01-16
Code: {*master*,next} git:// (browse)

This script was in daily use for SSL-secured POP3 download from mid 2011 until the begin of june 2013, but it is incomplete and stale. Please see its README first! ... read the complete documentation.

An unsophisticated end-user mail fetcher, processor and sender; it's easy configurable and offers a pretty good rule matching engine. This application has been written in Python.

In january 2011 this very project caused me to try Payprout 3000. Since then i stumbled over the discovery of dozens of bugs (in the standard library) and even got involved a little bit on their bug tracker. Payprout? Yes! (The details perhaps had had a negative influence on the emotional well-being of at least one of the mentioned characters.)

Well, it turns out that i curse the day when i decided that it try Python for this. The basic idea was that a Python installation always ships with the standard library, and that was supposed to already include most things ever needed. But if this projects takes that much longer there will not remain that many stdlib modules being used in the end.

S - X S y s t e m
Code: {*master*,next} git:// (browse)

The basic library of my S programming framework. It encapsulates almost all CPU and OS dependencies under its generic interface.

This will in fact be my free clone of my (our) internal non-free C++ library; different to that with full Unicode support, however. It will also be possible to use that as a normal library, whereas the C++ version does not link against any libraries and ships it's own threading module. That however implies that there will be speed impacts, because then the extremely time critical s_thr_self() function will depend on some system thread-library provided thing etc. The situation is somewhat better on platforms with ELF TLS, but still worse than a fast manual driven thread-local storage (though completely automatic, on the other hand). And then of course: signals. As if that wouldn't be hard enough in a monolithic situation, only interacting with the OS. Horror in the pipeline!

It's pretty large so that i need at least until 2020 to port it:

 $ du -s cvsroot/cpp-sys
 5460 cvsroot/cpp-sys
S - X C o n s o l e
Code: {*master*,next} git:// (browse)

S library for (basic) console screen management. This is my free clone of my (our) internal non-free C++ library, which is conversely called Terminal. Different to that with full Unicode support.

It is already finished in theory, but since it requires the S-XSystem library it will take some years until it is really usable. At that time it will be clear wether it runs in practice, too. :-). (P.S.: this simple minded port would not really work with real Unicode characters but only with single Unicode codepoints. The README file talks a bit about that.)

It'll be extended once a normal development cycle is possible: there will be (non-overlapping) windows, a single-line-only mode will perhaps be implemented. The drawing mechanism can be improved in respect to performance. And a termcap interface would be a win, too. (I.e., dynamically extending the static builtin database of supported terminals.)

T R 9 n support
OpenBSD/[de] / OpenSSH/[de]

In november 2011 i became the maintainer of the german translations of the OpenBSD website, and after a lot of work i got synchronized with the tree on 2012-01-22. Because of character deficiencies of Mr. de Raadt the OpenBSD/[de] maintainership was discontinued on 2014-01-24. Thus the following paragraphs are partially outdated.

I hope i can realize and afford it to provide a uniform style through all pages until the release of OpenBSD 5.7. See DEification for a wordlist of changes that have been applied to [de], or will be applied to it in the future.

Unfortunately i had to drop my anti-german-language-reform attitude after one and a half decade for this maintainership (and only for it), so this reveals that the first paragraph of my homepage is actually a lie. But hey, do you know what: this standard committee currently is and furthermore always was deeply broken! You possibly don't want to see a photo of the current chairman. But anyway, it's official german, now. (Sob.)

I also did my part to maintain the daily changelog from the beginning of november 2011 until the end of june 2012; being part of a team that was formed out of the australian-based male Alpha Centaurian Brett Mahar (brett@), the swedish sweetheart Janne Johannson (jj@) and the flying german unionist Ingo Schwarze (schwarze@; all [openbsd DOT org]).

So i finally contribute(d) a little bit to the OpenBSD and OpenSSH projects, just as i'd promised in 2001 (that could also have been 2002).

From you to me

I love C, perl(1) and dig plain old sh(1) and awk(1) more and more. I'd love C++ (again) if it would be a plain C with classes; maybe not even automatic ctors and dtors. (If i really would go for learning a new language i think i would most likely have a go with nimrod.) I've discovered mksh(1) and use it exclusively on all my real-work boxes for quite a while now. Since about year 2000 i'm a fan (though not a sophisticated user) of vim(1), it gave up in an endless loop once.

I'm thankful for being able to use git(1) for version control today, after a lot of distress with other VCSs. And i'm also thankful for being able to use virtual machines via QEMU and VirtualBox: it is so handy to have a bunch of operating systems at hand for immediate testing, just as necessary! I'm using Dropbear SSH for most SSH work, because it is very small and is sufficient for most uses; it replaced all my remaining rlogin(1) / rsh(1) use cases, and i am still saving space in comparison.

When on X(1) i'm running the fantastic ahwm(1) window manager. My terminal emulator is rxvt-unicode, but / and mostly i stay within the screen(1) terminal multiplexer. For browsing there is lynx(1), if text based browsing is possible, and Chrome otherwise; i've switched my graphical browser in 2014 after way more than a decade of using Opera, since that now updates behind the back and shows some other oddities, and its initial replacement Firefox is nothing i really can stay with, thus. I ended up using solely groff(1) for document preparation, viewing PDFs with mupdf(1); i do use from the same source.

I've used FreeBSD for about ten years, and again since 2015. I love NetBSD, which offers a lot of nice stuff for people that are willing to dig — just take netpgp(1). I'm also using Crux-Linux and archlinux regulary. Of course i have VMs with some other systems, like DragonFly. While looking around for Unicode aware software i have been pointed to the Plan9 from Bell Labs operating system and its 9atom and 9front extenders, which appear to follow a "real UNIX — everything is a file" design! I'm fascinated! Unfortunately i'm not a billionaire …

Copyright (c) 1997 - 2015, Steffen (Daode) Nurpmeso <>
@(#)code.html-w42 1.212 2015-09-23T12:49:00+0000