1 **Paraslash user manual**
3 This document describes how to install, configure and use the paraslash
4 network audio streaming system. Most chapters start with a chapter
5 overview and conclude with an example section. We try to focus on
6 general concepts and on the interaction of the various pieces of the
7 paraslash package. Hence this user manual is not meant as a replacement
8 for the manual pages that describe all command line options of each
15 In this chapter we give an [overview](#Overview) of the interactions of
16 the two main programs contained in the paraslash package, followed by
17 [brief descriptions](#The.paraslash.executables) of all executables.
22 The core functionality of the para suite is provided by two main
23 executables, para_server and para_audiod. The former maintains a
24 database of audio files and streams these files to para_audiod which
25 receives and plays the stream.
27 In a typical setting, both para_server and para_audiod act as
28 background daemons whose functionality is controlled by client
29 programs: the para_audioc client controls para_audiod over a local
30 socket while the para_client program connects to para_server over a
31 local or remote networking connection.
33 Typically, these two daemons run on different hosts but a local setup
36 A simplified picture of a typical setup is as follows
38 server_host client_host
39 ~~~~~~~~~~~ ~~~~~~~~~~~
41 +-----------+ audio stream +-----------+
42 |para_server| -----------------------------> |para_audiod|
43 +-----------+ +-----------+
54 | connect +-----------+
55 +-------------------------------------- |para_client|
57 The paraslash executables
58 -------------------------
62 para_server streams binary audio data (MP3, ...) over local and/or
63 remote networks. It listens on a TCP port and accepts commands such
64 as play, stop, pause, next from authenticated clients. There are
65 many more commands though, see the man page of para_server for a
66 description of all commands.
68 It supports three built-in network streaming protocols
69 (senders/receivers): HTTP, DCCP, or UDP. This is explained in more
70 detail in the section on [networking](#Networking).
72 The built-in audio file selector of paraslash is used to manage your
73 audio files. It maintains statistics on the usage of all available
74 audio files such as last-played time, and the number of times each
77 Additional information may be added to the database to allow
78 fine-grained selection based on various properties of the audio file,
79 including information found in (ID3) tags. However, old-fashioned
80 playlists are also supported.
82 It is also possible to store images (album covers) and lyrics in the
83 database and associate these to the corresponding audio files.
85 The section on the [audio file selector](#The.audio.file.selector)
91 The client program to connect to para_server. paraslash commands
92 are sent to para_server and the response is dumped to STDOUT. This
93 can be used by any scripting language to produce user interfaces with
94 little programming effort.
96 All connections between para_server and para_client are encrypted
97 with a symmetric session key. For each user of paraslash you must
98 create a public/secret RSA key pair for authentication.
100 If para_client is started without non-option arguments, an interactive
101 session (shell) is started. Command history and command completion are
102 supported through libreadline.
106 The local daemon that collects information from para_server.
108 It runs on the client side and connects to para_server. As soon as
109 para_server announces the availability of an audio stream, para_audiod
110 starts an appropriate receiver, any number of filters and a paraslash
111 writer to play the stream.
113 Moreover, para_audiod listens on a local socket and sends status
114 information about para_server and para_audiod to local clients on
115 request. Access via this local socket may be restricted by using Unix
116 socket credentials, if available.
121 The client program which talks to para_audiod. Used to control
122 para_audiod, to receive status info, or to grab the stream at any
123 point of the decoding process. Like para_client, para_audioc supports
124 interactive sessions on systems with libreadline.
128 A command line HTTP/DCCP/UDP stream grabber. The http mode is
129 compatible with arbitrary HTTP streaming sources (e.g. icecast).
130 In addition to the three network streaming modes, para_recv can also
131 operate in local (afh) mode. In this mode it writes the content of
132 an audio file on the local file system in complete chunks to stdout,
133 optionally 'just in time'. This allows to cut an audio file without
134 first decoding it, and it enables third-party software which is unaware
135 of the particular audio format to send complete frames in real time.
139 A filter program that reads from STDIN and writes to STDOUT.
140 Like para_recv, this is an atomic building block which can be used to
141 assemble higher-level audio receiving facilities. It combines several
142 different functionalities in one tool: decoders for multiple audio
143 formats and a number of processing filters, among these a normalizer
148 A small stand-alone program that prints tech info about the given
149 audio file to STDOUT. It can be instructed to print a "chunk table",
150 an array of offsets within the audio file.
154 A modular audio stream writer. It supports a simple file writer
155 output plug-in and optional WAV/raw players for ALSA (Linux) and OSS.
156 para_write can also be used as a stand-alone WAV or raw audio player.
160 A command line audio player.
164 Curses-based gui that presents status information obtained in a curses
165 window. Appearance can be customized via themes. para_gui provides
166 key-bindings for the most common server commands and new key-bindings
171 An alarm clock and volume-fader for OSS and ALSA.
177 This chapter lists the [necessary software](#Requirements)
178 that must be installed to compile the paraslash package, describes
179 how to [compile and install](#Installation) the paraslash
180 source code and the steps that have to be performed in order to
181 [set up](#Configuration) a typical server and client.
185 ### For the impatient ###
187 git clone git://git.tuebingen.mpg.de/osl
188 cd osl && make && sudo make install && sudo ldconfig
189 sudo apt-get install autoconf libssl-dev m4 \
190 libmad0-dev libid3tag0-dev libasound2-dev libvorbis-dev \
191 libfaad-dev libspeex-dev libFLAC-dev libsamplerate-dev realpath \
192 libasound2-dev libao-dev libreadline-dev libncurses-dev \
195 ### Detailed description ###
197 In any case you will need
199 - [libosl](http://people.tuebingen.mpg.de/maan/osl/). The _object
200 storage layer_ library is used by para_server. To clone the source
201 code repository, execute
203 git clone git://git.tuebingen.mpg.de/osl
205 - [lopsub](http://people.tuebingen.mpg.de/maan/lopsub/). The long
206 option parser for subcommands generates the command line and config
207 file parsers for all paraslash executables. Clone the source code
210 git clone git://git.tuebingen.mpg.de/lopsub
212 - [gcc](ftp://ftp.gnu.org/pub/gnu/gcc) or
213 [clang](http://clang.llvm.org). All gcc versions >= 4.2 are currently
214 supported. Clang version 1.1 or newer should work as well.
216 - [gnu make](ftp://ftp.gnu.org/pub/gnu/make) is also shipped with the
217 disto. On BSD systems the gnu make executable is often called gmake.
219 - [bash](ftp://ftp.gnu.org/pub/gnu/bash). Some scripts which run
220 during compilation require the _Bourne again shell_. It is most
221 likely already installed.
223 - [m4](ftp://ftp.gnu.org/pub/gnu/m4/). Some source files are generated
224 from templates by the m4 macro processor.
228 - [openssl](http://www.openssl.org/) or
229 [libgcrypt](ftp://ftp.gnupg.org/gcrypt/libgcrypt/). At least one
230 of these two libraries is needed as the backend for cryptographic
231 routines on both the server and the client side. Both openssl and
232 libgcrypt are usually shipped with the distro, but you might have
233 to install the development package (`libssl-dev` or `libgcrypt-dev`
234 on debian systems) as well.
236 - [libmad](http://www.underbit.com/products/mad/). To compile in MP3
237 support for paraslash, the development package must be installed. It
238 is called `libmad0-dev` on debian-based systems. Note that libmad is
239 not necessary on the server side, i.e., for sending MP3 files.
241 - [libid3tag](http://www.underbit.com/products/mad/). For version-2
242 ID3 tag support, you willl need the libid3tag development package
243 `libid3tag0-dev`. Without libid3tag, only version-1 tags are
244 recognized. The mp3 tagger also needs this library for modifying
245 (id3v1 and id3v2) tags.
247 - [ogg vorbis](http://www.xiph.org/downloads/). For ogg vorbis streams
248 you need libogg, libvorbis, libvorbisfile. The corresponding Debian
249 packages are called `libogg-dev` and `libvorbis-dev`.
251 - [libfaad and mp4ff](http://www.audiocoding.com/). For aac files
252 (m4a) you need libfaad and libmp4ff (package: `libfaad-dev`). Note
253 that for some distributions, e.g. Ubuntu, mp4ff is not part of the
254 libfaad package. Install the faad library from sources (available
255 through the above link) to get the mp4ff library and header files.
257 - [speex](http://www.speex.org/). In order to stream or decode speex
258 files, libspeex (`libspeex-dev`) is required.
260 - [flac](http://flac.sourceforge.net/). To stream or decode files
261 encoded with the _Free Lossless Audio Codec_, libFLAC (`libFLAC-dev`)
264 - [libsamplerate](http://www.mega-nerd.com/SRC/index.html). The
265 resample filter will only be compiled if this library is
266 installed. Debian package: `libsamplerate-dev`.
268 - [alsa-lib](ftp://ftp.alsa-project.org/pub/lib/). On Linux, you will
269 need to have the ALSA development package `libasound2-dev` installed.
271 - [libao](http://downloads.xiph.org/releases/ao/). Needed to build
272 the ao writer (ESD, PulseAudio,...). Debian package: `libao-dev`.
274 - [curses](ftp://ftp.gnu.org/pub/gnu/ncurses). Needed for
275 para_gui. Debian package: `libncurses-dev`.
278 Readline](http://cnswww.cns.cwru.edu/php/chet/readline/rltop.html). If
279 this library (`libreadline-dev`) is installed, para_client, para_audioc
280 and para_play support interactive sessions.
284 To build the sources from a tarball, execute
288 To build from git or a gitweb snapshot, run this command instead:
292 There should be no errors but probably some warnings about missing
293 packages which usually implies that not all audio formats will be
294 supported. If headers or libs are installed at unusual locations you
295 might need to tell the configure script where to find them. Try
299 to see a list of options. If the paraslash package was compiled
300 successfully, execute (optionally)
304 to run the paraslash test suite. If all tests pass, execute as root
308 to install executables under /usr/local/bin and the man pages under
314 ### Create a paraslash user ###
316 In order to control para_server at runtime you must create a paraslash
317 user. As authentication is based on the RSA crypto system you'll have
318 to create an RSA key pair. If you already have a user and an RSA key
319 pair, you may skip this step.
321 In this section we'll assume a typical setup: You would like to run
322 para_server on some host called server_host as user foo, and you want
323 to connect to para_server from another machine called client_host as
326 As foo@server_host, create ~/.paraslash/server.users by typing the
330 target=~/.paraslash/server.users
331 key=~/.paraslash/id_rsa.pub.$user
332 perms=AFS_READ,AFS_WRITE,VSS_READ,VSS_WRITE
333 mkdir -p ~/.paraslash
334 echo "user $user $key $perms" >> $target
336 Next, change to the "bar" account on client_host and generate the
337 key pair with the commands
339 ssh-keygen -q -t rsa -b 2048 -N '' -f $key
341 This generates the two files id_rsa and id_rsa.pub in ~/.ssh. Note
342 that para_server won't accept keys shorter than 2048 bits. Moreover,
343 para_client rejects private keys which are world-readable.
345 para_server only needs to know the public key of the key pair just
346 created. Copy this public key to server_host:
348 src=~/.ssh/id_rsa.pub
349 dest=.paraslash/id_rsa.pub.$LOGNAME
350 scp $src foo@server_host:$dest
352 Finally, tell para_client to connect to server_host:
354 conf=~/.paraslash/client.conf
355 echo 'hostname server_host' > $conf
358 ### Start para_server ###
360 For this first try, we'll use the info loglevel to make the output
361 of para_server more verbose.
365 Now you can use para_client to connect to the server and issue
366 commands. Open a new shell as bar@client_host and try
371 to retrieve the list of available commands and some server info.
372 Don't proceed if this doesn't work.
374 ### Create and populate the database ###
376 An empty database is created with
380 This initializes a couple of empty tables under
381 ~/.paraslash/afs_database-0.4. You normally don't need to look at these
382 tables, but it's good to know that you can start from scratch with
384 rm -rf ~/.paraslash/afs_database-0.4
386 in case something went wrong.
388 Next, you need to add some audio files to that database so that
389 para_server knows about them. Choose an absolute path to a directory
390 containing some audio files and add them to the audio file table:
392 para_client add /my/mp3/dir
394 This might take a while, so it is a good idea to start with a directory
395 containing not too many files. Note that the table only contains data
396 about the audio files found, not the files themselves.
398 You may print the list of all known audio files with
402 ### Configure para_audiod ###
404 We will have to tell para_audiod that it should receive the audio
405 stream from server_host via http:
407 para_audiod -l info -r '.:http -i server_host'
409 You should now be able to listen to the audio stream once para_server
410 starts streaming. To activate streaming, execute
414 Since no playlist has been specified yet, the "dummy" mode which
415 selects all known audio files is activated automatically. See the
416 section on the [audio file selector](#The.audio.file.selector) for how
417 to use playlists and moods to specify which files should be streamed
423 If you receive a socket related error on server or audiod startup,
424 make sure you have write permissions to the /var/paraslash directory:
426 sudo chown $LOGNAME /var/paraslash
428 Alternatively, use the --afs-socket (para_server) or --socket
429 (para_audiod) option to specify a different socket pathname.
431 To identify streaming problems try to receive, decode and play the
432 stream manually using para_recv, para_filter and para_write as follows.
433 For simplicity we assume that you're running Linux/ALSA and that only
434 MP3 files have been added to the database.
436 para_recv -r 'http -i server_host' > file.mp3
437 # (interrupt with CTRL+C after a few seconds)
438 ls -l file.mp3 # should not be empty
439 para_filter -f mp3dec -f wav < file.mp3 > file.wav
440 ls -l file.wav # should be much bigger than file.mp3
441 para_write -w alsa < file.wav
443 Double check what is logged by para_server and use the --loglevel
444 option of para_recv, para_filter and para_write to increase verbosity.
450 para_server uses a challenge-response mechanism to authenticate
451 requests from incoming connections, similar to ssh's public key
452 authentication method. Authenticated connections are encrypted using
453 the AES stream cipher in integer counter mode.
455 In this chapter we briefly describe RSA and AES, and sketch the
456 [authentication handshake](#Client-server.authentication)
457 between para_client and para_server. User management is discussed
458 in the section on [the user_list file](#The.user_list.file).
459 These sections are all about communication between the client and the
460 server. Connecting para_audiod is a different matter and is described
461 in a [separate section](#Connecting.para_audiod).
466 A block cipher is a transformation which operates on fixed-length
467 blocks. For symmetric block ciphers the transformation is determined
468 by a single key for both encryption and decryption. For asymmetric
469 block ciphers, on the other hand, the key consists of two parts,
470 called the public key and the private key. A message can be encrypted
471 with either key and only the counterpart of that key can decrypt the
472 message. Asymmetric block ciphers can be used for both signing and
473 encrypting a message.
475 RSA is an asymmetric block cipher which is used in many applications,
476 including ssh and gpg. The RSA public key encryption and signatures
477 algorithms are defined in detail in RFC 2437. Paraslash relies on
478 RSA for authentication.
480 Stream ciphers XOR the input with a pseudo-random key stream to produce
481 the output. Decryption uses the same function calls as encryption.
482 Any block cipher can be turned into a stream cipher by generating the
483 pseudo-random key stream by encrypting successive values of a counter
486 AES, the advanced encryption standard, is a well-known symmetric block
487 cipher. Paraslash employs AES in counter mode as described above to
488 encrypt communications. Since a stream cipher key must not be used
489 twice, a random key is generated for every new connection.
491 Client-server authentication
492 ----------------------------
494 The authentication handshake between para_client and para_server goes
497 - para_client connects to para_server and sends an authentication
498 request for a user. It does so by connecting to TCP port 2990 of the
499 server host. This port is called the para_server _control port_.
501 - para_server accepts the connection and forks a child process which
502 handles the incoming request. The parent process keeps listening on the
503 control port while the child process (also called para_server below)
504 continues as follows.
506 - para_server loads the RSA public key of that user, fills a
507 fixed-length buffer with random bytes, encrypts that buffer using the
508 public key and sends the encrypted buffer to the client. The first
509 part of the buffer is the challenge which is used for authentication
510 while the second part is the session key.
512 - para_client receives the encrypted buffer and decrypts it with the
513 user's private key, thereby obtaining the challenge buffer and the
514 session key. It sends the SHA1 hash value of the challenge back to
515 para_server and stores the session key for further use.
517 - para_server also computes the SHA1 hash of the challenge and compares
518 it against what was sent back by the client.
520 - If the two hashes do not match, the authentication has failed and
521 para_server closes the connection.
523 - Otherwise the user is considered authenticated and the client is
524 allowed to proceed by sending a command to be executed. From this
525 point on the communication is encrypted using the stream cipher with
526 the session key known to both peers.
528 paraslash relies on the quality of the pseudo-random bytes provided
529 by the crypto library (openssl or libgcrypt), on the security of
530 the implementation of the RSA and AES crypto routines and on the
531 infeasibility to invert the SHA1 function.
533 Neither para_server or para_client create RSA keys on their
534 own. This has to be done once for each user as sketched in
535 [Quick start](#Quick.start) and discussed in more detail
536 [below](#The.user_list.file).
541 At startup para_server reads the user list file which contains one
542 line per user. The default location of the user list file may be
543 changed with the --user-list option.
545 There should be at least one user in this file. Each user must have
546 an RSA key pair. The public part of the key is needed by para_server
547 while the private key is needed by para_client. Each line of the
548 user list file must be of the form
550 user <username> <key> <perms>
552 where _username_ is an arbitrary string (usually the user's login
553 name), _key_ is the full path to that user's public RSA key, and
554 _perms_ is a comma-separated list of zero or more of the following
557 +---------------------------------------------------------+
558 | AFS_READ | read the contents of the databases |
559 +-----------+---------------------------------------------+
560 | AFS_WRITE | change database contents |
561 +-----------+---------------------------------------------+
562 | VSS_READ | obtain information about the current stream |
563 +-----------+---------------------------------------------+
564 | VSS_WRITE | change the current stream |
565 +---------------------------------------------------------+
567 The permission bits specify which commands the user is allowed to
568 execute. The output of
572 contains in the third column the permissions needed to execute the
575 It is possible to make para_server reread the user_list file by
576 executing the paraslash "hup" command or by sending SIGHUP to the
579 Connecting para_audiod
580 ----------------------
582 para_audiod listens on a Unix domain socket. Those sockets are
583 for local communication only, so only local users can connect to
584 para_audiod. The default is to let any user connect but this can be
585 restricted on platforms that support UNIX socket credentials which
586 allow para_audiod to obtain the Unix credentials of the connecting
589 Use para_audiod's --user-allow option to allow connections only for
590 a limited set of users.
592 =======================
593 The audio file selector
594 =======================
596 paraslash comes with a sophisticated audio file selector (AFS),
597 whose main task is to determine which file to stream next, based on
598 information on the audio files stored in a database. It communicates
599 also with para_client whenever an AFS command is executed, for example
600 to answer a database query.
602 Besides the traditional playlists, AFS supports audio file selection
603 based on _moods_ which act as a filter that limits the set of all
604 known audio files to those which satisfy certain criteria. It also
605 maintains tables containing images (e.g. album cover art) and lyrics
606 that can be associated with one or more audio files.
608 AFS employs [libosl](http://people.tuebingen.mpg.de/maan/osl/), the
609 object storage layer library, as the backend library for storing
610 information on audio files, playlists, etc. This library offers
611 functionality similar to a relational database, but is much more
612 lightweight than a full database backend.
614 In this chapter we sketch the setup of the [AFS
615 process](#The.AFS.process) during server startup and proceed with the
616 description of the [layout](#Database.layout) of the various database
617 tables. The section on [playlists and moods](#Playlists.and.moods)
618 explains these two audio file selection mechanisms in detail
619 and contains pratical examples. The way [file renames and content
620 changes](#File.renames.and.content.changes) are detected is discussed
621 briefly before the [Troubleshooting](#Troubleshooting) section
622 concludes the chapter.
627 On startup, para_server forks to create the AFS process which opens
628 the OSL database tables. The server process communicates with the
629 AFS process via pipes and shared memory. Usually, the AFS process
630 awakes only briefly whenever the current audio file changes. The AFS
631 process determines the next audio file, opens it, verifies it has
632 not been changed since it was added to the database and passes the
633 open file descriptor to the server process, along with audio file
634 meta-data such as file name, duration, audio format and so on. The
635 server process then starts to stream the audio file.
637 The AFS process also accepts connections from local clients via
638 a well-known socket. However, only child processes of para_server
639 may connect through this socket. All server commands that have the
640 AFS_READ or AFS_WRITE permission bits use this mechanism to query or
646 ### The audio file table ###
648 This is the most important and usually also the largest table of the
649 AFS database. It contains the information needed to stream each audio
650 file. In particular the following data is stored for each audio file.
652 - SHA1 hash value of the audio file contents. This is computed once
653 when the file is added to the database. Whenever AFS selects this
654 audio file for streaming the hash value is recomputed and checked
655 against the value stored in the database to detect content changes.
657 - The time when this audio file was last played.
659 - The number of times the file has been played so far.
661 - The attribute bitmask.
663 - The image id which describes the image associated with this audio
666 - The lyrics id which describes the lyrics associated with this
669 - The audio format id (MP3, OGG, ...).
671 - An amplification value that can be used by the amplification filter
672 to pre-amplify the decoded audio stream.
674 - The chunk table. It describes the location and the timing of the
675 building blocks of the audio file. This is used by para_server to
676 send chunks of the file at appropriate times.
678 - The duration of the audio file.
680 - Tag information contained in the audio file (ID3 tags, Vorbis
683 - The number of channels
685 - The encoding bitrate.
687 - The sampling frequency.
689 To add or refresh the data contained in the audio file table, the _add_
690 command is used. It takes the full path of either an audio file or a
691 directory. In the latter case, the directory is traversed recursively
692 and all files which are recognized as valid audio files are added to
695 ### The attribute table ###
697 The attribute table contains two columns, _name_ and _bitnum_. An
698 attribute is simply a name for a certain bit number in the attribute
699 bitmask of the audio file table.
701 Each of the 64 bits of the attribute bitmask can be set for each
702 audio file individually. Hence up to 64 different attributes may be
703 defined. For example, "pop", "rock", "blues", "jazz", "instrumental",
704 "german_lyrics", "speech", whatever. You are free to choose as
705 many attributes as you like and there are no naming restrictions
708 A new attribute "test" is created by
710 para_client addatt test
714 lists all available attributes. You can set the "test" attribute for
715 an audio file by executing
717 para_client setatt test+ /path/to/the/audio/file
719 Similarly, the "test" bit can be removed from an audio file with
721 para_client setatt test- /path/to/the/audio/file
723 Instead of a path you may use a shell wildcard pattern. The attribute
724 is applied to all audio files matching this pattern:
726 para_client setatt test+ '/test/directory/*'
730 para_client -- ls -l=v
732 gives you a verbose listing of your audio files also showing which
735 In case you wonder why the double-dash in the above command is needed:
736 It tells para_client to not interpret the options after the dashes. If
737 you find this annoying, just say
739 alias para='para_client --'
741 and be happy. In what follows we shall use this alias.
743 The "test" attribute can be dropped from the database with
752 for more information and a complete list of command line options to
757 The image, lyrics, moods and playlists tables are all blob tables.
758 Blob tables consist of three columns each: The identifier which is
759 a positive number that is auto-incremented, the name (an arbitrary
760 string) and the content (the blob).
762 All blob tables support the same set of actions: cat, ls, mv, rm
763 and add. Of course, _add_ is used for adding new blobs to the table
764 while the other actions have the same meaning as the corresponding
765 Unix commands. The paraslash commands to perform these actions are
766 constructed as the concatenation of the table name and the action. For
767 example addimg, catimg, lsimg, mvimg, rmimg are the commands that
768 manipulate or query the image table.
770 The add variant of these commands is special as these commands read
771 the blob contents from stdin. To add an image to the image table the
774 para addimg image_name < file.jpg
778 Note that the images and lyrics are not interpreted at all, and also
779 the playlist and the mood blobs are only investigated when the mood
780 or playlist is activated with the select command.
782 ### The score table ###
784 The score table describes those audio files which are admissible for
785 the current mood or playlist (see below). The table has two columns:
786 a pointer to a row of the audio file table and a score value.
788 Unlike all other tables of the database, the score table remains in
789 memory and is never stored on disk. It is initialized at startup and
790 recomputed when the select command loads a new mood or playlist.
792 When the audio file selector is asked to open the next audio file,
793 it picks the row with the highest score, opens the corresponding
794 file and passes the file descriptor to the virtual streaming system.
795 At this point the last_played and the num_played fields of the selected
796 file are updated and the score is recomputed.
801 Playlists and moods offer two different ways of specifying the set of
802 admissible files. A playlist in itself describes a set of admissible
803 files. A mood, in contrast, describes the set of admissible files in
804 terms of attributes and other type of information available in the
805 audio file table. As an example, a mood can define a filename pattern,
806 which is then matched against the names of audio files in the table.
810 Playlists are accommodated in the playlist table of the afs database,
811 using the aforementioned blob format for tables. A new playlist is
812 created with the addpl command by specifying the full (absolute)
813 paths of all desired audio files, separated by newlines. Example:
815 find /my/mp3/dir -name "*.mp3" | para addpl my_playlist
817 If _my_playlist_ already exists it is overwritten. To activate the
818 new playlist, execute
820 para select p/my_playlist
822 The audio file selector will assign scores to each entry of the list,
823 in descending order so that files will be selected in order. If a
824 file could not be opened for streaming, its entry is removed from
825 the score table (but not from the playlist).
829 A mood consists of a unique name and its *mood definition*, which is
830 a set of *mood lines* containing expressions in terms of attributes
831 and other data contained in the database.
833 At any time at most one mood can be *active* which means that
834 para_server is going to select only files from that subset of
837 So in order to create a mood definition one has to write a set of
838 mood lines. Mood lines come in three flavours: Accept lines, deny
839 lines and score lines.
841 The general syntax of the three types of mood lines is
844 accept [with score <score>] [if] [not] <mood_method> [options]
845 deny [with score <score>] [if] [not] <mood_method> [options]
846 score <score> [if] [not] <mood_method> [options]
849 Here <score> is either an integer or the string "random" which assigns
850 a random score to all matching files. The score value changes the
851 order in which admissible files are going to be selected, but is of
852 minor importance for this introduction.
854 So we concentrate on the first two forms, i.e. accept and deny
855 lines. As usual, everything in square brackets is optional, i.e.
856 accept/deny lines take the following form when ignoring scores:
858 accept [if] [not] <mood_method> [options]
860 and analogously for the deny case. The "if" keyword is only syntactic
861 sugar and has no function. The "not" keyword just inverts the result,
862 so the essence of a mood line is the mood method part and the options
863 following thereafter.
865 A *mood method* is realized as a function which takes an audio file
866 and computes a number from the data contained in the database.
867 If this number is non-negative, we say the file *matches* the mood
868 method. The file matches the full mood line if it either
870 - matches the mood method and the "not" keyword is not given,
872 - does not match the mood method, but the "not" keyword is given.
874 The set of admissible files for the whole mood is now defined as those
875 files which match at least one accept mood line, but no deny mood line.
876 More formally, an audio file F is admissible if and only if
878 (F ~ AL1 or F ~ AL2...) and not (F ~ DL1 or F ~ DN2 ...)
880 where AL1, AL2... are the accept lines, DL1, DL2... are the deny
881 lines and "~" means "matches".
883 The cases where no mood lines of accept/deny type are defined need
886 - Neither accept nor deny lines: This treats all files as
887 admissible (in fact, that is the definition of the dummy mood
888 which is activated automatically if no moods are available).
890 - Only accept lines: A file is admissible iff it matches at
891 least one accept line:
893 F ~ AL1 or F ~ AL2 or ...
895 - Only deny lines: A file is admissible iff it matches no
898 not (F ~ DL1 or F ~ DN2 ...)
902 ### List of mood_methods ###
906 Takes no arguments and matches an audio file if and only if no
909 is_set <attribute_name>
911 Takes the name of an attribute and matches iff that attribute is set.
913 path_matches <pattern>
915 Takes a filename pattern and matches iff the path of the audio file
918 artist_matches <pattern>
919 album_matches <pattern>
920 title_matches <pattern>
921 comment_matches <pattern>
923 Takes an extended regular expression and matches iff the text of the
924 corresponding tag of the audio file matches the pattern. If the tag
925 is not set, the empty string is matched against the pattern.
935 Takes a comparator ~ of the set {<, =, <=, >, >=, !=} and a number
936 <num>. Matches an audio file iff the condition <val> ~ <num> is
937 satisfied where val is the corresponding value of the audio file
938 (value of the year tag, bitrate in kbit/s, etc.).
940 The year tag is special as its value is undefined if the audio file
941 has no year tag or the content of the year tag is not a number. Such
942 audio files never match. Another difference is the special treatment
943 if the year tag is a two-digit number. In this case either 1900 or
944 2000 is added to the tag value, depending on whether the number is
945 greater than 2000 plus the current year.
950 To create a new mood called "my_mood", write its definition into
951 some temporary file, say "tmpfile", and add it to the mood table
954 para addmood my_mood < tmpfile
956 If the mood definition is really short, you may just pipe it to the
957 client instead of using temporary files. Like this:
959 echo "$MOOD_DEFINITION" | para addmood my_mood
961 There is no need to keep the temporary file since you can always use
962 the catmood command to get it back:
966 A mood can be activated by executing
968 para select m/my_mood
970 Once active, the list of admissible files is shown by the ls command
971 if the "-a" switch is given:
976 ### Example mood definition ###
978 Suppose you have defined attributes "punk" and "rock" and want to define
979 a mood containing only Punk-Rock songs. That is, an audio file should be
980 admissible if and only if both attributes are set. Since
984 is obviously the same as
986 not (not punk or not rock)
988 (de Morgan's rule), a mood definition that selects only Punk-Rock
991 deny if not is_set punk
992 deny if not is_set rock
996 File renames and content changes
997 --------------------------------
999 Since the audio file selector knows the SHA1 of each audio file that
1000 has been added to the afs database, it recognizes if the content of
1001 a file has changed, e.g. because an ID3 tag was added or modified.
1002 Also, if a file has been renamed or moved to a different location,
1003 afs will detect that an entry with the same hash value already exists
1004 in the audio file table.
1006 In both cases it is enough to just re-add the new file. In the
1007 first case (file content changed), the audio table is updated, while
1008 metadata such as the num_played and last_played fields, as well as
1009 the attributes, remain unchanged. In the other case, when the file
1010 is moved or renamed, only the path information is updated, all other
1011 data remains as before.
1013 It is possible to change the behaviour of the add command by using the
1014 "-l" (lazy add) or the "-f" (force add) option.
1019 Use the debug loglevel (-l debug) to show debugging info. All paraslash
1020 executables have a brief online help which is displayed when -h is
1021 given. The --detailed-help option prints the full help text.
1023 If para_server crashed or was killed by SIGKILL (signal 9), it
1024 may refuse to start again because of "dirty osl tables". In this
1025 case you'll have to run the oslfsck program of libosl to fix your
1028 oslfsck -fd ~/.paraslash/afs_database-0.4
1030 However, make sure para_server isn't running before executing oslfsck.
1032 If you don't mind to recreate your database you can start
1033 from scratch by removing the entire database directory, i.e.
1035 rm -rf ~/.paraslash/afs_database-0.4
1037 Be aware that this removes all attribute definitions, all playlists
1038 and all mood definitions and requires to re-initialize the tables.
1040 Although oslfsck fixes inconsistencies in database tables it doesn't
1041 care about the table contents. To check for invalid table contents, use
1045 This prints out references to missing audio files as well as invalid
1046 playlists and mood definitions.
1048 Similarly, para_audiod refuses to start if its socket file exists, since
1049 this indicates that another instance of para_audiod is running. After
1050 a crash a stale socket file might remain and you must run
1056 =======================================
1057 Audio formats and audio format handlers
1058 =======================================
1063 The following audio formats are supported by paraslash:
1067 Mp3, MPEG-1 Audio Layer 3, is a common audio format for audio storage,
1068 designed as part of its MPEG-1 standard. An MP3 file is made up of
1069 multiple MP3 frames, which consist of a header and a data block. The
1070 size of an MP3 frame depends on the bit rate and on the number
1071 of channels. For a typical CD-audio file (sample rate of 44.1 kHz
1072 stereo), encoded with a bit rate of 128 kbit, an MP3 frame is about
1077 OGG is a standardized audio container format, while Vorbis is an
1078 open source codec for lossy audio compression. Since Vorbis is most
1079 commonly made available via the OGG container format, it is often
1080 referred to as OGG/Vorbis. The OGG container format divides data into
1081 chunks called OGG pages. A typical OGG page is about 4KB large. The
1082 Vorbis codec creates variable-bitrate (VBR) data, where the bitrate
1083 may vary considerably.
1087 Speex is an open-source speech codec that is based on CELP (Code
1088 Excited Linear Prediction) coding. It is designed for voice
1089 over IP applications, has modest complexity and a small memory
1090 footprint. Wideband and narrowband (telephone quality) speech are
1091 supported. As for Vorbis audio, Speex bit-streams are often stored
1092 in OGG files. As of 2012 this codec is considered obsolete since the
1093 Oppus codec, described below, surpasses its performance in all areas.
1097 Opus is a lossy audio compression format standardized through RFC
1098 6716 in 2012. It combines the speech-oriented SILK codec and the
1099 low-latency CELT (Constrained Energy Lapped Transform) codec. Like
1100 OGG/Vorbis and OGG/Speex, Opus data is usually encapsulated in OGG
1101 containers. All known software patents which cover Opus are licensed
1102 under royalty-free terms.
1106 Advanced Audio Coding (AAC) is a standardized, lossy compression
1107 and encoding scheme for digital audio which is the default audio
1108 format for Apple's iPhone, iPod, iTunes. Usually MPEG-4 is used as
1109 the container format and audio files encoded with AAC have the .m4a
1110 extension. A typical AAC frame is about 700 bytes large.
1114 Windows Media Audio (WMA) is an audio data compression technology
1115 developed by Microsoft. A WMA file is usually encapsulated in the
1116 Advanced Systems Format (ASF) container format, which also specifies
1117 how meta data about the file is to be encoded. The bit stream of WMA
1118 is composed of superframes, each containing one or more frames of
1119 2048 samples. For 16 bit stereo a WMA superframe is about 8K large.
1123 The Free Lossless Audio Codec (FLAC) compresses audio without quality
1124 loss. It gives better compression ratios than a general purpose
1125 compressor like zip or bzip2 because FLAC is designed specifically
1126 for audio. A FLAC-encoded file consists of frames of varying size, up
1127 to 16K. Each frame starts with a header that contains all information
1128 necessary to decode the frame.
1133 Unfortunately, each audio format has its own conventions how meta
1134 data is added as tags to the audio file.
1136 For MP3 files, ID3, version 1 and 2 are widely used. ID3 version 1
1137 is rather simple but also very limited as it supports only artist,
1138 title, album, year and comment tags. Each of these can only be at most
1139 32 characters long. ID3, version 2 is much more flexible but requires
1140 a separate library being installed for paraslash to support it.
1142 Ogg vorbis, ogg speex and flac files contain meta data as Vorbis
1143 comments, which are typically implemented as strings of the form
1144 "[TAG]=[VALUE]". Unlike ID3 version 1 tags, one may use whichever
1145 tags are appropriate for the content.
1147 AAC files usually use the MPEG-4 container format for storing meta
1148 data while WMA files wrap meta data as special objects within the
1149 ASF container format.
1151 paraslash only tracks the most common tags that are supported by
1152 all tag variants: artist, title, year, album, comment. When a file
1153 is added to the AFS database, the meta data of the file is extracted
1154 and stored in the audio file table.
1156 Chunks and chunk tables
1157 -----------------------
1159 paraslash uses the word "chunk" as common term for the building blocks
1160 of an audio file. For MP3 files, a chunk is the same as an MP3 frame,
1161 while for OGG files a chunk is an OGG page, etc. Therefore the chunk
1162 size varies considerably between audio formats, from a few hundred
1163 bytes (MP3) up to 16K (FLAC).
1165 The chunk table contains the offsets within the audio file that
1166 correspond to the chunk boundaries of the file. Like the meta data,
1167 the chunk table is computed and stored in the database whenever an
1168 audio file is added.
1170 The paraslash senders (see below) always send complete chunks. The
1171 granularity for seeking is therefore determined by the chunk size.
1173 Audio format handlers
1174 ---------------------
1176 For each audio format paraslash contains an audio format handler whose
1177 first task is to tell whether a given file is a valid audio file of
1178 this type. If so, the audio file handler extracts some technical data
1179 (duration, sampling rate, number of channels etc.), computes the
1180 chunk table and reads the meta data.
1182 The audio format handler code is linked into para_server and executed
1183 via the _add_ command. The same code is also available as a stand-alone
1184 tool, para_afh, which prints the technical data, the chunk table
1185 and the meta data of a file. Moreover, all audio format handlers are
1186 combined in the afh receiver which is part of para_recv and para_play.
1192 Paraslash uses different network connections for control and data.
1193 para_client communicates with para_server over a dedicated TCP control
1194 connection. To transport audio data, separate data connections are
1195 used. For these data connections, a variety of transports (UDP, DCCP,
1196 HTTP) can be chosen.
1198 The chapter starts with the [control
1199 service](#The.paraslash.control.service), followed by a section
1200 on the various [streaming protocols](#Streaming.protocols)
1201 in which the data connections are described. The way
1202 audio file headers are embedded into the stream is discussed
1203 [briefly](#Streams.with.headers.and.headerless.streams) before the
1204 [example section](#Networking.examples) which illustrates typical
1205 commands for real-life scenarios.
1207 Both IPv4 and IPv6 are supported.
1209 The paraslash control service
1210 -----------------------------
1212 para_server is controlled at runtime via the paraslash control
1213 connection. This connection is used for server commands (play, stop,
1214 ...) as well as for afs commands (ls, select, ...).
1216 The server listens on a TCP port and accepts connections from clients
1217 that connect the open port. Each connection causes the server to fork
1218 off a client process which inherits the connection and deals with that
1219 client only. In this classical accept/fork approach the server process
1220 is unaffected if the child dies or goes crazy for whatever reason. In
1221 fact, the child process can not change address space of server process.
1223 The section on [client-server
1224 authentication](#Client-server.authentication) above described the
1225 early connection establishment from the crypto point of view. Here
1226 it is described what happens after the connection (including crypto
1227 setup) has been established. There are four processes involved during
1228 command dispatch as sketched in the following diagram.
1230 server_host client_host
1231 ~~~~~~~~~~~ ~~~~~~~~~~~
1233 +-----------+ connect +-----------+
1234 |para_server|<------------------------------ |para_client|
1235 +-----------+ +-----------+
1238 +----------> |AFS| |
1242 | | connect (cookie) |
1245 | fork +-----+ inherited connection |
1246 +---------->|child|<--------------------------+
1249 Note that the child process is not a child of the afs process,
1250 so communication of these two processes has to happen via local
1251 sockets. In order to avoid abuse of the local socket by unrelated
1252 processes, a magic cookie is created once at server startup time just
1253 before the server process forks off the AFS process. This cookie is
1254 known to the server, AFS and the child, but not to unrelated processes.
1256 There are two different kinds of commands: First there are commands
1257 that cause the server to respond with some answer such as the list
1258 of all audio files. All but the addblob commands (addimg, addlyr,
1259 addpl, addmood) are of this kind. The addblob commands add contents
1260 to the database, so they need to transfer data the other way round,
1261 from the client to the server.
1263 There is no knowledge about the server commands built into para_client,
1264 so it does not know about addblob commands. Instead, the server sends
1265 a special "awaiting data" packet for these commands. If the client
1266 receives this packet, it sends STDIN to the server, otherwise it
1267 dumps data from the server to STDOUT.
1272 A network (audio) stream usually consists of one streaming source,
1273 the _sender_, and one or more _receivers_ which read data over the
1274 network from the streaming source.
1276 Senders are thus part of para_server while receivers are part of
1277 para_audiod. Moreover, there is the stand-alone tool para_recv which
1278 can be used to manually download a stream, either from para_server
1279 or from a web-based audio streaming service.
1281 The following three streaming protocols are supported by paraslash:
1283 - HTTP. Recommended for public streams that can be played by any
1284 player like mpg123, xmms, itunes, winamp, etc. The HTTP sender is
1285 supported on all operating systems and all platforms.
1287 - DCCP. Recommended for LAN streaming. DCCP is currently available
1290 - UDP. Recommended for multicast LAN streaming.
1292 See the Appendix on [network protocols](/#Network.protocols)
1293 for brief descriptions of the various protocols relevant for network
1294 audio streaming with paraslash.
1296 It is possible to activate more than one sender simultaneously.
1297 Senders can be controlled at run time and via config file and command
1300 Note that audio connections are _not_ encrypted. Transport or Internet
1301 layer encryption should be used if encrypted data connections are
1304 Since DCCP and TCP are both connection-oriented protocols, connection
1305 establishment/teardown and access control are very similar between
1306 these two streaming protocols. UDP is the most lightweight option,
1307 since in contrast to TCP/DCCP it is connectionless. It is also the
1308 only protocol supporting IP multicast.
1310 The HTTP and the DCCP sender listen on a (TCP/DCCP) port waiting for
1311 clients to connect and establish a connection via some protocol-defined
1312 handshake mechanism. Both senders maintain two linked lists each:
1313 The list of all clients which are currently connected, and the list
1314 of access control entries which determines who is allowed to connect.
1315 IP-based access control may be configured through config file and
1316 command line options and via the "allow" and "deny" sender subcommands.
1318 Upon receiving a GET request from the client, the HTTP sender sends
1319 back a status line and a message. The body of this message is the
1320 audio stream. This is common practice and is supported by many popular
1321 clients which can thus be used to play a stream offered by para_server.
1322 For DCCP things are a bit simpler: No messages are exchanged between
1323 the receiver and sender. The client simply connects and the sender
1326 DCCP is an experimental protocol which offers a number of new features
1327 not available for TCP. Both ends can negotiate these features using
1328 a built-in negotiation mechanism. In contrast to TCP/HTTP, DCCP is
1329 datagram-based (no retransmissions) and thus should not be used over
1330 lossy media (e.g. WiFi networks). One useful feature offered by DCCP
1331 is access to a variety of different congestion-control mechanisms
1332 called CCIDs. Two different CCIDs are available per default on Linux:
1335 - _CCID 2_. A Congestion Control mechanism similar to that of TCP. The
1336 sender maintains a congestion window and halves this window in response
1340 - _CCID-3_. Designed to be fair when competing for bandwidth.
1341 It has lower variation of throughput over time compared with TCP,
1342 which makes it suitable for streaming media.
1344 Unlike the HTTP and DCCP senders, the UDP sender maintains only a
1345 single list, the _target list_. This list describes the set of clients
1346 to which the stream is sent. There is no list for access control and
1347 no "allow" and "deny" commands for the UDP sender. Instead, the "add"
1348 and "delete" commands can be used to modify the target list.
1350 Since both UDP and DCCP offer an unreliable datagram-based transport,
1351 additional measures are necessary to guard against disruptions over
1352 networks that are lossy or which may be subject to interference (as
1353 is for instance the case with WiFi). Paraslash uses FEC (Forward
1354 Error Correction) to guard against packet losses and reordering. The
1355 stream is FEC-encoded before it is sent through the UDP socket and
1356 must be decoded accordingly on the receiver side.
1358 The packet size and the amount of redundancy introduced by FEC can
1359 be configured via the FEC parameters which are dictated by server
1360 and may also be configured through the "sender" command. The FEC
1361 parameters are encoded in the header of each network packet, so no
1362 configuration is necessary on the receiver side. See the section on
1363 [FEC](#Forward.error.correction) below.
1365 Streams with headers and headerless streams
1366 -------------------------------------------
1368 For OGG/Vorbis, OGG/Speex and wma streams, some of the information
1369 needed to decode the stream is only contained in the audio file
1370 header of the container format but not in each data chunk. Clients
1371 must be able to obtain this information in case streaming starts in
1372 the middle of the file or if para_audiod is started while para_server
1373 is already sending a stream.
1375 This is accomplished in different ways, depending on the streaming
1376 protocol. For connection-oriented streams (HTTP, DCCP) the audio file
1377 header is sent prior to audio file data. This technique however does
1378 not work for the connectionless UDP transport. Hence the audio file
1379 header is periodically being embedded into the UDP audio data stream.
1380 By default, the header is resent after five seconds. The receiver has
1381 to wait until the next header arrives before it can start decoding
1387 The "si" (server info) command lists some information about the
1388 currently running server process.
1390 -> Show PIDs, number of connected clients, uptime, and more:
1394 The sender command of para_server prints information about senders,
1395 like the various access control lists, and it allows to (de-)activate
1396 senders and to change the access permissions at runtime.
1402 -> Obtain general help for the sender command:
1404 para_client help sender
1406 -> Get help for a specific sender (contains further examples):
1408 s=http # or dccp or udp
1409 para_client sender $s help
1411 -> Show status of the http sender
1413 para_client sender http status
1415 By default para_server activates both the HTTP and th DCCP sender on
1416 startup. This can be changed via command line options or para_server's
1419 -> List config file options for senders:
1423 All senders share the "on" and "off" commands, so senders may be
1424 activated and deactivated independently of each other.
1426 -> Switch off the http sender:
1428 para_client sender http off
1430 -> Receive a DCCP stream using CCID2 and write the output into a file:
1432 host=foo.org; ccid=2; filename=bar
1433 para_recv --receiver "dccp --host $host --ccid $ccid" > $filename
1435 Note the quotes around the arguments for the dccp receiver. Each
1436 receiver has its own set of command line options and its own command
1437 line parser, so arguments for the dccp receiver must be protected
1438 from being interpreted by para_recv.
1440 -> Start UDP multicast, using the default multicast address:
1442 para_client sender udp add 224.0.1.38
1444 -> Receive FEC-encoded multicast stream and write the output into a file:
1447 para_recv -r udp > $filename
1449 -> Add an UDP unicast for a client to the target list of the UDP sender:
1452 para_client sender udp add $t
1454 -> Receive this (FEC-encoded) unicast stream:
1457 para_recv -r 'udp -i 0.0.0.0' > $filename
1459 -> Create a minimal config for para_audiod for HTTP streams:
1461 c=$HOME/.paraslash/audiod.conf.min; s=server.foo.com
1462 echo receiver \".:http -i $s\" > $c
1463 para_audiod --config $c
1469 A paraslash filter is a module which transforms an input stream into
1470 an output stream. Filters are included in the para_audiod executable
1471 and in the stand-alone tool para_filter which usually contains the
1474 While para_filter reads its input stream from STDIN and writes
1475 the output to STDOUT, the filter modules of para_audiod are always
1476 connected to a receiver which produces the input stream and a writer
1477 which absorbs the output stream.
1479 Some filters depend on a specific library and are not compiled in
1480 if this library was not found at compile time. To see the list of
1481 supported filters, run para_filter and para_audiod with the --help
1482 option. The output looks similar to the following:
1485 compress wav amp fecdec wmadec prebuffer oggdec aacdec mp3dec
1487 Out of these filter modules, a chain of filters can be constructed,
1488 much in the way Unix pipes can be chained, and analogous to the use
1489 of modules in gstreamer: The output of the first filter becomes the
1490 input of the second filter. There is no limitation on the number of
1491 filters and the same filter may occur more than once.
1493 Like receivers, each filter has its own command line options which
1494 must be quoted to protect them from the command line options of
1495 the driving application (para_audiod or para_filter). Example:
1497 para_filter -f 'mp3dec --ignore-crc' -f 'compress --damp 1'
1499 For para_audiod, each audio format has its own set of filters. The
1500 name of the audio format for which the filter should be applied can
1501 be used as the prefix for the filter option. Example:
1503 para_audiod -f 'mp3:prebuffer --duration 300'
1505 The "mp3" prefix above is actually interpreted as a POSIX extended
1506 regular expression. Therefore
1508 para_audiod -f '.:prebuffer --duration 300'
1510 activates the prebuffer filter for all supported audio formats (because
1511 "." matches all audio formats) while
1513 para_audiod -f 'wma|ogg:prebuffer --duration 300'
1515 activates it only for wma and ogg streams.
1520 For each supported audio format there is a corresponding filter
1521 which decodes audio data in this format to 16 bit PCM data which
1522 can be directly sent to the sound device or any other software that
1523 operates on undecoded PCM data (visualizers, equalizers etc.). Such
1524 filters are called _decoders_ in general, and xxxdec is the name of
1525 the paraslash decoder for the audio format xxx. For example, the mp3
1526 decoder is called mp3dec.
1528 Note that the output of the decoder is about 10 times larger than
1529 its input. This means that filters that operate on the decoded audio
1530 stream have to deal with much more data than filters that transform
1531 the audio stream before it is fed to the decoder.
1533 Paraslash relies on external libraries for most decoders, so these
1534 libraries must be installed for the decoder to be included in the
1535 executables. For example, the mp3dec filter depends on the mad library.
1537 Forward error correction
1538 ------------------------
1540 As already mentioned [earlier](#Streaming.protocols), paraslash
1541 uses forward error correction (FEC) for the unreliable UDP and
1542 DCCP transports. FEC is a technique which was invented already in
1543 1960 by Reed and Solomon and which is widely used for the parity
1544 calculations of storage devices (RAID arrays). It is based on the
1545 algebraic concept of finite fields, today called Galois fields, in
1546 honour of the mathematician Galois (1811-1832). The FEC implementation
1547 of paraslash is based on code by Luigi Rizzo.
1549 Although the details require a sound knowledge of the underlying
1550 mathematics, the basic idea is not hard to understand: For positive
1551 integers k and n with k < n it is possible to compute for any k given
1552 data bytes d_1, ..., d_k the corresponding r := n -k parity bytes p_1,
1553 ..., p_r such that all data bytes can be reconstructed from *any*
1556 {d_1, ..., d_k, p_1, ..., p_r}.
1558 FEC-encoding for unreliable network transports boils down to slicing
1559 the audio stream into groups of k suitably sized pieces called _slices_
1560 and computing the r corresponding parity slices. This step is performed
1561 in para_server which then sends both the data and the parity slices
1562 over the unreliable network connection. If the client was able
1563 to receive at least k of the n = k + r slices, it can reconstruct
1564 (FEC-decode) the original audio stream.
1566 From these observations it is clear that there are three different
1567 FEC parameters: The slice size, the number of data slices k, and the
1568 total number of slices n. It is crucial to choose the slice size
1569 such that no fragmentation of network packets takes place because
1570 FEC only guards against losses and reordering but fails if slices are
1573 FEC decoding in paralash is performed through the fecdec filter which
1574 usually is the first filter (there can be other filters before fecdec
1575 if these do not alter the audio stream).
1577 Volume adjustment (amp and compress)
1578 ------------------------------------
1580 The amp and the compress filter both adjust the volume of the audio
1581 stream. These filters operate on uncompressed audio samples. Hence
1582 they are usually placed directly after the decoding filter. Each
1583 sample is multiplied with a scaling factor (>= 1) which makes amp
1584 and compress quite expensive in terms of computing power.
1588 The amp filter amplifies the audio stream by a fixed scaling factor
1589 that must be known in advance. For para_audiod this factor is derived
1590 from the amplification field of the audio file's entry in the audio
1591 file table while para_filter uses the value given at the command line.
1593 The optimal scaling factor F for an audio file is the largest real
1594 number F >= 1 such that after multiplication with F all samples still
1595 fit into the sample interval [-32768, 32767]. One can use para_filter
1596 in combination with the sox utility to compute F:
1598 para_filter -f mp3dec -f wav < file.mp3 | sox -t wav - -e stat -v
1600 The amplification value V which is stored in the audio file table,
1601 however, is an integer between 0 and 255 which is connected to F
1606 To store V in the audio file table, the command
1608 para_client -- touch -a=V file.mp3
1610 is used. The reader is encouraged to write a script that performs
1611 these computations :)
1615 Unlike the amplification filter, the compress filter adjusts the volume
1616 of the audio stream dynamically without prior knowledge about the peak
1617 value. It maintains the maximal volume of the last n samples of the
1618 audio stream and computes a suitable amplification factor based on that
1619 value and the various configuration options. It tries to chose this
1620 factor such that the adjusted volume meets the desired target level.
1622 Note that it makes sense to combine amp and compress.
1624 Misc filters (wav and prebuffer)
1625 --------------------------------
1627 These filters are rather simple and do not modify the audio stream at
1628 all. The wav filter is only useful with para_filter and in connection
1629 with a decoder. It asks the decoder for the number of channels and the
1630 sample rate of the stream and adds a Microsoft wave header containing
1631 this information at the beginning. This allows to write wav files
1632 rather than raw PCM files (which do not contain any information about
1633 the number of channels and the sample rate).
1635 The prebuffer filter simply delays the output until the given time has
1636 passed (starting from the time the first byte was available in its
1637 input queue) or until the given amount of data has accumulated. It
1638 is mainly useful for para_audiod if the standard parameters result
1639 in buffer underruns.
1641 Both filters require almost no additional computing time, even when
1642 operating on uncompressed audio streams, since data buffers are simply
1643 "pushed down" rather than copied.
1648 -> Decode an mp3 file to wav format:
1650 para_filter -f mp3dec -f wav < file.mp3 > file.wav
1652 -> Amplify a raw audio file by a factor of 1.5:
1654 para_filter -f amp --amp 32 < foo.raw > bar.raw
1660 Once an audio stream has been received and decoded to PCM format,
1661 it can be sent to a sound device for playback. This part is performed
1662 by paraslash _writers_ which are described in this chapter.
1667 A paraslash writer acts as a data sink that consumes but does not
1668 produce audio data. Paraslash writers operate on the client side and
1669 are contained in para_audiod and in the stand-alone tool para_write.
1671 The para_write program reads uncompressed audio data from STDIN. If
1672 this data starts with a wav header, sample rate, sample format and
1673 channel count are read from the header. Otherwise CD audio (44.1KHz
1674 16 bit little endian, stereo) is assumed but this can be overridden
1675 by command line options. para_audiod, on the other hand, obtains
1676 the sample rate and the number of channels from the decoder.
1678 Like receivers and filters, each writer has an individual set of
1679 command line options, and for para_audiod writers can be configured
1680 per audio format separately. It is possible to activate more than
1681 one writer for the same stream simultaneously.
1686 Unfortunately, the various flavours of Unix on which paraslash
1687 runs on have different APIs for opening a sound device and starting
1688 playback. Hence for each such API there is a paraslash writer that
1689 can play the audio stream via this API.
1691 - *ALSA*. The _Advanced Linux Sound Architecture_ is only available on
1692 Linux systems. Although there are several mid-layer APIs in use by
1693 the various Linux distributions (ESD, Jack, PulseAudio), paraslash
1694 currently supports only the low-level ALSA API which is not supposed
1695 to be change. ALSA is very feature-rich, in particular it supports
1696 software mixing via its DMIX plugin. ALSA is the default writer on
1699 - *OSS*. The _Open Sound System_ is the only API on \*BSD Unixes and
1700 is also available on Linux systems, usually provided by ALSA as an
1701 emulation for backwards compatibility. This API is rather simple but
1702 also limited. For example only one application can open the device
1703 at any time. The OSS writer is activated by default on BSD Systems.
1705 - *FILE*. The file writer allows to capture the audio stream and
1706 write the PCM data to a file on the file system rather than playing
1707 it through a sound device. It is supported on all platforms and is
1710 - *AO*. _Libao_ is a cross-platform audio library which supports a wide
1711 variety of platforms including PulseAudio (gnome), ESD (Enlightened
1712 Sound Daemon), AIX, Solaris and IRIX. The ao writer plays audio
1713 through an output plugin of libao.
1718 -> Use the OSS writer to play a wav file:
1720 para_write --writer oss < file.wav
1722 -> Enable ALSA software mixing for mp3 streams:
1724 para_audiod --writer 'mp3:alsa -d plug:swmix'
1731 para_gui executes an arbitrary command which is supposed to print
1732 status information to STDOUT. It then displays this information in
1733 a curses window. By default the command
1735 para_audioc -- stat -p
1737 is executed, but this can be customized via the --stat-cmd option. In
1738 particular it possible to use
1740 para_client -- stat -p
1742 to make para_gui work on systems on which para_audiod is not running.
1747 It is possible to bind keys to arbitrary commands via custom
1748 key-bindings. Besides the internal keys which can not be changed (help,
1749 quit, loglevel, version...), the following flavours of key-bindings
1752 - external: Shutdown curses before launching the given command.
1753 Useful for starting other ncurses programs from within para_gui,
1754 e.g. aumix or dialog scripts. Or, use the mbox output format to write
1755 a mailbox containing one mail for each (admissible) file the audio
1756 file selector knows about. Then start mutt from within para_gui to
1757 browse your collection!
1759 - display: Launch the command and display its stdout in para_gui's
1762 - para: Like display, but start "para_client <specified command>"
1763 instead of "<specified command>".
1765 The general form of a key binding is
1769 which maps key k to command c using mode m. Mode may be x, d or p
1770 for external, display and paraslash commands, respectively.
1775 Currently there are only two themes for para_gui. It is easy, however,
1776 to add more themes. To create a new theme one has to define the
1777 position, color and geometry for for each status item that should be
1778 shown by this theme. See gui_theme.c for examples.
1780 The "." and "," keys are used to switch between themes.
1785 -> Show server info:
1789 -> Jump to the middle of the current audio file by pressing F5:
1791 key_map "<F5>:p:jmp 50"
1793 -> vi-like bindings for jumping around:
1796 key_map "h:p:ff 10-"
1798 key_map "b:p:ff 60-"
1800 -> Print the current date and time:
1804 -> Call other curses programs:
1807 key_map "!:x:/bin/bash"
1808 key_map "^E:x:/bin/sh -c 'vi ~/.paraslash/gui.conf'"
1817 Paraslash is an open source project and contributions are
1818 welcome. Here's a list of things you can do to help the project:
1820 - Report problems with building, installing or running the software.
1821 In particular, test the experimental git branches ("next" and "pu").
1822 This helps to identify and fix problems before the code gets merged
1823 and thus keeps the master branch as stable as possible.
1824 - Proofread the documentation (manual, web pages, man pages, source
1825 code documentation) and point out unclear or poorly written parts. If
1826 you are a native English speaker you will easily find a lot of text
1827 that could be improved.
1828 - Run analysis tools (coverity, afl, sparse, etc.) and report issues
1829 found by those tools.
1830 - Suggest new features you would like to see implemented.
1831 - Compile and test on your favorite architecture or operating
1832 system. The code is tested only on a limited set of systems, so you
1833 will probably encounter problems when building on different systems.
1834 - Post about about paraslash on your blog or on social networks.
1835 - Build and maintain Debian/RPM packages for your favorite distribution.
1837 Note that there is no mailing list, no bug tracker and no discussion
1838 forum for paraslash. If you'd like to contribute, or have questions
1839 about contributing, send email to Andre Noll <maan@tuebingen.mpg.de>.
1844 In order to compile the sources from the git repository (rather than
1845 from tar balls) and for contributing non-trivial changes to the
1846 paraslash project, some additional tools should be installed on a
1849 - [git](http://git.or.cz/). As described in more detail
1850 [below](#Git.branches), the git source code management tool is used for
1851 paraslash development. It is necessary for cloning the git repository
1852 and for getting updates.
1854 - [autoconf](ftp://ftp.gnu.org/pub/gnu/autoconf/) GNU autoconf creates
1855 the configure file which is shipped in the tarballs but has to be
1856 generated when compiling from git.
1858 - [discount](http://www.pell.portland.or.us/~orc/Code/discount). The
1859 HTML version of this manual and some of the paraslash web pages are
1860 written in the Markdown markup language and are translated into html
1861 with the converter of the *Discount* package.
1863 - [doxygen](http://www.stack.nl/~dimitri/doxygen/). The documentation
1864 of paraslash's C sources uses the doxygen documentation system. The
1865 conventions for documenting the source code is described in the
1866 [Doxygen section](#Doxygen).
1868 - [global](ftp://ftp.gnu.org/pub/gnu/global). This is used to generate
1869 browsable HTML from the C sources. It is needed by doxygen.
1874 Paraslash has been developed using the git source code management
1875 tool since 2006. Development is organized roughly in the same spirit
1876 as the git development itself, as described below.
1878 The following text passage is based on "A note from the maintainer",
1879 written by Junio C Hamano, the maintainer of git.
1881 There are four branches in the paraslash repository that track the
1882 source tree: "master", "maint", "next", and "pu".
1884 The "master" branch is meant to contain what is well tested and
1885 ready to be used in a production setting. There could occasionally be
1886 minor breakages or brown paper bag bugs but they are not expected to
1887 be anything major, and more importantly quickly and easily fixable.
1888 Every now and then, a "feature release" is cut from the tip of this
1889 branch, named with three dotted decimal digits, like 0.4.2.
1891 Whenever changes are about to be included that will eventually lead to
1892 a new major release (e.g. 0.5.0), a "maint" branch is forked off from
1893 "master" at that point. Obvious, safe and urgent fixes after the major
1894 release are applied to this branch and maintenance releases are cut
1895 from it. New features never go to this branch. This branch is also
1896 merged into "master" to propagate the fixes forward.
1898 A trivial and safe enhancement goes directly on top of "master".
1899 New development does not usually happen on "master", however.
1900 Instead, a separate topic branch is forked from the tip of "master",
1901 and it first is tested in isolation; Usually there are a handful such
1902 topic branches that are running ahead of "master". The tip of these
1903 branches is not published in the public repository to keep the number
1904 of branches that downstream developers need to worry about low.
1906 The quality of topic branches varies widely. Some of them start out as
1907 "good idea but obviously is broken in some areas" and then with some
1908 more work become "more or less done and can now be tested by wider
1909 audience". Luckily, most of them start out in the latter, better shape.
1911 The "next" branch is to merge and test topic branches in the latter
1912 category. In general, this branch always contains the tip of "master".
1913 It might not be quite rock-solid production ready, but is expected to
1914 work more or less without major breakage. The maintainer usually uses
1915 the "next" version of paraslash for his own pleasure, so it cannot
1916 be _that_ broken. The "next" branch is where new and exciting things
1919 The two branches "master" and "maint" are never rewound, and "next"
1920 usually will not be either (this automatically means the topics that
1921 have been merged into "next" are usually not rebased, and you can find
1922 the tip of topic branches you are interested in from the output of
1923 "git log next"). You should be able to safely build on top of them.
1925 However, at times "next" will be rebuilt from the tip of "master" to
1926 get rid of merge commits that will never be in "master". The commit
1927 that replaces "next" will usually have the identical tree, but it
1928 will have different ancestry from the tip of "master".
1930 The "pu" (proposed updates) branch bundles the remainder of the
1931 topic branches. The "pu" branch, and topic branches that are only in
1932 "pu", are subject to rebasing in general. By the above definition
1933 of how "next" works, you can tell that this branch will contain quite
1934 experimental and obviously broken stuff.
1936 When a topic that was in "pu" proves to be in testable shape, it
1937 graduates to "next". This is done with
1940 git merge that-topic-branch
1942 Sometimes, an idea that looked promising turns out to be not so good
1943 and the topic can be dropped from "pu" in such a case.
1945 A topic that is in "next" is expected to be polished to perfection
1946 before it is merged to "master". Similar to the above, this is
1950 git merge that-topic-branch
1951 git branch -d that-topic-branch
1953 Note that being in "next" is not a guarantee to appear in the next
1954 release (being in "master" is such a guarantee, unless it is later
1955 found seriously broken and reverted), nor even in any future release.
1960 The preferred coding style for paraslash coincides more or less
1961 with the style of the Linux kernel. So rather than repeating what is
1962 written [there](http://www.kernel.org/doc/Documentation/process/coding-style.rst),
1963 here are the most important points.
1965 - Burn the GNU coding standards.
1966 - Never use spaces for indentation.
1967 - Tabs are 8 characters, and thus indentations are also 8 characters.
1968 - Don't put multiple assignments on a single line.
1969 - Avoid tricky expressions.
1970 - Don't leave whitespace at the end of lines.
1971 - The limit on the length of lines is 80 columns.
1972 - Use K&R style for placing braces and spaces:
1978 - Use a space after (most) keywords.
1979 - Do not add spaces around (inside) parenthesized expressions.
1980 - Use one space around (on each side of) most binary and ternary operators.
1981 - Do not use cute names like ThisVariableIsATemporaryCounter, call it tmp.
1982 - Mixed-case names are frowned upon.
1983 - Descriptive names for global variables are a must.
1985 - Functions should be short and sweet, and do just one thing.
1986 - The number of local variables shouldn't exceed 10.
1987 - Gotos are fine if they improve readability and reduce nesting.
1988 - Don't use C99-style "// ..." comments.
1989 - Names of macros defining constants and labels in enums are capitalized.
1990 - Enums are preferred when defining several related constants.
1991 - Always use the paraslash wrappers for allocating memory.
1992 - If the name of a function is an action or an imperative.
1993 command, the function should return an error-code integer
1994 (<0 means error, >=0 means success). If the name is a
1995 predicate, the function should return a "succeeded" boolean.
2000 Doxygen is a documentation system for various programming
2001 languages. The API reference on the paraslash web page is generated
2004 It is more illustrative to look at the source code for examples than
2005 to describe the conventions in this manual, so we only describe which
2006 parts of the code need doxygen comments, but leave out details on
2007 documentation conventions.
2009 As a rule, only the public part of the C source is documented with
2010 Doxygen. This includes structures, defines and enumerations in header
2011 files as well as public (non-static) C functions. These should be
2012 documented completely. For example, each parameter and the return
2013 value of a public function should get a descriptive doxygen comment.
2015 No doxygen comments are necessary for static functions and for
2016 structures and enumerations in C files (which are used only within
2017 this file). This does not mean, however, that those entities need
2018 no documentation at all. Instead, common sense should be applied to
2019 document what is not obvious from reading the code.
2030 The _Internet Protocol_ is the primary networking protocol used for
2031 the Internet. All protocols described below use IP as the underlying
2032 layer. Both the prevalent IPv4 and the next-generation IPv6 variant
2033 are being deployed actively worldwide.
2035 ### Connection-oriented and connectionless protocols ###
2037 Connectionless protocols differ from connection-oriented ones in
2038 that state associated with the sending/receiving endpoints is treated
2039 implicitly. Connectionless protocols maintain no internal knowledge
2040 about the state of the connection. Hence they are not capable of
2041 reacting to state changes, such as sudden loss or congestion on the
2042 connection medium. Connection-oriented protocols, in contrast, make
2043 this knowledge explicit. The connection is established only after
2044 a bidirectional handshake which requires both endpoints to agree
2045 on the state of the connection, and may also involve negotiating
2046 specific parameters for the particular connection. Maintaining an
2047 up-to-date internal state of the connection also in general means
2048 that the sending endpoints perform congestion control, adapting to
2049 qualitative changes of the connection medium.
2053 In IP networking, packets can be lost, duplicated, or delivered
2054 out of order, and different network protocols handle these
2055 problems in different ways. We call a transport-layer protocol
2056 _reliable_, if it turns the unreliable IP delivery into an ordered,
2057 duplicate- and loss-free delivery of packets. Sequence numbers
2058 are used to discard duplicates and re-arrange packets delivered
2059 out-of-order. Retransmission is used to guarantee loss-free
2060 delivery. Unreliable protocols, in contrast, do not guarantee ordering
2063 ### Classification ###
2065 With these definitions the protocols which are used by paraslash for
2066 steaming audio data may be classified as follows.
2068 - HTTP/TCP: connection-oriented, reliable,
2069 - UDP: connectionless, unreliable,
2070 - DCCP: connection-oriented, unreliable.
2072 Below we give a short descriptions of these protocols.
2076 The _Transmission Control Protocol_ provides reliable, ordered delivery
2077 of a stream and a classic window-based congestion control. In contrast
2078 to UDP and DCCP (see below), TCP does not have record-oriented or
2079 datagram-based syntax, i.e. it provides a stream which is unaware
2080 and independent of any record (packet) boundaries. TCP is used
2081 extensively by many application layers. Besides HTTP (the Hypertext
2082 Transfer Protocol), also FTP (the File Transfer protocol), SMTP (Simple
2083 Mail Transfer Protocol), SSH (Secure Shell) all sit on top of TCP.
2087 The _User Datagram Protocol_ is the simplest transport-layer protocol,
2088 built as a thin layer directly on top of IP. For this reason, it offers
2089 the same best-effort service as IP itself, i.e. there is no detection
2090 of duplicate or reordered packets. Being a connectionless protocol,
2091 only minimal internal state about the connection is maintained, which
2092 means that there is no protection against packet loss or network
2093 congestion. Error checking and correction (if at all) are performed
2098 The _Datagram Congestion Control Protocol_ combines the
2099 connection-oriented state maintenance known from TCP with the
2100 unreliable, datagram-based transport of UDP. This means that it
2101 is capable of reacting to changes in the connection by performing
2102 congestion control, offering multiple alternative approaches. But it
2103 is bound to datagram boundaries (the maximum packet size supported
2104 by a medium), and like UDP it lacks retransmission to protect
2105 against loss. Due to the use of sequence numbers, it is however
2106 able to react to loss (interpreted as a congestion indication) and
2107 to ignore out-of-order and duplicate packets. Unlike TCP it allows
2108 to negotiate specific, binding features for a connection, such as
2109 the choice of congestion control: classic, window-based congestion
2110 control known from TCP is available as CCID-2, rate-based, "smooth"
2111 congestion control is offered as CCID-3.
2115 The _Hypertext Transfer Protocol_ is an application layer protocol
2116 on top of TCP. It is spoken by web servers and is most often used
2117 for web services. However, as can be seen by the many Internet radio
2118 stations and YouTube/Flash videos, http is by far not limited to the
2119 delivery of web pages only. Being a simple request/response based
2120 protocol, the semantics of the protocol also allow the delivery of
2121 multimedia content, such as audio over http.
2125 IP multicast is not really a protocol but a technique for one-to-many
2126 communication over an IP network. The challenge is to deliver
2127 information to a group of destinations simultaneously using the
2128 most efficient strategy to send the messages over each link of the
2129 network only once. This has benefits for streaming multimedia: the
2130 standard one-to-one unicast offered by TCP/DCCP means that n clients
2131 listening to the same stream also consume n-times the resources,
2132 whereas multicast requires to send the stream just once, irrespective
2133 of the number of receivers. Since it would be costly to maintain state
2134 for each listening receiver, multicast often implies connectionless
2135 transport, which is the reason that it is currently only available
2138 Abstract socket namespace
2139 -------------------------
2140 UNIX domain sockets are a traditional way to communicate between
2141 processes on the same machine. They are always reliable (see above)
2142 and don't reorder datagrams. Unlike TCP and UDP, UNIX domain sockets
2143 support passing open file descriptors or process credentials to
2146 The usual way to set up a UNIX domain socket (as obtained from
2147 socket(2)) for listening is to first bind the socket to a file system
2148 pathname and then call listen(2), then accept(2). Such sockets are
2149 called _pathname sockets_ because bind(2) creates a special socket
2150 file at the specified path. Pathname sockets allow unrelated processes
2151 to communicate with the listening process by binding to the same path
2152 and calling connect(2).
2154 There are two problems with pathname sockets:
2156 * The listing process must be able to (safely) create the
2157 socket special in a directory which is also accessible to
2158 the connecting process.
2160 * After an unclean shutdown of the listening process, a stale
2161 socket special may reside on the file system.
2163 The abstract socket namespace is a non-portable Linux feature which
2164 avoids these problems. Abstract sockets are still bound to a name,
2165 but the name has no connection with file system pathnames.
2170 Paraslash is licensed under the GPL, version 2. Most of the code
2171 base has been written from scratch, and those parts are GPL V2
2172 throughout. Notable exceptions are FEC and the WMA decoder. See the
2173 corresponding source files for licencing details for these parts. Some
2174 code sniplets of several other third party software packages have
2175 been incorporated into the paraslash sources, for example log message
2176 coloring was taken from the git sources. These third party software
2177 packages are all published under the GPL or some other license
2178 compatible to the GPL.
2183 Many thanks to Gerrit Renker who read an early draft of this manual
2184 and contributed significant improvements.
2192 - [Polynomial Codes over Certain Finite
2193 Fields](http://kom.aau.dk/~heb/kurser/NOTER/KOFA01.PDF) by Reed, Irving
2194 S.; Solomon, Gustave (1960), Journal of the Society for Industrial
2195 and Applied Mathematics (SIAM) 8 (2): 300-304, doi:10.1137/0108018)
2200 - [RFC 768](http://www.ietf.org/rfc/rfc768.txt) (1980): User Datagram
2203 - [RFC 791](http://www.ietf.org/rfc/rfc791.txt) (1981): Internet
2206 - [RFC 2437](http://www.ietf.org/rfc/rfc2437.txt) (1998): RSA
2207 Cryptography Specifications
2209 - [RFC 4340](http://www.ietf.org/rfc/rfc4340.txt) (2006): Datagram
2210 Congestion Control Protocol (DCCP)
2212 - [RFC 4341](http://www.ietf.org/rfc/rfc4341.txt) (2006): Congestion
2213 Control ID 2: TCP-like Congestion Control
2215 - [RFC 4342](http://www.ietf.org/rfc/rfc4342.txt) (2006): Congestion
2216 Control ID 3: TCP-Friendly Rate Control (TFRC)
2218 - [RFC 6716](http://www.ietf.org/rfc/rfc6716.txt) (2012): Definition
2219 of the Opus Audio Codec
2221 Application web pages
2222 ---------------------
2224 - [paraslash](http://people.tuebingen.mpg.de/maan/paraslash/)
2225 - [xmms](http://xmms2.org/wiki/Main_Page)
2226 - [mpg123](http://www.mpg123.de/)
2227 - [gstreamer](http://gstreamer.freedesktop.org/)
2228 - [icecast](http://www.icecast.org/)
2229 - [Audio Compress](http://beesbuzz.biz/code/audiocompress.php)
2231 External documentation
2232 ----------------------
2234 - [The mathematics of
2235 Raid6](http://kernel.org/pub/linux/kernel/people/hpa/raid6.pdf)
2238 - [Effective Erasure Codes for reliable Computer Communication
2239 Protocols](http://info.iet.unipi.it/~luigi/fec_ccr.ps.gz) by Luigi
2245 implementation](http://info.iet.unipi.it/~luigi/vdm.tar.gz) by