\documentclass[12pt, a4paper, oneside, titlepage, listof=totoc, headinclude,
open=right, glossaries=totoc, captions=tableheading, final, BCOR=10mm,
parskip=never]{paper}
\usepackage[english]{babel}
\usepackage[ngerman]{babel}
\usepackage[utf8]{inputenc}
\usepackage[T1]{fontenc}
\usepackage{textcomp}
\usepackage[usenames,dvipsnames,rgb,svgnames,table]{xcolor}
\definecolor{osc-out}{RGB}{150,0,255}
\definecolor{osc-in}{RGB}{0,0,255}
\definecolor{audio-in}{RGB}{255,0,0}
\definecolor{audio-out}{RGB}{0,206,0}
\definecolor{pubsub-in}{RGB}{128,0,0}
\definecolor{controller-in}{RGB}{0,128,0}
\definecolor{table-background-one}{RGB}{194,195,194}
\definecolor{table-background-two}{RGB}{210,213,210}

\usepackage[newfloat=true]{minted}
\usepackage{mdframed}

\usepackage{fancyhdr}
\usepackage{url}
\usepackage{graphicx}
\usepackage{pdfpages}
\usepackage[author={David Runge}]{pdfcomment}
\usepackage{mathtools}

\usepackage{float}
\floatstyle{boxed}
\restylefloat{figure}

\usepackage{tocloft}
\setcounter{secnumdepth}{4}

\usepackage{hyperref}
\hypersetup{hidelinks, colorlinks = false}
\usepackage[authoryear]{natbib}

% caption
\usepackage[font=scriptsize]{caption}
% glossary
\usepackage[acronym,nonumberlist,toc,xindy]{glossaries}
\makeglossaries
\newacronym{aap}{AAP}{Ambisonics Amplitude Panning}
\newacronym{adat}{ADAT}{Alesis Digital Audio Tape}
\newacronym{alsa}{ALSA}{Advanced Linux Sound Architecture}
\newacronym{apf}{APF}{Audio Processing Framework}
\newacronym{api}{API}{Application programming interface}
\newacronym{asdf}{ASDF}{Audio Scene Description Format}
\newacronym{bs}{BS}{Binaural Synthesis}
\newacronym{brir}{BRIR}{Binaural Room Impulse Response}
\newacronym{brs}{BRS}{Binaural Room Synthesis}
\newacronym{cc}{CC}{Creative Commons}
\newacronym{cicm}{CICM}{Centre de recherche Informatique et Création Musicale}
\newacronym{cnmat}{CNMAT}{Center for New Music and Audio Technologies}
\newacronym{cpu}{CPU}{Central Processing Unit}
\newacronym{fdl}{FDL}{GNU Free Documentation License}
\newacronym{gpl}{GPL}{GNU General Public License}
\newacronym{gui}{GUI}{Graphical user interface}
\newacronym{hrir}{HRIR}{Head Related Impulse Response}
\newacronym{hrtf}{HRTF}{Head Related Transfer Function}
\newacronym{iana}{IANA}{Internet Assigned Numbers Authority}
\newacronym{ide}{IDE}{Integrated Development Environment}
\newacronym{lgpl}{LGPL}{GNU Lesser General Public License}
\newacronym{lts}{LTS}{Long Term Support}
\newacronym{hoa}{HOA}{Higher Order Ambisonics}
\newacronym{ip}{IP}{Internet Protocol}
\newacronym{jack}{JACK}{JACK Audio Connection Kit}
\newacronym{madi}{MADI}{Multichannel Audio Digital Interface}
\newacronym{midi}{MIDI}{Musical Instrument Digital Interface}
\newacronym{nfc-hoa}{NFC-HOA}{Near-field-compensated Higher Order Ambisonics}
\newacronym{oop}{OOP}{Object-oriented Programming}
\newacronym{os}{OS}{Operating System}
\newacronym{osc}{OSC}{Open Sound Control}
\newacronym{posix}{POSIX}{Portable Operating System Interface}
\newacronym{pubsub}{PubSub}{Publish-subscribe message pattern}
\newacronym{pd}{Pd}{PureData}
\newacronym{raii}{RAII}{Ressource acquisition is initialization}
\newacronym{ssr}{SSR}{SoundScape Renderer}
\newacronym{tcp}{TCP}{Transmission Control Protocol}
\newacronym{tu-berlin}{TU Berlin}{Technische Universität Berlin}
\newacronym{udp}{UDP}{User Datagram Protocol}
\newacronym{vbap}{VBAP}{Vector Based Amplitude Panning}
\newacronym{wfs}{WFS}{Wave Field Synthesis}
\newacronym{xml}{XML}{Extensible Markup Language}
\newglossaryentry{ascii}{
  name={ASCII},
  description={American Standard Code for Information Interchange --- a
  character encoding standard}
}
\newglossaryentry{sclang}{
  name={sclang},
  description={Name of the SuperCollider programming language and the
  interpreter executable of the SuperCollider programming language}
}
\newglossaryentry{id}{
  name={ID},
  description={A name or number, that identifies an object},
  plural=IDs
}
\newglossaryentry{stdout}{
  name={stdout},
  description={The standard output is a stream where a program writes its
  output data to. This can be a log file or a terminal}
}
\newglossaryentry{faust}{
  name={FAUST},
  description={Functional Audio Stream is a functional programming language
  specifically designed for realtime signal processing and synthesis}
}
\newglossaryentry{quark}{
  name={Quark},
  description={Name for Classes extending the SuperCollider programming
  language, usually developed in a separate version controlled code
  repository},
  plural=Quarks
}
\newglossaryentry{supercollider}{
  name={SuperCollider},
  description={A programming language, \gls{ide} and synthesis server for
  realtime audio processing and synthesis}
}
\newglossaryentry{python}{
  name={Python},
  description={An multi-purpose, object-oriented programming language}
}
\newglossaryentry{qt4}{
  name={Qt4},
  description={Version 4 (legacy) of the cross-platform application framework
  for creating desktop applications.}
}
\newglossaryentry{qt5}{
  name={Qt5},
  description={Version 5 of the cross-platform application framework for
  creating desktop applications.}
}

\graphicspath{{../images//}}

\AtBeginDocument{\hypersetup{pdftitle = {A Networking Extension for the
    SoundScape Renderer}, pdfauthor = {David Runge}
  }
}

\begin{document}
  \pagestyle{empty}
  \begin{titlepage}
    \centering
    {\huge\bfseries A Networking Extension for the SoundScape Renderer\par}
    \vspace{1cm}
    {\scshape\Large Master Thesis\par}
    \vspace{1.5cm}
    \includegraphics[width=0.3\textwidth]{tu-berlin-logo.pdf}\par\vspace{1cm}
    Technische Universität Berlin\\
    Fakultät I - Geisteswissenschaften\\
    Fachgebiet Audiokommunikation\\
    Audiokommunikation und -technologie M.Sc.\\
    \vspace{1.5cm}
    \begin{table}[!htb]
      \begin{tabular}{l l}
      \item \textbf{Vorgelegt von}: &David Runge\\
      \item \textbf{Matrikelnummer}: &340592\\
      \item \textbf{E-Mail}: &\href{dave@sleepmap.de}{dave@sleepmap.de}\\
      \end{tabular}
    \end{table}
    \begin{table}[!htb]
      \begin{tabular}{l l}
      \item \textbf{Erstgutachter}: &Prof. Dr. Stefan Weinzierl\\
      \item \textbf{Zweitgutachter}: &Henrik von Coler\\
      \item \textbf{Datum}: &\today\\
      \end{tabular}
    \end{table}
  \end{titlepage}

  \nextpage
  \mbox{}
  \cleardoublepage
  \nextpage

  \section*{Eidesstattliche Erklärung}
  \vspace{1cm}
  Hiermit erkläre ich, dass ich die vorliegende Arbeit selbstständig und
  eigenhändig sowie ohne unerlaubte fremde Hilfe und ausschließlich unter
  Verwendung der aufgeführten Quellen und Hilfsmittel angefertigt habe.\\
  Berlin, den \today\par\\
  \vspace{2cm}
  \noindent\ldots\ldots\ldots\ldots\ldots\ldots\ldots\ldots\ldots\ldots\ldots\\
  David Runge
  \newpage

  \begin{abstract}
    Different types of software have been designed in the field of spatial
    audio reproduction to cope with the needs of several rendering algorithms,
    conceived over the last century. Their feasibility and reproducability in a
    real environment have been evaluated and tested and their implementations
    extended, or improved accordingly.\\
    Today, spatial audio rendering software is often a single-purpose
    application, that attends to the needs of a specific setup in place.
    Unfortunately, to some of these solutions, no further work is applied,
    leaving them unsupported at some point in time. The affected hardware
    setups range from those for large scale Wave Field Synthesis to smaller
    ones, found in scientific research, involving Dynamic Binaural Synthesis.
    All of them are in need of extensively supported and reliable software,
    that can even be used, once the system's hardware has to be upgraded, due
    to old age, or requires new software, that can not be build on older
    Operating Systems.\\
    In the following work, a set of current free and open-source realtime
    spatial audio renderers, actively used in scientific and artistic contexts,
    is evaluated for usability, applicability and realtime context: sWONDER,
    HoaLibrary (for PureData), 3Dj (for SuperCollider), WFSCollider and the
    SoundScape Renderer. The latter --- in contrast to the other candidates, a
    multi-purpose renderer --- is chosen for a rewrite of its network based
    messaging system, as it implements many different rendering algorithms,
    while sitll being actively maintained.\\
    Its new functionality allows it to be used in other environments: While
    rendering the same virtual audio scene, large scale multi-loudspeaker
    setups, in which several instances work collectively, are possible as well
    as networked individual setups.\\
    The SoundScape Renderer's new networking extension, along with a
    client-server architecture, its messaging system with tests and workflow
    examples is elaborated. New setup possibilities are contrasted with the
    automation available through the network interface currently in use.\\
    Closing, an outlook on future work with the help of the new networking
    extension and general improvement suggestions for the SoundScape Renderer
    are discussed.
  \end{abstract}
  \newpage

  \renewcommand{\abstractname}{Zusammenfassung}
  \begin{abstract}
    Unterschiedlichste Software wurde im Feld der Klangverräumlichung
    geschrieben, um die vielzähligen Algorithmen des letzten Jahrhunderts
    abbilden zu können. Die Praktikabilität und Reproduzierbarkeit dieser
    Algorithmen in einer realen Umgebung wurde evaluiert und getetest und ihre
    Implementationen entsprechend erweitert, oder verbessert.\\
    Heutige Klangverräumlichungssoftware ist häufig eine Einzweck-Software,
    welche einem spezifischen Aufbau dient. Unglücklicherweise werden einige
    dieser Lösungen nicht weiter entwickelt, was zukünftig zwangsläufig zu
    ihrer Unbrauchbarkeit führt. Die betroffenen Hardwareaufbauten reichen von
    großen Anlagen, wie jene für die Umsetzung einer Wellenfeldsynthese zu
    kleineren, in der wisschenaftlichen Forschung auffindbare, die Dynamische
    Binauralsynthese anwenden. All jene Hardwareumgebungen benötigen Software,
    die weitläufig unterstützt ist und verlässlich arbeitet, auch wenn die
    Betriebssysteme, auf denen sie genutzt werden, auf einen neuen Stand
    gebracht werden müssen (Umstände, die durch veraltete oder defekte Hardware
    oder der zwangsläufigen Nutzung neuerer Software, die nicht auf alten
    Betriebssystemen kompilierbar ist, hervorgerufen werden können).\\
    In der nachfolgenden Arbeit wird die folgende Sammlung aus freien und
    quelloffenen Anwendungen zur Echtzeit-Klangverräumlichung, die derzeit
    aktiv in künstlerischen und wisschenschaftlichen Umgebungen Anwendung
    finden, anhand ihrer Nutzungsmöglichkeiten, ihrer Anwendbarkeitkeit und
    ihrer Umgebung evaluiert: sWONDER, HoaLibrary (für PureData), 3Dj (für
    SuperCollider), WFSCollider und der SoundScape Renderer. Der letztere ---
    im Gegensatz zu den anderen Kandidaten ein Mehrzweck-Werkzeug --- wurde
    für eine Neuausarbeitung seiner Netzwerkfähigkeit gewählt, aufgrund seiner
    vielzähligen Implementationen von Verräumlichungs-Algorithmen und seiner
    noch betriebenen Weiterentwicklung.\\
    Seine neue Funktionalität erlaubt es ihm in anderen Umgebung eingesetzt zu
    werden: Während die gleiche virtuelle Audio-Szene verräumlicht wird, ist
    die Anwendung nun gleichermaßen einsetzbar in einer Großanlage, in der
    viele Instanzen zusammen arbeiten und vernetzte, individuellen Aufbauten,
    die getrennt voneinander arbeiten.\\
    Die neuen Netzwerkmöglichkeiten der Anwendung, zusammen mit einer
    Client-Server Architektur, sowie seines Nachrichtensystem, anhand von
    Arbeitsabläufen und Tests, werden ausgearbeitet. Neue Möglichkeiten der
    Vernetzung werden der derzeitig verwendeten Netzwerkschnittstelle gegenüber
    gestellt.\\
    Abschließend werden ein Ausblick auf weitere Arbeit anhand der neuen
    Netzwerkschnittstelle und generelle Verbesserungsvorschläge für den
    SoundScape Renderer diskutiert.


  \end{abstract}
  \newpage

  \section*{Acknowlegdements}
  I would like to thank my supervisors Prof. Dr. Stefan Weinzierl and Henrik
  von Coler for their tremendous efforts in making this work happen.\\
  Special thanks to Henrik for pushing in the right moments.\\
  \\
  I would also like to thank Marc Voigt for bouncing off ideas every once in a
  while and helping in realizing a test setup. Hopefully this work will be
  useful for the setups you have to maintain in the future.\\
  Thanks to Matthias Geier for being relentless and yet supportive.\\
  \\
  Special thanks to my family for their loving support over all of this time.\\
  Thank you Nanni, for taking my mind off.\\
  \\
  Thanks to Sabine and Peter for reading through all of this nonsense.\\
  \\
  To all, whom I forgot: You know, who you are. I could not have done it
  without you.

  \newpage


  \pagestyle{empty}
  {
    \setcounter{tocdepth}{4}
    \renewcommand{\thispagestyle}[1]{}
    \tableofcontents
    \newpage
  }

  \pagestyle{headings}
  \setcounter{page}{1}
  \pagenumbering{arabic}
  \renewcommand{\theFancyVerbLine}
    {\textcolor{gray}{\scriptsize\arabic{FancyVerbLine}}}
  \RecustomVerbatimEnvironment{Verbatim}{Verbatim}{xleftmargin=5mm}

  \section{Introduction}
    \label{sec:introduction}
    From the early days of stereo audio reproduction onwards, different kinds
    of spatial audio reproduction techniques have been developed and
    established. Leading from plain stereophony to three-dimensional,
    multi-channel setups.
    Their applications range from research to artistic and conventionally
    commercial fields, such as cinema and home entertainment.\\

    With the rise of dynamic two and three-dimensional rendering algorithms
    (see~\ref{subsec:spatialaudiorenderingalgorithms}), the need for
    specialized software, implementing them, grew. Opposed to encoding of
    spatial information of sources in only two channels --- static in the case
    of commercially produced audio for radio and film --- encoding for massive
    multi loudspeaker systems would not be feasible, when done statically, or
    not applicable in the case of dynamic setups, reacting to user input in
    realtime.\\

    Generally, software can be divided into two groups: One, which is developed
    freely and open-source (see~\ref{subsec:whyfreesoftwarematters}) and the
    other, which is developed non-free, also known as closed-source. Both types
    are usually covered by a license, explicitely describing the rights of its
    end-user. In scientific research, free software is usually preferred, due
    to its ease of reproducability (the sources can be retrieved, modified and
    used, free of charge).\\

    Spatial audio rendering software exists for different \glspl{os}, in
    several stages of completeness and feature richness, while covered by free
    and non-free licenses. The following work will focus on free software, used
    in scientific research and artistic contexts. Several spatial audio
    renderers, currently in use, were evaluated and compared, of which one was
    chosen for extension.\\

    The \gls{ssr} is a multi-purpose spatial audio renderer, developed at the
    \gls{tu-berlin}. To improve its usability and networking capabilities, a
    new networking extension was developed, facilitating a \gls{osc} based
    messaging system, that incorporates features for distributed processing in
    massive multi-loudspeaker setups.\\

    \cleardoublepage
  \section{Free and open-source spatial audio renderers}
    \label{sec:freespatialaudiorenderers}
    \gls{jack} \citep{website:jackaudio2016} is a low-latency audio server,
    that allows for software using its environment to connect their in- and
    outputs with any other application using it. It is licensed under the
    \gls{gpl} and can be built for various \glspl{os} (e.g. Linux, Mac OSX,
    Windows). As of today, a plethora of applications exist, that extend
    \gls{jack}'s functionality graphically, or make use of it musically and
    productively. Due to a large set of audio drivers it can use (i.e.
    \gls{alsa}, coreaudio, freebob, oss sun and portaudio) and its general
    availability, the audio server has become the de-facto standard for free
    and open-source, production ready applications on all major \glspl{os}.\\
    To date there exist five (known of) free and open-source spatial audio
    renderers, which are all \gls{jack} clients:

    \begin{itemize}
      \item sWONDER \citep{baalman2007}, developed at the \gls{tu-berlin},
        Germany
      \item WFSCollider \citep{website:wfscollider}, developed by the Game Of
        Life Foundation \citep{website:gameoflife}, The Hague, Netherlands
      \item HoaLibrary for \gls{pd} \citep{github:hoalibraryforpd} developed at
        the \gls{cicm}, Paris, France
      \item 3Dj for \gls{supercollider} \citep{thesis:perezlopez3dj2014},
        developed at the Universitat Pompeu Fabra, Barcelona
      \item \gls{ssr} \citep{website:ssr2016}, developed at the Quality \&
        Usability Lab, Telekom Innovation Laboratories, \gls{tu-berlin} and
        Institut für Nachrichtentechnik, Universität Rostock and Division of
        Applied Acoustics, Chalmers University of Technology
    \end{itemize}

    Different concepts and contexts apply to all of the renderers, which are
    briefly explained in the following sections, prefixed by a section about
    spatial audio rendering algorithms and followed by one about free software
    and its pitfalls.

    \subsection{Spatial audio rendering algorithms}
      \label{subsec:spatialaudiorenderingalgorithms}
      In the following subsections several spatial audio rendering
      algorithms are introduced briefly.
      As they serve as a mere introduction, they were merged where applicable.

      \subsubsection{Dynamic Binaural Synthesis and Dynamic Binaural Room
      Synthesis}
        \label{subsubsec:binaural}
        \gls{bs} describes a stereophonic audio reproduction, in which ---
        usually using headphones --- acoustic signals are recreated at the ears
        of the listener.\\
        For humans, sound source localization and distance estimation takes
        place according to auditory cues from each ear. The signals perceived
        by inner and outer ear are correlated by the brain, to account for
        locations in all three dimensions and their distances from the
        listener.\\
        The differences between the cues perceived by each ear can be measured
        as a \gls{hrir} for every human individually (as it is dependant on
        physiology). Its Fourier transform, the \gls{hrtf}, can then be used
        to modify audio signals to become a directional audio source, perceived
        as in free field conditions.\\
        \gls{brs} is a special form of \gls{bs}, in which \glspl{brir}, encode
        all of the virtual source's characteristics, such as position,
        alongside the room's acoustic characteristics.
        This way, recordings from real rooms can be reproduced authentically.\\
        \glspl{hrir} and \glspl{brir} are by default applied seperately for
        each ear. Therefore, if a resolution of 1\textdegree~is desired, it can
        be achieved by a set of 720 impulse responses, that are applied to the
        source with the help of a head tracker, measuring the azimuth of the
        listener towards it.

      \subsubsection{(Higher Order) Ambisonics Amplitude Panning and
      Near-field-compensated Higher Order Ambisonics}
        \label{subsubsec:aapandnfchoa}
        \gls{aap} and \gls{hoa} are spatial rendering algorithms, that
        reproduce audio on multi-speaker setups. Those are usually circular or
        spherical.\\
        Depending on a loudspeaker's position in the setup, relative to the
        spheres's center (the listening area or \textit{sweet spot} \citep[Fig.
        1.4]{HagenWierstorf2014}), a linear combination of all loudspeakers is
        used to achieve a localized representation of a virtual sound source.\\
        The relatively small listening area can be extended by using additional
        sets of loudspeakers, which in turn lead to more spatial aliasing.\\
        Due to the perceptebility of localization cues, mentioned
        in~\ref{subsubsec:binaural}, it is required to apply spatial
        equalization for the rendered sources, to account for differences in
        low- and high-frequency localization capabilities of the human ear.\\
        For ambisonics, plane-wave sources are assumed, which means, their
        distance is infinite. Due to the proximity effect, this leads to a bass
        boost in the listening area. \gls{nfc-hoa} accounts for this by a set
        of driving functions, applying a per speaker near-field compensation.\\

      \subsubsection{Vector Based Amplitude Panning}
        \label{subsubsec:vbap}
        \gls{vbap} is another rendering method for multiple loudspeakers. Up to
        three loudspeakers are used to reproduce a virtual sound source in a
        three-dimensional setup, while only two are needed in a horizontal one.\\
        It enables for “virtual source positioning in a three-dimensional sound
        field formed by loudspeakers in an arbitrary three-dimensional
        placement“, while being ”computationally efficient and accurate“
        \citep[p. 464]{VillePulkki1997-05-31}.\\
        However, according to \citet{MatthiasGeier1888} ”\gls{vbap} has a very
        small sweet spot, out of which localization of sources is distorted
        towards the nearest active loudspeaker“ and ”works best for circular
        setups“.

      \subsubsection{Wave Field Synthesis}
        \label{subsubsec:wavefieldsynthesis}
        \gls{wfs} is a spatial audio rendering technique, which is based on the
        Huygens-Fresnel principle. It states that any wave front can be
        synthesized by the superposition of elementary spherical waves.\\
        Setups mainly focus on horizontal, preferably spatially discrete,
        speaker arrays of rectangular or circular shape as the human hearing is
        most capable to localize acoustic sources in this this plane.\\
        According to \citet{HagenWierstorf1899}, localization is accurately and
        evenly distributed in the listening area with loudspeaker spacings of
        up to 40cm.\\
        Although \gls{wfs} does not suffer from a pronounced sweet spot and
        spatial aliasing is distributed over a relatively large listening area,
        compared to e.g.\ \gls{nfc-hoa}, the spatial sampling artifacts may
        still be perceived as coloration of the sound field, which can be
        improved by prefiltering especially high-frequency content
        \citep{HelmutWittek1900}.\\
        Due to the relatively high amount of loudspeakers (and thereby
        computing power to calculate as many audio channels) needed for a
        medium to large-scale setup, \gls{wfs} is not yet very widely
        distributed.

    \subsection{sWONDER}
      \label{subsec:swonder}
      sWONDER \citep{baalman2007} consists of a set of C++ applications that
      provide \gls{bs} and \gls{wfs} rendering. In 2007 it was specifically
      redesigned \citep{baalmanetal2007} to cope with large scale \gls{wfs}
      setups in which several (computer) nodes, providing several speakers
      each, drive a system together.\\
      In these setups each node receives all available audio streams (which
      represent one virtual audio source respectively) redundantly and a master
      application signals which node is responsible for rendering what source
      on which speaker.\\
      It uses \gls{osc} for messaging between its components and for setting
      its controls. Additionally, it can be controlled through a \gls{gui},
      that was specifically designed for it.\\
      Sound sources can be moved dynamically, or according to a \gls{xml} based
      score.\\
      For example sWONDER has been in use for the medium and large scale
      \gls{wfs} systems in the Electronic Music Studio
      \citep{website:tu-electronic_studio} and lecture hall H0103
      \citep{website:tu-wfs} at \gls{tu-berlin} and a medium
      scale system at the Wave Field Synthesis Lab at HAW in Hamburg
      \citep{Fohl2013}.\\
      The included convolution engine fWonder is applied in “Assessing the
      Authenticity of Individual Dynamic Binaural Synthesis” \citep[pp.
      223-246]{lindau2014}.\\
      Unfortunately, the spatial audio renderer has not been actively
      maintained for several years. Hence it is limited to its two rendering
      algorithms and has many bugs, that are not likely to get fixed in the
      future.\\

    \subsection{HoaLibrary (PureData extension)}
      \label{subsec:hoalibrary_puredata_extension}
      The HoaLibrary is “a collection of C++ and \gls{faust} classes and
      objects for Max, PureData and VST destined to high order ambisonics sound
      reproduction” \citep{website:hoalibrary}. By using its \gls{pd}
      extension, it enables for \gls{hoa} reproduction, while harnessing the
      rich feature set of the audio programming language still enables for
      implementing other forms of spatial rendering alongside the HoaLibrary.\\
      \gls{pd} is \gls{osc} capable with the help of extensions, such as
      \textit{mrpeach}\footnote{ \href{https://puredata.info/downloads/mrpeach}
      {https://puredata.info/downloads/mrpeach}} or \textit{IEMnet}\footnote{
        \href{https://puredata.info/downloads/iemnet}
      {https://puredata.info/downloads/iemnet}}.\\

    \subsection{3Dj (SuperCollider Quark)}
      \label{subsec:3dj_supercollider_quark}
      3Dj is a \gls{supercollider} \gls{quark} conceived in the course of a Master
      Thesis at Universitat Pompeu Fabra, Barcelona
      \citep{thesis:perezlopez3dj2014} for interactive performance live
      spatialization purposes. It implements \gls{hoa} and \gls{vbap} rendering
      \citep[p 45]{thesis:perezlopez3dj2014} and uses a specific scene format
      \citep[pp. 45--46]{thesis:perezlopez3dj2014} to allow sound sources to
      have static, linear, random, brownian, simple harmonic and orbital
      motion.\\
      Due to being a language extension to \gls{sclang}, 3Dj can be used in
      conjunction with other spatial rendering algorithms provided by
      \gls{supercollider} or any of its \glspl{quark}.\\
      \gls{supercollider} is \gls{osc} enabled by default, which renders 3Dj a
      dynamically accessible solution.

    \subsection{WFSCollider}
      \label{subsec:wfscollider}
      WFSCollider was built on top of
      \href{https://supercollider.github.io}{SuperCollider} 3.5
      \citep{website:supercollider} and as its name suggests, it is an
      application for \gls{wfs} reproduction. It “allows soundfiles, live input
      and synthesis processes to be placed in a score editor where start times,
      and durations can be set and trajectories or positions assigned to each
      event. It also allows realtime changement of parameters and on the fly
      starting and stopping of events via \gls{gui} or \gls{osc} control. Each
      event can be composed of varous objects (“units”) in a processing chain“
      \citep{website:wfscollider}. According to its current manual, it is
      also capable of using a \gls{vbap} renderer for other multi-speaker
      setups \citep[p. 8]{manual:wfscollider}.\\
      ”WFSCollider is the driving software of the Wave Field Synthesis system
      of the Game Of Life Foundation“ \citep{website:gameoflife}. In
      multi-computer setups, it can synchronize the involved processes and a
      dynamic latency can be introduced to account for high network throughput
      \citep[p. 22]{manual:wfscollider}. By nature WFSCollider is \gls{osc}
      capable and extendable by what \gls{sclang} has to offer. Its scores are
      saved as \gls{supercollider} code, as well.\\
      It is currently only tested on MacOSX and is based upon a several year
      old version of \href{https://supercollider.github.io}{SuperCollider}.

    \subsection{SoundScape Renderer}
      \label{subsec:soundscaperenderer}
      The \gls{ssr}, written in C++, is a multi-purpose spatial audio renderer,
      that runs on Linux and MacOSX\@. Based on its underlying \gls{apf}
      \citep{MatthiasGeierTorbenHohn1890}, it is able to use \gls{bs},
      \gls{brs}, \gls{aap}, \gls{wfs}, \gls{nfc-hoa} and \gls{vbap}. However,
      all rendering algorithms with potentially orthogonal sound fields, are
      currently only available in 2D \citep{MatthiasGeier1888}.\\
      It can be used with a \gls{qt4} based \gls{gui} or headless (without
      one), depicting the virtual sources, their volumes and positions. If a
      loudspeaker based renderer is chosen, the \gls{gui} also illustrates
      which speakers are currently used for rendering a selected source.\\
      The \gls{bs} and \gls{brs} renderers are frequently used in scientific
      research, such as \citep{DavidAckermann1895} or
      \citep{DmitryGrigoriev1896}. The \gls{wfs} renderer has been improved by
      the work of several research papers, dealing with enhancements of spatial
      aliasing, active listening room and loudspeaker compensation and active
      noise control \citep{SaschaSporsRudolfRabensteinJensAhrens1822} and
      analyzing and pre-equalizing in 2.5-dimensional \gls{wfs}
      \citep{SaschaSporsJensAhrens1821}.\\
      The \gls{ssr} uses \gls{xml} based configuration files for reproduction
      (i.e.\ how something is played back) and scene (i.e.\ what is played
      back). The \gls{asdf} however is not (yet) able to represent dynamic
      setups.\\
      The application can be controlled through a \gls{tcp}/\gls{ip} socket.
      \gls{osc} functionality can only be achieved using the capapilities of
      other applications such as \gls{pd} \citep{website:puredata2016} in
      combination with it.\\
      Unlike \nameref{subsec:swonder} or \nameref{subsec:wfscollider}, the
      \gls{ssr} is not able to run medium or large-scale \gls{wfs} setups, as
      it lacks the features to communicate between instances of itself on
      several computers, while these instances serve a subset of the
      available loudspeakers.

    \subsection{Why free software matters and what its pitfalls are}
      \label{subsec:whyfreesoftwarematters}
      Free software is the terminology for software published under a free
      license. Licenses, such as the \gls{gpl} are considered free, because
      they allow for anyone to copy, modify and redistribute the source code
      (under the same license).\\
      Research is a field of work, in which reproducability is very important,
      as findings need to be independently verifiable.  Scientific work is
      published and shared (sometimes also under free licenses, such as
      \gls{cc}) amongst research groups of institutions all around the world.
      In an ideal world, all scientific research would be published under a
      free documentation license, such as the \gls{fdl}, allowing access to
      anyone, free of charge.\\
      The software used in scientific institutions is unfortunately rarely free
      (e.g.\ word processing, statistics, mathmatical calculations, realtime
      audio synthesis and audio production) and additionally mostly bound to
      proprietary \glspl{os}, such as Microsoft Windows or Apple's Mac OSX,
      preventing interoperability, development and an open society.\\
      However, free software enables students and researchers to learn from the
      source code directly (if they can and want to), to modify (and improve)
      it and to share their findings. More than with proprietary software, it
      is possible to have a community develop around it, that takes care of the
      project for a long time.\\
      Free software nonetheless can not be considered superior. It is after all
      only a way of developing software and not a way to grade its efficiency
      or code quality. Additionally it has to be noted, that especially in a
      scientific context it can happen, that software is conceived by an
      institution, put to use, but later lacks the developers to drive the
      project onwards (e.g. \nameref{subsec:swonder}).
      This way a high responsibility weighs on these institutions, as they
      need to ensure further development on systems, not easily accessible to
      the public, or not feasible for home use (e.g. \gls{wfs}). This situation
      however also holds a great opportunity for coorporation.

  \cleardoublepage
  \section{Implementation}
    \label{sec:implementation}
    This section covers the implementation of a networking interface for the
    \gls{ssr} and the considerations leading to it.
    The application was chosen to be extended by a \gls{osc} based networking
    interface, because it runs on multiple \glspl{os}, offers a wide set of
    rendering algorithms (in various stages of completion) by using the
    \gls{apf} \citep{MatthiasGeierTorbenHohn1890}, is used extensively in
    scientific research, has the future possibility to run medium and large
    scale \gls{wfs} setups and was still actively maintained by its creators at
    the time of writing.\\
    Software, such as the HoaLibrary
    (see~\ref{subsec:hoalibrary_puredata_extension}) or 3Dj,
    (see~\ref{subsec:3dj_supercollider_quark}) were not considered, as they
    were too reliant on their environment (i.e.\ \gls{pd} or
    \gls{supercollider}) and only implemented a small set of spatial audio
    renderers, while sWONDER was additionally long unmaintained
    (see~\ref{subsec:swonder}) and WFSCollider bound to a non-free operating
    system (see~\ref{subsec:wfscollider}).

    \subsection{Outline}
      \label{subsec:outline}
      Initially, it was aimed to extend the \gls{ssr}'s features in the scope
      of creating a replacement for the aging sWONDER software, enabling it to
      run networked instances to drive a medium or large scale \gls{wfs} setup.
      However, the approach appeared too narrow, as the application offers many
      different rendering algorithms. A networking extension therefore would
      have to be available to all of them with an equal feature set.
      Additionally, extending a rendering framework by a networking feature,
      with the help of only one of its engines proved to be linked to a
      massive, but avoidable overhead (see~\ref{subsubsec:preliminaries}).\\
      The \gls{ssr}, being a multi-purpose spatial audio renderer, can be used
      in diverse setup scenarios (see~\ref{subsubsec:setups}). Therefore not
      only classic server-client relationships
      (see~\ref{subsubsec:remote_controlling_a_server}), but also client-only
      and local (see~\ref{subsubsec:remote_controlling_a_client}) setups had to
      be taken account of. In addition the case of medium and
      large scale loudspeaker based rendering setups and their specifics had to
      be considered (see~\ref{subsubsec:rendering_on_dedicated_speakers}).

      \subsubsection{Prelimenaries}
        \label{subsubsec:preliminaries}
        In preparation to this work, an implemention of a side-by-side
        installation to the \gls{os} currently driving the \gls{wfs} system
        setup of the Electronic Studio at \gls{tu-berlin}
        \citep{website:tu-electronic_studio} was attempted for testing
        purposes.\\
        Arch Linux \citep{website:archlinux} was installed and configured to
        run the medium scale setup. Unfortunately, the proprietary Dante
        \citep{website:audinate} driver for Linux, offered by Four Audio
        \citep{website:fouraudio}, creates non-trivial circumstances for using
        it on an up-to-date Linux kernel, due to \gls{alsa} \gls{api} changes
        not accounted for.\\
        While the current \gls{os} --- an Ubuntu \citep{website:ubuntu} Studio
        2012 \gls{lts} --- still runs well in its own parameters, its support has
        run out and it is therefore becoming harder, if not impossible, to
        build newer software on it, using newer versions of free software
        compilers.\\
        For research purposes however, it is desirable to be able to try new
        kernel and software features on a regular basis. It is essential to
        find the most stable and secure setup possible involving realtime
        enabled kernels and building new versions of (spatialisation)
        software.\\
        The hardware of the large scale setup at \gls{tu-berlin} in lecture
        hall H0104 was being updated and unusable at the time of writing. In
        the future it will however become a valuable candidate for testing of
        the sought after \gls{ssr} features, as its setup involves no Dante
        network, but is instead run by several rendering computers connected to
        \gls{madi} and \gls{adat} lightpipe enabled speaker systems.\\
        Although a \gls{wfs} setup for testing purposes was eligible, it
        generally is not needed for implementing the features described in
        the following sections and subsections, as they can be tested using two
        machines running Linux, \gls{jack} and a development version of the
        \gls{ssr}.\\

      \subsubsection{Remote controlling a server}
        \label{subsubsec:remote_controlling_a_server}
        A \gls{ssr} server instance in the notion of a medium to large scale
        reproduction setup is supposed to have a set of \textit{n}
        (pre-)assigned clients. Generally, controlling it should be possible
        through either \gls{udp} or \gls{tcp}.
        Every message sent to it should be distributed to all of its clients
        (if applicable), preferably using the same protocol used to communicate
        with the server. The messaging system should be flexible and
        scriptable.\\
        All audio inputs available to the server should be available to its
        clients as well. A server instance should be able to render audio just
        as a client would. It should be able to receive messages from its
        clients and act upon them (e.g.\ updating \gls{gui} elements).

      \subsubsection{Remote controlling clients}
        \label{subsubsec:remote_controlling_a_client}
        A \gls{ssr} client can either be local (on the same machine) or
        somewhere on the same network, as the server or application controlling
        it. It should not make a difference, if the client instance is
        controlled by a server instance or any other application, implementing
        the messaging system it uses. A client should send an update to its
        server or the application controlling it, upon receiving a valid
        message.

      \subsubsection{Rendering on dedicated speakers}
        \label{subsubsec:rendering_on_dedicated_speakers}
        In a medium or large scale setup, \textit{n} clients collectively
        render audio on \textit{l} loudspeakers, while all should be using the
        same \textit{i} inputs and each have \textit{c} hardware outputs.
        \textit{l} is preferably a multiple of \textit{c}, but definitely
        larger than \textit{c}.\\
        As the described setups usually have too many loudspeakers for only one
        machine (i.e.\ a client), a system has to be conceived, by which each
        client will only render on its specifically assigned subset of size
        \textit{c} of the \textit{n} loudspeakers.

    \subsection{Publisher/Subscriber interface}
      \label{subsec:publisher_subscriber_interface}
      The \gls{ssr} internally uses a \gls{pubsub}, which is a design pattern,
      to implement control through and over several parts of its components.\\
      In \gls{oop} \gls{pubsub} --- also called observer, listener messaging
      --- is usually comprised of a publisher class, handling the messages,
      without explicitely implementing how they will be used and a subscriber
      class, that allows for its implementations to subscribe to the messages
      provided. Filtering takes place to enable subscribers to only receive a
      certain subset of the messages.\\
      The \gls{ssr} implements a content-based filtering system, in which each
      subscriber evaluates the messages received and acts depending on its own
      constraints to implement further actions upon it.\\

      \begin{figure}[!htb]
        \centering
        \includegraphics[scale=0.6, trim = 6mm 130mm 12mm 4mm, clip]
        {ssr-publisher-with-legacy-subscribers.pdf}
        \caption{A diagram depicting a simplified version of the \gls{pubsub}
          used within the \gls{ssr}.\\
          {\color{pubsub-in}\textbf{--}} Calls from Publisher to Subscriber
          {\color{controller-in}\textbf{--}} Calls from Subscribers to
          Controller (Publisher)
        }
        \label{fig:ssr-publisher-with-legacy-subscribers}
      \end{figure}

      The abstract class \mintinline{c++}{Publisher} defines the messages
      possible to send and provides means to subscribe to them. The global
      \mintinline{c++}{Controller} class is its only implementation within the
      \gls{ssr}.\\
      The abstract class \mintinline{c++}{Subscriber} in turn defines the
      messages understood, while its implementations in
      \mintinline{c++}{RenderSubscriber}, \mintinline{c++}{Scene},
      \mintinline{c++}{OscSender} and \mintinline{c++}{NetworkSubscriber} take
      care of how they are used.\\
      This system enables a versatile messaging layout, in which components can
      call the \mintinline{c++}{Publisher} functionality in
      \mintinline{c++}{Controller}, which in turn will send out messages to all
      of its subscribers.\\
      Depending on the design of an application, \gls{pubsub} is not
      necessarily a one-way-road. As shown in
      Figure~\ref{fig:ssr-publisher-with-legacy-subscribers}, subscribers can
      also be able to call functions of the \mintinline{c++}{Publisher}, if the
      implementation permits it.\\
      In the \gls{ssr} this is possible, because each
      \mintinline{c++}{Subscriber} holds a reference to the
      \mintinline{c++}{Controller} instance and is therefore able to call its
      public functions.\\
      According to the principle of encapsulation in \gls{oop}, this type of
      functionality is handled by a separate class. In
      Figure~\ref{fig:ssr-publisher-with-all-subscribers}, the
      \mintinline{c++}{OscHandler} and \mintinline{c++}{Server} instances
      delegate calls to \mintinline{c++}{Controller} functionality to their
      utilities \mintinline{c++}{OscReceiver} and
      \mintinline{c++}{CommandParser} (respectively).

    \subsection{IP interface}
      \label{subsec:ip-interface}
      The \gls{ssr} from early on incorporated a network interface, that
      accepts specially terminated \gls{xml}-formatted strings over a \gls{tcp}
      port, called “\gls{ip} interface”. This has the benefit of reusing the
      same \gls{xml} parser code in use for scene and reproduction
      description.\\
      A downside is however, that --- from the perspective of other software
      --- it is complicated to use, as a conversion to \gls{xml} has to be
      attempted before sending a message to the \gls{ssr}. Additionally, the
      message has to be linted (error checked) before sending and parsed again,
      after receiving an answer from the application.\\
      The \gls{ip} interface achieves to offer more or less direct access to
      the \nameref{subsec:publisher_subscriber_interface}. However, it has no
      notion of a networked setup and could therefore be described as a
      two-directional message system between two destinations. With it, only
      setups with up to \textit{n} clients are possible.

      \subsubsection{OSC through PureData}
        \label{subsubsec:osc_through_puredata}
        To allow \gls{osc} communication, the \gls{ssr} incorporates a Lua
        based \gls{pd} external. It uses two externals
        (\textit{IEMnet}\footnote{
            \href{https://puredata.info/downloads/iemnet}
        {https://puredata.info/downloads/iemnet}} and
        \textit{pdlua}\footnote{\href{https://puredata.info/downloads/pdlua}
      {https://puredata.info/downloads/pdlua}}) alongside a Lua library for
      parsing and creating \gls{xml}
      (\textit{SLAXML}\footnote{\href{https://github.com/Phrogz/SLAXML}
      {https://github.com/Phrogz/SLAXML}}).

      \subsubsection{Sending and receiving}
        \label{subsubsec:sending_and_receiving}
        As mentioned in section~\ref{subsec:publisher_subscriber_interface},
        the \mintinline{c++}{NetworkSubscriber} class (part of the \gls{ip}
        interface) implements the \mintinline{c++}{Subscriber} interface. This
        means: The network interface subscribes to the messages the
        \mintinline{c++}{Publisher} (the \mintinline{c++}{Controller} instance)
        has to offer. Every time a function of the \gls{ssr}'s
        \mintinline{c++}{Controller} instance, that was inherited from
        \mintinline{c++}{Publisher}, is called, it will issue the call on all
        of its subscribers, too. Every message, available to the
        \gls{ssr}'s~\ref{subsec:ip-interface} is therefore directly bound to
        its \gls{pubsub} interface's set of functions.\\

  \cleardoublepage
    \subsection{Open Sound Control interface}
      \label{subsec:osc-interface}
      The networking interface conceived in the course of this work was
      developed in several branches, using the git version control system
      (written by Linus Torvalds, now maintained by Junio
      Hamano\footnote{\href{https://git-scm.com}{https://git-scm.com}}),
      publicly on Github\footnote{\href{https://github.com/dvzrv/ssr}
      {https://github.com/dvzrv/ssr}}. Internally the liblo library (further
      explained in~\ref{subsubsec:liblo}) was harnessed to implement \gls{osc}
      functionality (see~\ref{subsubsec:open-sound-control}) for the
      \gls{ssr}.\\
      After initial conversations with the current maintainer Matthias Geier
      through the project's Github issue
      tracker\footnote{ \href{https://github.com/soundscaperenderer/ssr}
      {https://github.com/soundscaperenderer/ssr}}, different ideas were worked
      out to achieve a broad solution to the server-client and client-only
      setups and to get a better understanding of the underlying design.
      Initial attempts, such as the mapping of a networking setup in the scene
      description\footnote{
      \href{https://github.com/dvzrv/ssr/tree/distributed\_reproduction}
      {https://github.com/dvzrv/ssr/tree/distributed\_reproduction}}
      proved too restrictive though, as it would allow the networking
      functionality only to renderers, that use loudspeakers and mix scene
      description with networking description.\\
      A nearly configuration-less approach, based on subscribing clients on
      sending poll messages to them proved more open (in the sense that it can
      be interfaced with by any \gls{osc}-capable application or programming
      language) and have less configuration overhead. With it, a diverse set of
      setups can be achieved (further described in~\ref{subsubsec:setups}),
      which at the same time remain dynamically configurable (using a plethora
      of \gls{osc} implementations) and debuggable using tests (further
      explored in~\ref{subsec:automated_tests}).

      \begin{figure}[!htb]
        \centering
        \includegraphics[scale=0.6, trim = 6mm 91mm 12mm 10mm, clip]
        {ssr-publisher-with-all-subscribers.pdf}
        \caption{A diagram depicting a simplified version of the \gls{pubsub}
          used within the \gls{ssr}.\\
          {\color{pubsub-in}\textbf{--}} Calls from \mintinline{c++}{Publisher}
          to \mintinline{c++}{Subscriber} {\color{controller-in}\textbf{--}}
          Calls from \mintinline{c++}{Subscribers} to
          \mintinline{c++}{Controller} (\mintinline{c++}{Publisher})
        }
        \label{fig:ssr-publisher-with-all-subscribers}
      \end{figure}

      The main implementations of the interface, further described in the
      following subsections, can be found in the classes
      \mintinline{c++}{OscHandler} (handling the \gls{osc} server),
      \mintinline{c++}{OscReceiver} (handling incoming \gls{osc} messages and
      acting upon them in the context of the \gls{ssr} instance) and
      \mintinline{c++}{OscSender} (responsible for reacting to calls from the
      \textbf{\nameref{subsec:publisher_subscriber_interface}} and sending of
      \gls{osc} messages to clients and server).\\
      The class \mintinline{c++}{OscClient} implements the representation of a
      client (or server) to the message interface. It holds information about
      the client's address and port, along with its
      \mintinline{c++}{MessageLevel} (a concept elaborated
      in~\ref{subsubsec:message_levels}) and its alive counter (used to check,
      whether a given client is still available on the network).\\
      As shown in Figure~\ref{fig:ssr-publisher-with-all-subscribers}, the
      \mintinline{c++}{OscSender} is another implementation of the
      \mintinline{c++}{Subscriber} interface. This way, every call made through
      the \mintinline{c++}{Publisher} (i.e.\ the \mintinline{c++}{Controller}),
      will be made on the corresponding function in \mintinline{c++}{OscSender}
      as well. With \mintinline{c++}{OscReceiver} the \gls{osc} interface has
      direct access to the \mintinline{c++}{Controller} and can make calls to
      it, on receiving a message.\\
      In its implementation approach the \gls{osc} interface follows that of
      the \textbf{\nameref{subsec:ip-interface}}. However, it expands in
      creating a client-server architecture, controlled by
      \textbf{\nameref{subsubsec:message_levels}}, using a unified
      \textbf{\nameref{subsubsec:message_interface}}.\\
      \gls{ssr} client instances only evaluate messages of server instances
      they are subscribed to. Server instances only evaluate messages of client
      instances, that are subscribed to them.

      \subsubsection{Open Sound Control}
        \label{subsubsec:open-sound-control}
        \gls{osc} is an “open, transport-independent, message-based protocol
        developed for communication among computers, sound synthesizers, and
        other multimedia devices” \citep{website:oscv1.0} developed at the
        \gls{cnmat}.  Its 1.0 specification was published by Matthew Wright in
        2002 \citep{website:oscv1.0} and the protocol has found widespread
        implementations (as libraries) in several programming languages and
        through that many use-cases in free and closed audio and video related
        applications (e.g. Ardour \citep{website:ardour}, Max/MSP
          \citep{website:cycling74}, \gls{supercollider}
        \citep{website:supercollider}) since then.\\ \gls{osc}'s syntax is
        defined by several parts, which are discussed briefly in this
        section.\\

        \begin{itemize}
          \item Atomic data types, which are also reflected in type tags (see
            Table~\ref{tab:ssr-osc-data-type-acronyms} for details)
          \item Address patterns (a \gls{osc}-string starting with a “/”)
          \item Type tag string (a string, beginning with a “,”, holding a set
            of type tags, describing a collection of atomic data types)
          \item Arguments, a set of binary representations of each argument
          \item Messages, consisting (in sequence) of an address pattern, a
            type tag string and \textit{n} \gls{osc} arguments.
          \item Bundles, consisting of a set of Messages.
          \item Packets, the unit of transmission (sent over \gls{udp} or
            \gls{tcp}), consisting of a message or a bundle.
        \end{itemize}

        According to the specification, applications sending \gls{osc} packets
        are considered a client and the ones receiving packets a server.
        Applications can therefore be client and server at the same time.\\

        \begin{table}[!htb]
          \renewcommand{\arraystretch}{1.2}
          \scriptsize
          \centering
          \rowcolors{2}{table-background-one}{table-background-two}
          \begin{tabular}{ p{2cm} | p{8cm} }
            \textbf{\gls{osc} type tag} & \textbf{Type} \\
            \hline
            \texttt{i} & \mintinline{c++}{int32} \\
            \texttt{f} & \mintinline{c++}{float32} \\
            \texttt{s} & \gls{osc}-string \\
            \texttt{b} & \gls{osc}-blob \\
            \hline
            \hline
            \texttt{h} & 64 bit big-endian two's complement integer\\
            \texttt{t} & \gls{osc}-timetag\\
            \texttt{d} & 64 bit (“double”) IEEE 754 floating point number\\
            \texttt{S} & Alternate type represented as an \gls{osc}-string (for
            example, for systems that differentiate “symbols” from “strings”)\\
            \texttt{c} & An \gls{ascii} character, sent as 32 bits\\
            \texttt{r} & 32 bit RGBA color\\
            \texttt{m} & 4 byte \gls{midi} message. Bytes from MSB to LSB are:
            port id, status byte, data1, data2\\
            \texttt{T} & True. No bytes are allocated in the argument data.\\
            \texttt{F} & False. No bytes are allocated in the argument data.\\
            \texttt{N} & Nil. No bytes are allocated in the argument data.\\
            \texttt{I} & Infinitum. No bytes are allocated in the argument
            data.\\
            \texttt{[} & Indicates the beginning of an array. The tags
              following are for data in the Array until a close brace tag is
              reached.\\
            \texttt{]} & Indicates the end of an array.\\
          \end{tabular}
          \caption{Acronyms (type tags) for atomic data types, used in
          \gls{osc} messages and bundles \citep{website:oscv1.0}.}\\
          The first four types define the standard \gls{osc} type tags, which
          should be understood by all implementations. The remaining are
          non-standard types, that are implemented by most (e.g.
          \nameref{subsubsec:liblo} implements all but array and RGBA color
          type).
          \label{tab:ssr-osc-data-type-acronyms}
        \end{table}

        As shown in Table~\ref{tab:ssr-osc-data-type-acronyms}, only
        \mintinline{c++}{int32}, \mintinline{c++}{float32}, \gls{osc}-string
        and \gls{osc}-blob are considered standardized. Most of the remaining
        non-standard types are however implemented and used by many different
        clients.  For implementing the \gls{ssr} \gls{osc} interface, described
        in the following subsections, it was necessary to use the
        non-standard types \textit{True} and \textit{False} alongside the
        standard-types.

      \subsubsection{liblo}
        \label{subsubsec:liblo}
        Liblo \citep{website:liblo2017} is an implementation of the \gls{osc}
        protocol for \gls{posix} systems. It was initially developed by Steve
        Harris and is now actively maintained by Stephen Sinclair.\\
        The library, written in C, offers a C++ abstraction layer and is
        released under the \gls{lgpl} v2.1 or greater. Additionally, there are
        wrappers for the Perl and \gls{python} programming languages.\\
        Due to its long standing availability and usage in many small and
        large-scale software projects, alongside its fairly straight forward
        implementability, it was chosen as the candidate for establishing a
        \gls{osc} interface for the \gls{ssr}.\\
        At the time of writing liblo's lastet stable release (0.28) was issued
        on 27th January 2014. Many changes and improvements have been applied
        to the codebase since then. One of them is the implementation of a
        \mintinline{c++}{ServerThread} for the C++ abstraction layer, which
        runs a \mintinline{c++}{Server} instance on a separate thread
        automatically.\\
        In programming, threads are a way to implement simultaneous and/ or
        asynchroneous execution of code. The liblo \mintinline{c++}{Server}
        class, at the core of the C++ side of the library, is responsible for
        assigning a network port to listen to for incoming messages, listening
        for messages, executing code on their arrival (i.e.\ callback handling)
        and sending messages to clients. Many applications, facilitating
        liblo, use \gls{osc} only as a messaging system. This usually means,
        that such an application itself is not single-purpose and is busy
        computing something else most of the time. Therefore it makes sense to
        run a Server instance on a separate background thread, to not interfere
        with the executional flow of the rest of the program.\\
        The \mintinline{c++}{ServerThread} class is able to free its ressources
        upon going ot of scope (i.e.\ their ressources are not used by any
        object or function anymore), known as \gls{raii}. For this reason, the
        latest development version, instead of the current stable version of
        liblo, was chosen for the implementation.

      \subsubsection{Starting the SSR}
        \label{subsubsec:starting-the-ssr}
        The \gls{ssr} can be started with a rendering engine preselected (an
        executable postfixed by the supported rendering algorithm is provided
        by the software --- e.g. \textbf{ssr-wfs}), or by selecting one through
        the configuration file, when using the standard executable named
        \textbf{ssr}. This way \gls{aap}, \gls{bs}, \gls{brs}, generic,
        \gls{nfc-hoa}, \gls{vbap}, \gls{wfs} renderers become available.\\
        Additional features can be activated with the help of several flags to
        the executables. The customized ones, belonging to the \gls{osc}
        interface will be discussed in the following subsections. More
        information on the interplay between \gls{osc} messaging and the
        \nameref{subsec:publisher_subscriber_interface} can be found in
        \textbf{\nameref{subsubsec:message_interface}}.

        \paragraph{Client instance}
          \label{para:client-instance}
          By default the \gls{ssr} is started as an \gls{osc} client on network
          port 50001 and only allows using ephemeral ports (in the range
          49152--65535), suggested by the \gls{iana} according to
          \citet{rfc6335}. As shown in
          Listing~\ref{lst:ssr-binaural-client-start}, it is possible to use a
          different port, by defining it with the help of the \textbf{-p}
          flag.\\

          \begin{listing}[!htb]
            \begin{mdframed}
              \begin{minted}[fontsize=\footnotesize]{shell}
ssr-binaural -p “50002”
              \end{minted}
            \end{mdframed}
            \caption{Starting the \gls{ssr} using the \gls{bs} renderer as an
            \gls{osc} client (default) on the non-standard port 50002.}
            \label{lst:ssr-binaural-client-start}
          \end{listing}

          Once started, the client instance waits to receive a poll message
          from a server instance (or an application, mimicking one), upon which
          it will subscribe to it. Only then it is possible for the server
          application to control the client instance to the full extent via
          \gls{osc}.

        \paragraph{Server instance}
          \label{para:server-instance}
          With the help of the \textbf{-N} flag, it is possible to start the
          \gls{ssr} as an \gls{osc} server. For additional future purposes of
          the flag, such as a \gls{gui} only mode,
          see~\ref{subsubsec:non-renderer}. In
          Listing~\ref{lst:ssr-binaural-server-start} additionally, flag
          \textbf{-C} is used to specify an initial client \gls{ip} and its
          port (the flag actually accepts a comma-separated list of
          \gls{ip}-port pairs).\\
          The \textbf{-p} flag, for setting a specific port is also available,
          when starting a server instance.

          \begin{listing}[!htb]
            \begin{mdframed}
              \begin{minted}[fontsize=\footnotesize]{shell}
ssr-aap -N “server” -C “127.0.0.1:50002”
              \end{minted}
            \end{mdframed}
            \caption{Starting the \gls{ssr} using the \gls{aap} renderer as an
              \gls{osc} server, with an initial client on localhost, port 50002
            provided. }
            \label{lst:ssr-binaural-server-start}
          \end{listing}

          When the server instance starts, it instantly starts sending out poll
          messages to all of its active clients. Clients provided by the
          \textbf{-C} flag are considered instantly active.\\
          Additionally, it is possible for clients (\gls{ssr} client instances,
          or \gls{osc} capable applications) to subscribe to the server
          instance, or be subscribed to it by another client, using a message
          level system further explained in~\ref{subsubsec:message_levels}.
          Every valid \gls{osc} message sent to the server instance will be
          delegated to all of its clients upon evaluation, again according to
          the aforementioned message level system.\\
          If a client instance has not answered the sent out poll message of a
          server 10 times, it is considered to be unavailable and will be
          deactivated. No messages will be sent to it anymore, until the client
          subscribes/ is subscribed again.

        \paragraph{Verbosity}
          \label{para:verbosity}
          The \gls{ssr} can be started with several levels of verbosity. These
          are accessed by using the flag \textbf{-v}, up to three times (i.e.
          \textbf{-vvv}).\\
          The higher the level of verbosity, the more messages will be printed
          by the application. This especially applies to the \gls{osc}
          interface part of the \gls{ssr}, as most incoming and outgoing
          messages will be printed to \gls{stdout} at a level of \textbf{-vv}.
          At a level of \textbf{-vvv}, additionally all incoming and outgoing
          messages, that are issued in very short intervals per default
          (see~\ref{subsubsec:message_levels} for details) will be printed.

      \subsubsection{Setups}
        \label{subsubsec:setups}
        The \gls{ssr} offers the possibility for many different \gls{osc}
        enabled client-server and client-only setups. They will be explained in
        the following subsections.\\
        All examples provide audio input via a \gls{jack} client, which can be
        local (on each client's or server's host computer) or provided through
        external audio inputs from another host computer (e.g.\ through
        \gls{adat} or \gls{madi}). However, this is not mandatory, as the
        \gls{ssr} is capable of playing back audio files directly.\\
        The differences between server and client messaging is further
        elaborated in \textbf{\nameref{subsubsec:message_interface}}.\\
        A special networked setup, in which the server instance is not
        rendering any audio, is discussed in
        \textbf{\nameref{subsubsec:non-renderer}}.

        \paragraph{Client-Server, shared rendering}
          \label{para:client_server_shared_rendering}
          In Figure~\ref{fig:ssr-client-server-shared-output}, the setup shows
          \textit{1} to \textit{n} client instances, controlled by a server
          instance. All instances are receiving audio from an external
          \gls{jack} client or from reading local files. Collectively, the
          \textit{n} clients and the server are rendering audio on a shared
          output system (e.g. \gls{wfs} or \gls{hoa}).\\
          The server instance is controlled through its \gls{gui}, sends out
          \gls{osc} messages to all \textit{n} clients and receives their
          updated information (again through \gls{osc}).

          \begin{figure}[!htb]
            \centering
            \includegraphics[scale=1.0, trim = 20mm 204mm 10mm 10mm, clip]
            {ssr-client-server-shared-output.pdf}
            \caption{A diagram displaying a \gls{ssr} client/server setup, in
              which the server and the clients render audio collectively (e.g.
              \gls{wfs}). The server instance is not controlled via \gls{osc},
              but controls its clients through it.\\
              {\color{osc-in}\textbf{--}} \gls{osc} input
              {\color{osc-out}\textbf{--}} \gls{osc} output
              {\color{audio-in}\textbf{--}} Audio input
              {\color{audio-out}\textbf{--}} Audio output
            }
            \label{fig:ssr-client-server-shared-output}
          \end{figure}

          \begin{figure}[!htb]
            \centering
            \includegraphics[scale=1.0, trim = 20mm 204mm 10mm 10mm, clip]
            {ssr-external-client-server-shared-output.pdf}
            \caption{A diagram displaying a \gls{ssr} client/server setup, in
              which the server and the clients render audio collectively (e.g.
              \gls{wfs}). The server instance is controlled by an \gls{osc}
              capable application (acting as another client) and controls its
              clients through \gls{osc} as well.\\
              {\color{osc-in}\textbf{--}} \gls{osc} input
              {\color{osc-out}\textbf{--}} \gls{osc} output
              {\color{audio-in}\textbf{--}} Audio input
              {\color{audio-out}\textbf{--}} Audio output
            }
            \label{fig:ssr-external-client-server-shared-output}
          \end{figure}

          The setup shown in
          Figure~\ref{fig:ssr-external-client-server-shared-output} is similar
          to the previous one, with the exception, that the server instance is
          controlled by an external \gls{osc} capable application. This way,
          the server instance can also be run headless (without a \gls{gui}).\\
          The set of \textit{n} clients report back to the server instance,
          which in turn reports back to the \gls{osc} enabled application
          (acting as another client).
        \cleardoublepage

        \paragraph{Client-Server, separate rendering}
          \label{para:client_server_separate_rendering}
          As shown in Figure~\ref{fig:ssr-client-server-separate-output}, it is
          possible to have a setup, in which, similar to the one described in
          Figure~\ref{fig:ssr-client-server-shared-output}, server and
          \textit{n} clients render the same sources, but on separate output
          systems (e.g.\ several \gls{bs}/\gls{brs} renderers or even a mixture
          of a \gls{wfs}/\gls{hoa} system and several \gls{bs}/\gls{brs}
          renderers).\\

          \begin{figure}[!htb]
            \centering
            \includegraphics[scale=1.0, trim = 20mm 204mm 10mm 10mm, clip]
            {ssr-client-server-separate-output.pdf}
            \caption{A diagram displaying a \gls{ssr} client/server setup, in
              which the server and the clients render audio to separate outputs
              (e.g.\ multiple \gls{bs} renderers). The server instance is not
              controlled via \gls{osc}, but controls its clients through it.\\
              {\color{osc-in}\textbf{--}} \gls{osc} input
              {\color{osc-out}\textbf{--}} \gls{osc} output
              {\color{audio-in}\textbf{--}} Audio input
              {\color{audio-out}\textbf{--}} Audio output
            }
            \label{fig:ssr-client-server-separate-output}
          \end{figure}

          Figure~\ref{fig:ssr-external-client-server-separate-output} is an
          example of a similar setup, but again using an external \gls{osc}
          capable application to control the server instance.\\

          \begin{figure}[!htb]
            \centering
            \includegraphics[scale=1.0, trim = 20mm 204mm 10mm 10mm, clip]
            {ssr-external-client-server-separate-output.pdf}
            \caption{A diagram displaying a \gls{ssr} client/server setup, in
              which the server and the clients render audio separately (e.g.\
              multiple \gls{bs} renderers). The server instance is controlled
              by an \gls{osc} capable application (acting as another client)
              and controls its clients through \gls{osc} as well.\\
              {\color{osc-in}\textbf{--}} \gls{osc} input
              {\color{osc-out}\textbf{--}} \gls{osc} output
              {\color{audio-in}\textbf{--}} Audio input
              {\color{audio-out}\textbf{--}} Audio output
            }
            \label{fig:ssr-external-client-server-separate-output}
          \end{figure}
        \cleardoublepage

        \paragraph{Clients only}
          \label{para:clients_only}
          Using the new \gls{osc} interface, it is also possible to have
          client-only setups, in which an \gls{osc} capable application mimics
          a \gls{ssr} server. This way a set of \textit{n} clients can
          collectively (see
          Figure~\ref{fig:ssr-external-clients-only-shared-output}) or
          separately (see
          Figure~\ref{fig:ssr-external-clients-only-separate-output}) render
          audio, without the specific need of a \gls{ssr} server instance
          controlling them. The clients send their update information back to
          the controlling application.\\
          Much of the functionality implemented in the server-side of the
          \gls{osc} interface however has to be reapplied to the controlling
          software and its behavior, when dealing with \gls{ssr} clients.

          \begin{figure}[!htb]
            \centering
            \includegraphics[scale=1.0, trim = 20mm 204mm 10mm 10mm, clip]
            {ssr-external-clients-only-shared-output.pdf}
            \caption{A diagram displaying a \gls{ssr} client cluster setup, in
              which a set of clients render audio collectively (e.g.\ medium or
              large-scale \gls{wfs} setup). An \gls{osc} capable application
              acts as a \gls{ssr} server instance and controls the clients.\\
              {\color{osc-in}\textbf{--}} \gls{osc} input
              {\color{osc-out}\textbf{--}} \gls{osc} output
              {\color{audio-in}\textbf{--}} Audio input
              {\color{audio-out}\textbf{--}} Audio output
            }
            \label{fig:ssr-external-clients-only-shared-output}
          \end{figure}

          \begin{figure}[!htb]
            \centering
            \includegraphics[scale=1.0, trim = 20mm 204mm 10mm 10mm, clip]
            {ssr-external-clients-only-separate-output.pdf}
            \caption{A diagram displaying a \gls{ssr} client cluster setup, in
              which a set of clients render audio separately (e.g.\ multiple
              \gls{bs} renderers). An \gls{osc} capable application acts as a
              \gls{ssr} server instance and controls the clients.\\
              {\color{osc-in}\textbf{--}} \gls{osc} input
              {\color{osc-out}\textbf{--}} \gls{osc} output
              {\color{audio-in}\textbf{--}} Audio input
              {\color{audio-out}\textbf{--}} Audio output
            }
            \label{fig:ssr-external-clients-only-separate-output}
          \end{figure}

  \cleardoublepage
      \subsubsection{Message levels}
        \label{subsubsec:message_levels}
        To be able to distinguish between types of
        clients and servers, several message levels were implemented for the
        \gls{osc} interface conceived in the course of this work.\\
        The \mintinline{c++}{enumeration class MessageLevel} (see
        Listing~\ref{lst:ssr_global.h}) defines the four types
        \mintinline{c++}{CLIENT}, \mintinline{c++}{GUI_CLIENT},
        \mintinline{c++}{SERVER}, \mintinline{c++}{GUI_SERVER}, which are
        represented as non-negative integers (in ascending order), starting
        from 0.\\

        \begin{listing}[!htb]
          \begin{mdframed}
            \inputminted[numbers=left, firstline=54, lastline=61,
              fontsize=\footnotesize]{c++}{../../ssr/src/ssr_global.h}
          \end{mdframed}
          \caption{src/ssr\_global.h: \mintinline{c++}{enum class
          MessageLevel}}
          \label{lst:ssr_global.h}
        \end{listing}\\

        \noindent\gls{ssr} client instances subscribe to \gls{ssr} server
        instances with the \mintinline{c++}{MessageLevel}
        \mintinline{c++}{CLIENT} by default. Server instances get the
        \mintinline{c++}{MessageLevel} \mintinline{c++}{SERVER} assigned to by
        each client on subscribing to it.\\
        In the \gls{osc} interface it is implemented as follows: A (recycable
        and reconfigurable) list of clients is held by a server instance, which
        enables for the \mintinline{c++}{MessageLevel} to change for each
        client. Every client instance holds a (reconfigurable) server
        representation, that enables for the \mintinline{c++}{MessageLevel} to
        change for each client towards its server.\\
        Several messages, such as information related to \gls{cpu} load or
        master signal level are not useful for a rendering client (additionally
        they are issued in very short intervals, which can lead to performance
        issues), which is why they are only sent to clients with a
        \mintinline{c++}{MessageLevel} \mintinline{c++}{GUI_CLIENT} or servers
        with a \mintinline{c++}{MessageLevel} \mintinline{c++}{GUI_SERVER}.\\
        Lightweight \gls{osc} capable applications used to control a \gls{ssr}
        server instance are clients to said server instance. An elevated
        \mintinline{c++}{MessageLevel} of \mintinline{c++}{SERVER} (instead of
        \mintinline{c++}{CLIENT}) enables them to send messages to the server
        and have them evaluated.\\
        Analogous to a server instance holding a \mintinline{c++}{MessageLevel}
        of \mintinline{c++}{GUI_SERVER} towards its clients, a client instance
        can hold the same \mintinline{c++}{MessageLevel} towards a server
        instance to receive the above mentioned performance heavy \gls{osc}
        messages.
        How the setting up of message levels is achieved, is further elaborated
        in the following section.

      \subsubsection{Message interface}
        \label{subsubsec:message_interface}
        \gls{osc} offers the possibility of a hierarchical path tree, that can
        be used to group messages by type (i.e.\ context). In conjunction with
        messages only understood by client or server (or a context dependant
        meaning), most of the messages understood by the
        \nameref{subsec:ip-interface} are implemented. Additional features,
        related to server-client and client-only functionality, were integrated
        as well.\\
        In general, it can be distinguished between \textit{direct} messages -
        sent from a server (or an application mimicking one) to a client or a
        server to trigger processing (see
        Table~\ref{tab:ssr-osc-processing-tracker-transport}), reference (see
        Table~\ref{tab:ssr-osc-reference}), scene (see
        Table~\ref{tab:ssr-osc-scene}), source (see
        Table~\ref{tab:ssr-osc-source}), tracker (see
        Table~\ref{tab:ssr-osc-processing-tracker-transport}) or transport (see
        Table~\ref{tab:ssr-osc-processing-tracker-transport}) related
        operations in the \gls{ssr} --- and \textit{update} messages (see
        Table~\ref{tab:ssr-osc-update}) --- sent from a client to a server upon
        successful processing an operation related to a \textit{direct}
        message.\\

        \begin{table}[!htb]
          \renewcommand{\arraystretch}{1.2}
          \scriptsize
          \centering
          \rowcolors{2}{table-background-one}{table-background-two}
          \begin{tabular}{ p{2cm} | p{1cm} | p{3.5cm} | p{3cm} }
            \textbf{Path} & \textbf{Types} & \textbf{Description} &
            \textbf{Example}\\
            \hline
            \texttt{/alive} & & Alive notification from client (in response to
            a /poll) & \texttt{[/alive]} \\
            \texttt{/message\_level} & i & Set message level of sender &
            \texttt{[/message\_level, 1]} \\
            \texttt{/message\_level} & ssi & Set message level of a specific
            client & \texttt{[/message\_level, “127.0.0.1”,
            “50002”, 1]} \\
            \texttt{/subscribe} & F & Unsubscribe sender & \texttt{[/subscribe,
            false]} \\
            \texttt{/subscribe} & Fss & Unsubscribe specific client &
            \texttt{[/subscribe, false, “127.0.0.1”, “50002”]}\\
            \texttt{/subscribe} & T & Subscribe sender & \texttt{[/subscribe,
            true]} \\
            \texttt{/subscribe} & Ti & Subscribe sender with specific message
            level & \texttt{[/subscribe, true, 1]} \\
            \texttt{/subscribe} & Tssi & Subscribe specific client with
            specific message level & \texttt{[/subscribe, true, “127.0.0.1”,
            “50002”, 1]} \\
          \end{tabular}
          \caption{\gls{osc} messages relevant for subscribing and setting of
            message levels for clients.\\
          }
          \caption*{Understood by server.\\
            Data types and their acronyms are listed in
            Table~\ref{tab:ssr-osc-data-type-acronyms}.
          }
          \label{tab:ssr-osc-subscribe}
        \end{table}

        \begin{table}[!htb]
          \renewcommand{\arraystretch}{1.2}
          \scriptsize
          \centering
          \rowcolors{2}{table-background-one}{table-background-two}
          \begin{tabular}{ p{2cm} | p{1cm} | p{3.5cm} | p{3cm} }
            \textbf{Path} & \textbf{Types} & \textbf{Description} &
            \textbf{Example}\\
            \hline
            \texttt{/message\_level} & i & Set message level of sender (the
            server) & \texttt{[/message\_level, 1]} \\
            \texttt{/poll} & & Poll client (continously sent), triggering
            subscribe (see Table~\ref{tab:ssr-osc-subscribe}) &
            \texttt{[/poll]} \\
          \end{tabular}
          \caption{\gls{osc} messages relevant for polling and setting of
            message levels for servers subscribed to.\\
          }
          \caption*{Understood by client.\\
            Data types and their acronyms are listed in
            Table~\ref{tab:ssr-osc-data-type-acronyms}.
          }
          \label{tab:ssr-osc-client-poll-message-level}
        \end{table}

        A special set of \textit{direct} message are the \textit{subscribe} and
        \textit{message level} (see Table~\ref{tab:ssr-osc-subscribe}) and
        \textit{poll} and \textit{message level} (see
        Table~\ref{tab:ssr-osc-client-poll-message-level}) messages. The former
        --- understood only by \gls{ssr} server instances --- enable clients to
        subscribe (with a certain message level) or subscribe other clients
        (with a predefined message level) and set their own message level or
        that of another client. The latter set --- only understood by clients
        --- enables servers (or applications mimicking one) to poll yet
        unsubscribed clients to have them subscribe and subscribed clients to
        reply with an alive message. Similar to the \textit{message level}
        message understood by server instances, the one understood by clients
        sets the message level (of the server representation in the client).\\

        \begin{table}[!htb]
          \renewcommand{\arraystretch}{1.2}
          \scriptsize
          \centering
          \rowcolors{2}{table-background-one}{table-background-two}
          \begin{tabular}{ p{3cm} | p{1.2cm} | p{3.5cm} | p{3cm} }
            \textbf{Path} & \textbf{Types} & \textbf{Description} &
            \textbf{Example}\\
            \hline
            \texttt{/processing/state} & F & Unset processing state &
            \texttt{[/processing/state, false]} \\
            \texttt{/processing/state} & T & Set processing state &
            \texttt{[/processing/state, true]} \\
            \texttt{/tracker/reset} & & Reset tracker &
            \texttt{[/tracker/reset]} \\
            \texttt{/transport/rewind} & & Rewind the \gls{jack} transport &
            \texttt{[/transport/rewind]} \\
            \texttt{/transport/seek} & s & Seek to time code in
            \gls{jack} transport & \texttt{[/transport/seek, “42:00:00”]} \\
            \texttt{/transport/state} & F & Unset \gls{jack} transport
            state & \texttt{[/transport/state, false]} \\
            \texttt{/transport/state} & T & Set \gls{jack} transport
            state & \texttt{[/transport/state, true]} \\
          \end{tabular}
          \caption{\gls{osc} messages relevant for processing, tracker and
            (\gls{jack}) transport related settings.\\
          }
          \caption*{Understood by server and client.\\
            Data types and their acronyms are listed in
            Table~\ref{tab:ssr-osc-data-type-acronyms}.
          }
          \label{tab:ssr-osc-processing-tracker-transport}
        \end{table}

        \begin{table}[!htb]
          \renewcommand{\arraystretch}{1.2}
          \scriptsize
          \centering
          \rowcolors{2}{table-background-one}{table-background-two}
          \begin{tabular}{ p{4.3cm} | p{1cm} | p{2.5cm} | p{4.4cm} }
            \textbf{Path} & \textbf{Types} & \textbf{Description} &
            \textbf{Example}\\
            \hline
            \texttt{/reference/orientation} & f & Set azimuth of reference
            point & \texttt{[/reference/orientation, -90.0]} \\
            \texttt{/reference/position} & ff & Set position of reference &
            \texttt{[/reference/position, 1.5, 2.0]} \\
            \texttt{/reference\_offset/orientation} & f & Set azimuth of
            reference offset position &
            \texttt{[/reference\_offset/orientation, -90.0]} \\
            \texttt{/reference\_offset/position} & ff & Set position of
            reference offset & \texttt{[/reference\_offset/position, 1.5,
            2.0]}\\
          \end{tabular}
          \caption{\gls{osc} messages relevant for reference management.\\
          }
          \caption*{Understood by server and client.\\
            Data types and their acronyms are listed in
            Table~\ref{tab:ssr-osc-data-type-acronyms}.
          }
          \label{tab:ssr-osc-reference}
        \end{table}

        \begin{table}[!htb]
          \renewcommand{\arraystretch}{1.2}
          \scriptsize
          \centering
          \rowcolors{2}{table-background-one}{table-background-two}
          \begin{tabular}{ p{4.5cm} | p{0.9cm} | p{2.5cm} | p{4.3cm} }
            \textbf{Path} & \textbf{Types} & \textbf{Description} &
            \textbf{Example} \\
            \hline
            \texttt{/scene/amplitude\_reference
            \_distance} & f & Set amplitude
            reference distance. &
            \texttt{[/scene/amplitude\_reference
            \_distance, 6.0]}\\
            \texttt{/scene/auto\_rotate\_sources} & F & Disable automatic
            rotation of sources. & \texttt{[/scene/auto\_rotate\_sources,
            false]}\\
            \texttt{/scene/auto\_rotate\_sources} & T & Enable automatic
            rotation of sources. & \texttt{[/scene/auto\_rotate\_sources,
            true]}\\
            \texttt{/scene/clear} & & Delete all sources &
            \texttt{[/scene/clear]}\\
            \texttt{/scene/decay\_exponent} & f & Set amplitude decay exponent
            in virtual space ($1/r^{exp}$). & \texttt{[/scene/decay\_exponent,
            2.0]}\\
            \texttt{/scene/load} & s & Load scene from \gls{asdf} file. &
            \texttt{[/scene/load, “example.asd”]}\\
            \texttt{/scene/master\_signal\_level} & f & Set the renderers
            signal level. & \texttt{[/scene/master\_signal\_level, -20]}\\
            \texttt{/scene/save} & s & Save scene to \gls{asdf} file. &
            \texttt{[/scene/save, “example.asd”]}\\
            \texttt{/scene/volume} & f & Set scene master volume. &
            \texttt{[/scene/volume, 0.23]}\\
          \end{tabular}
          \caption{\gls{osc} messages relevant for scene management.\\
          }
          \caption*{Understood by server and client.\\
            Data types and their acronyms are listed in
            Table~\ref{tab:ssr-osc-data-type-acronyms}.
          }
          \label{tab:ssr-osc-scene}
        \end{table}

        When starting a \gls{ssr} server instance
        (see~\ref{para:server-instance}), it responds to the messages shown in
        Table~\ref{tab:ssr-osc-subscribe}
        ,~\ref{tab:ssr-osc-scene}
        ,~\ref{tab:ssr-osc-source}
        ,~\ref{tab:ssr-osc-processing-tracker-transport}
        ,~\ref{tab:ssr-osc-reference} and~\ref{tab:ssr-osc-update}.

        \begin{table}[!htb]
          \renewcommand{\arraystretch}{1.2}
          \scriptsize
          \centering
          \rowcolors{2}{table-background-one}{table-background-two}
          \begin{tabular}{ p{3.5cm} | p{1.5cm} | p{2.5cm} | p{4.3cm} }
            \textbf{Path} & \textbf{Types} & \textbf{Description} &
            \textbf{Example} \\
            \hline
            \texttt{/source/delete} & i & Delete source with given id &
            \texttt{[/source/delete, 1]}\\
            \texttt{/source/file\_channel} & ii & Set a source's file channel &
            \texttt{[/source/file\_channel, 1, 2]}\\
            \texttt{/source/file\_name\_ or\_port\_number} & is & Set a source's
            file name or port number &
            \texttt{[/source/file\_name\_ or\_port\_number, 1, “1”]}\\
            \texttt{/source/port\_name } & is & Set a source's \gls{jack} input
            port name & \texttt{[/source/port\_name, 1, “system:capture\_2”]}\\
            \texttt{/source/gain} & if & Set a source's gain &
            \texttt{[/source/gain, 1, 0.2]}\\
            \texttt{/source/model} & is & Set a source's model &
            \texttt{[/source/model, 1, “point”]}\\
            \texttt{/source/mute} & iF & Unmute a source &
            \texttt{[/source/mute, 1, false]}\\
            \texttt{/source/mute} & iT & Mute a source &
            \texttt{[/source/mute, 1, true]}\\
            \texttt{/source/name} & is & Set a source's name &
            \texttt{[/source/name, 1, “Daisy”]}\\
            \texttt{/source/new} & i & Create a new source stub using id &
            \texttt{[/source/new, 1]}\\
            \texttt{/source/new} & sssffffTFF & Create a new source
            (auto-generated id) with name, model, port number, X-coordinate,
            Y-coordinate, orientation, gain, movability, orientation movability
            and mute status & \texttt{[/source/new, “Daisy”, “point”, “1”, 1.0,
            2.5, 90.0, 0.2, true, false, false]} \\
            \texttt{/source/new} & sssffffisTFF & Create a new source
            (auto-generated id) with name, model, port number, X-coordinate,
            Y-coordinate, orientation, gain, file channel, properties file,
            movability, orientation movability and mute status &
            \texttt{[/source/new, “Daisy”, “point”, “1”, 1.0, 2.5, 90.0, 0.2,
            2, “properties.xml”, true, false, false]} \\
            \texttt{/source/orientation} & if & Set a source's orientation &
            \texttt{[/source/orientation, 1, -90.0]}\\
            \texttt{/source/position} & iff & Set a source's position &
            \texttt{[/source/position, 1, 1.5, 2.0]}\\
            \texttt{/source/position\_fixed} & iF & Set a source movable &
            \texttt{[/source/position\_fixed, 1, false]}\\
            \texttt{/source/position\_fixed} & iT & Set a source immovable &
            \texttt{[/source/position\_fixed, 1, true]}\\
            \texttt{/source/properties\_file} & is & Set a source's properties
            file & \texttt{[/source/properties\_file, 1,
            “source-properties.xml”]}\\
          \end{tabular}
          \caption{\gls{osc} messages relevant for source management.\\
          }
          \caption*{Understood by server and client.\\
            Data types and their acronyms are listed in
            Table~\ref{tab:ssr-osc-data-type-acronyms}.
          }
          \label{tab:ssr-osc-source}
        \end{table}

        \noindent A client instance (see~\ref{para:client-instance}) will only
        respond to the \textit{direct} messages listed in
        Table~\ref{tab:ssr-osc-client-poll-message-level}
        ,~\ref{tab:ssr-osc-scene} ,~\ref{tab:ssr-osc-source}
        ,~\ref{tab:ssr-osc-processing-tracker-transport}
        and~\ref{tab:ssr-osc-reference}, but is able to send \textit{update}
        messages.

        \begin{table}[!htb]
          \renewcommand{\arraystretch}{1.2}
          \scriptsize
          \centering
          \rowcolors{2}{table-background-one}{table-background-two}
          \begin{tabular}{ p{6.5cm} | p{0.9cm} | p{4.3cm} }
            \textbf{Path} & \textbf{Types} & \textbf{Description} \\
            \hline
            \texttt{/update/cpu\_load} & f & \gls{cpu} load changes.\\
            \texttt{/update/processing/state} & T & Processing state is set.\\
            \texttt{/update/processing/state} & F & Processing state is
            unset.\\
            \texttt{/update/reference/orientation} & f & Reference orientation
            changes.\\
            \texttt{/update/reference/position} & ff & Reference position
            changes.\\
            \texttt{/update/reference\_offset/orientation} & f & Reference
            offset orientation changes.\\
            \texttt{/update/reference\_offset/position} & ff & Reference offset
            position changes.\\
            \texttt{/update/scene/amplitude\_reference\_distance} & f &
            Amplitude reference distance changes.\\
            \texttt{/update/scene/auto\_rotate\_sources} & T & Auto rotation of
            sources is set.\\
            \texttt{/update/scene/auto\_rotate\_sources} & F & Auto rotation of
            sources is unset.\\
            \texttt{/update/scene/decay\_exponent} & f & The scene's decay
            exponent has changed.\\
            \texttt{/update/scene/master\_signal\_level} & f & Master signal
            level has changed.\\
            \texttt{/update/scene/sample\_rate} & i & Sample rate of the scene
            changed.\\
            \texttt{/update/scene/volume} & f & Volume of the scene has
            changed.\\
            \texttt{/update/source/delete} & i & A source with given id was
            deleted. \\
            \texttt{/update/source/file\_channel} & ii & A source's file
            channel was set.\\
            \texttt{/update/source/file\_name\_or\_port\_number} & is & A
            source's file name or port number was set.\\
            \texttt{/update/source/gain} & if & A source's gain was set.\\
            \texttt{/update/source/length} & ii & A source's length was set.\\
            \texttt{/update/source/level} & if & A source's output level has
            changed.\\
            \texttt{/update/source/model} & is & A source's model was set.\\
            \texttt{/update/source/mute} & iF & A source was unmuted.\\
            \texttt{/update/source/mute} & iT & A source was muted.\\
            \texttt{/update/source/name} & is & A source's name was set.\\
            \texttt{/update/source/orientation} & if & A source's orientation
            was set. \\
            \texttt{/update/source/new} & i & A new source with given id was
            created. \\
            \texttt{/update/source/port\_name} & is & A source's \gls{jack}
            port\_name was set. \\
            \texttt{/update/source/position} & iff & A source's position was
            set. \\
            \texttt{/update/source/position\_fixed} & iF & A source was set to
            be movable. \\
            \texttt{/update/source/position\_fixed} & iT & A source was set to
            be immovable. \\
            \texttt{/update/source/properties\_file} & is & A source's
            properties\_file was set.\\
            \texttt{/update/transport/seek} & s & \gls{jack} transport seeked
            to a timecode position.\\
            \texttt{/update/transport/state} & F & \gls{jack} transport was
            stopped.\\
            \texttt{/update/transport/state} & T & \gls{jack} transport was
            started.\\
          \end{tabular}
          \caption{\gls{osc} messages for updating information on \gls{cpu} load,
            processing, reference, scene, source, and transport of clients on a
            server.\\
            No examples are given, as they are mostly analogous to the ones in
            Table~\ref{tab:ssr-osc-processing-tracker-transport}
            ,~\ref{tab:ssr-osc-scene} and~\ref{tab:ssr-osc-source}.
          }
          \caption*{Understood by server.\\
            Data types and their acronyms are listed in
            Table~\ref{tab:ssr-osc-data-type-acronyms}.
          }
          \label{tab:ssr-osc-update}
        \end{table}

        \cleardoublepage

      \subsubsection{Workflow Examples}
        \label{subsubsec:workflow_examples}
        Using any \gls{osc} capable programming language or application enables
        for communication with the \gls{ssr}. The following examples illustrate
        simple workflows using \gls{sclang} and should therefore be \gls{os}
        agnostic.\\

        \paragraph{Controlling a server}
          \label{para:controlling_a_server}

          \begin{listing}[!htb]
            \begin{mdframed}
              \inputminted[numbers=left, firstline=3, lastline=29,
              fontsize=\footnotesize]{supercollider}
              {../../ssr/supercollider/workflows.scd}
            \end{mdframed}
            \caption{supercollider/workflows.scd: \gls{sclang} as client
            controlling a \gls{ssr} server instance}
            \label{lst:ssr-workflow-sclang-controls-server}
          \end{listing}\\

          As shown in Listing~\ref{lst:ssr-workflow-sclang-controls-server}, it
          is necessary to subscribe to the server instance with a
          \mintinline{c++}{MessageLevel} of \mintinline{c++}{SERVER} or higher.\\
          After doing so, also all \textit{direct} \gls{osc} messages (i.e.
          Table~\ref{tab:ssr-osc-source},~\ref{tab:ssr-osc-scene}
          ,~\ref{tab:ssr-osc-reference},~\ref{tab:ssr-osc-subscribe}) are
          evaluated when sent to the \gls{ssr}.\\
          The server instance will relay valid messages to all of its active
          clients.

        \paragraph{Server mimicry}
          \label{para:server_mimicry}

          \begin{listing}[!htb]
            \begin{mdframed}
              \inputminted[numbers=left, firstline=31, lastline=61,
              fontsize=\footnotesize]{supercollider}
              {../../ssr/supercollider/workflows.scd}
            \end{mdframed}
            \caption{supercollider/workflows.scd: \gls{sclang} mimics server,
            controlling a \gls{ssr} client instance}
            \label{lst:ssr-workflow-sclang-is-server-controls-client}
          \end{listing}\\

          When mimicking a \gls{ssr} server instance in a client-only setup
          (e.g.\ Figure~\ref{fig:ssr-external-clients-only-shared-output} or
          Figure~\ref{fig:ssr-external-clients-only-separate-output}), it is
          necessary to send a poll message to the client instance to make it
          subscribe (which sets the server's address and port up internally).\\
          Afterwards --- similar to the example in the subsection
          \nameref{para:clients_only} --- all \textit{direct} \gls{osc}
          messages are accepted by the client instance, when coming from the
          server address and port.\\
          An interesting concept here is to (temporarily) set a different
          \mintinline{c++}{MessageLevel} for the application acting as a server
          (e.g.\ to \mintinline{c++}{GUI_SERVER}), to receive \gls{gui}
          relevant messages, as explained in
          \nameref{subsubsec:message_interface}.\\

  \cleardoublepage
  \section{Discussion}
    \label{sec:discussion}
    The \gls{osc} based networking extension created for the \gls{ssr} can be
    considered a valuable usability improvement. Its implemented features are
    discussed further in the following section, followed by an outlook on
    related future work. The extension is additionally extensively documented
    in the source code, to ensure the ease of further development.\\
    Due to the versatility of how the \gls{ssr} can be used in a networking
    context, it is likely, that some of its possibilities are not even
    accounted for.

    \subsection{Implemented Features}
      \label{subsec:implemented_features}
      The \gls{osc} interface described in~\ref{subsec:osc-interface} can be
      seen as a full replacement (with one minor exception, detailed
      in~\ref{subsubsec:scene_transfer}) for the \gls{ip} interface, already in
      place. Its additional features are what set it trully apart though, when
      not only regarding non-reliance on external software to enable \gls{osc}
      capabilities.\\

      The implementation follows the internal \gls{pubsub} interface, as
      described in~\ref{subsec:publisher_subscriber_interface} and extends it,
      where appropriate. Additionally, an open client-server architecture has
      been created, according to a message level system, further elaborated
      in~\ref{subsubsec:message_levels}. An attempt at giving extensive
      examples on the various setup possibilities, that are now available, is
      made in~\ref{subsubsec:setups}, some of which are still dependant on
      various missing features (see~\ref{subsec:future_work}).\\
      The \gls{osc} messaging system is adhering to the aforementioned
      client-server architecture by distinguishing between client-only,
      server-only and messages available to clients and servers alike
      (see~\ref{subsubsec:message_interface}). Examples for different workflows
      are given in~\ref{subsubsec:workflow_examples} to illustrate simple
      use-cases.\\
      This puts the \gls{osc} interface in the unique position of providing a
      native messaging interface and a flexible architecture. It can be used
      from single local instances up to large scale networked setups (with the
      limitations discussed in~\ref{subsubsec:alien_loudspeaker}
      and~\ref{subsubsec:assigning_in_and_outputs_on_the_fly}).\\
      While with the \gls{ip} interface, multiple instances are only
      controllable by using an \gls{osc} capable application or one, that is
      able to send \gls{xml}-formatted strings over \gls{tcp}/\gls{ip}, the
      \gls{osc} interface can deal with \textit{n} clients natively, while only
      one instance has to be controlled using \gls{osc}. The behavior implies,
      that setups are possible, in which a large collection of different types
      of renderers can share the same scene, which is particularly useful for
      e.g.\ auditioning different rendering algorithms over the same system, or
      rather in the same room.\\
      Message sending takes place over \gls{udp}, instead of \gls{tcp}, which
      lowers the complexity of the network topology (\gls{udp} does not perform
      handshakes for every packet sent, unlike \gls{tcp}) and thus the size of
      each message sent.\\
      The \gls{osc} interface therefore implements messaging, using lower
      bandwidth, at a greater feature set. In~\ref{subsec:automated_tests} a
      test environment is introduced, that further elaborates on the overall
      functionality and feasibility of the message interface.

    \subsection{Automated tests}
      \label{subsec:automated_tests}
      The \gls{ssr} was developed without the help of a test framework, which
      is responsible to test its components, after they have been changed.
      This means, that internal (e.g.\ the \gls{pubsub} interface), or external
      (e.g.\ the \gls{ip} or \gls{osc} interface) functionality might or might
      not work as expected. To test the \gls{osc} interface's logical coherency
      and robustness automatically, a set of tests was written in
      \gls{sclang}.\\
      The tests are divided into those probing robustness of the \gls{osc}
      interface and others probing its functionality. The robustness tests
      further divide into server and client specific tests, where authorized
      and unauthorized access is tried. The functionality tests are grouped by
      tests for general operability, i.e.\ testing certain features or
      workflows once and long-running tests, where features are tried
      repeatedly.\\
      \subsubsection{Robustness}
        \label{subsubsec:robustness}

        Listing~\ref{lst:ssr-tests-sclang-unsubscribed-controls-server}
        and~\ref{lst:ssr-tests-sclang-subscribed-controls-server} describe
        server-side tests for robustness. While the first will not lead to any
        processed action by the server, the latter will. This is explained by
        \gls{sclang} not being a subscribed client with a
        \mintinline{c++}{MessageLevel} of \mintinline{c++}{SERVER} or higher in
        the first case. However, in the second test \gls{sclang} subscribes to
        the \gls{ssr} server instance, which is why the \gls{osc} messages are
        evaluated in this case.\\

        \begin{listing}[!htb]
          \begin{mdframed}
            \inputminted[numbers=left, firstline=145, lastline=158,
            fontsize=\footnotesize]{supercollider}
            {../../ssr/supercollider/tests.scd}
          \end{mdframed}
          \caption{supercollider/tests.scd: \gls{sclang} (unsubscribed) tries
          to control a \gls{ssr} server}
          \label{lst:ssr-tests-sclang-unsubscribed-controls-server}
        \end{listing}\\

        \begin{listing}[!htb]
          \begin{mdframed}
            \inputminted[numbers=left, firstline=160, lastline=183,
            fontsize=\footnotesize]{supercollider}
            {../../ssr/supercollider/tests.scd}
          \end{mdframed}
          \caption{supercollider/tests.scd: \gls{sclang} (subscribed) tries to
          control a \gls{ssr} server}
          \label{lst:ssr-tests-sclang-subscribed-controls-server}
        \end{listing}\\

        The tests described in
        Listing~\ref{lst:ssr-tests-sclang-controls-client-unpolled}
        and~\ref{lst:ssr-tests-sclang-controls-client-polled} are client-side
        tests for robustness, that work in a similar fashion to the
        aforementioned server-side tests. While in the first case, the sent
        \gls{osc} messages are not evaluated, because \gls{sclang}, mimicking a
        server instance (see~\ref{para:server_mimicry}), did not poll the
        \gls{ssr} client instance up front, in the second case the messages are
        evaluated, because it did poll the client first.

        \begin{listing}[!htb]
          \begin{mdframed}
            \inputminted[numbers=left, firstline=185, lastline=197,
            fontsize=\footnotesize]{supercollider}
            {../../ssr/supercollider/tests.scd}
          \end{mdframed}
          \caption{supercollider/tests.scd: \gls{sclang} tries to control a
          \gls{ssr} client (without polling it)}
          \label{lst:ssr-tests-sclang-controls-client-unpolled}
        \end{listing}\\

        \begin{listing}[!htb]
          \begin{mdframed}
            \inputminted[numbers=left, firstline=199, lastline=214,
            fontsize=\footnotesize]{supercollider}
            {../../ssr/supercollider/tests.scd}
          \end{mdframed}
          \caption{supercollider/tests.scd: \gls{sclang} tries to control a
          \gls{ssr} client (with previously polling it)}
          \label{lst:ssr-tests-sclang-controls-client-polled}
        \end{listing}\\

        In all tests for robustness, the attempt is made to force errors in the
        implementation of the \nameref{subsubsec:message_interface}. This is
        achieved by purposely using ranges of data types for messages, that are
        not allowed, or not defined in the \gls{ssr}'s internal
        implementation.\\
        Two examples for weak spot exploitations were the use of negative
        integers for \glspl{id} in source related messages (only non-zero,
        non-negative \glspl{id} are allowed internally) or supplying an empty
        string as hostname or port number for subscription messages.\\
        The first example will lead to undefined behavior, if the range is not
        checked in the implementation, as a \textit{static\_cast} is used
        internally to cast the value of the message data type (\textit{unsigned
        int}) to the one expected by the \gls{ssr}'s Controller implementation
        (\textit{signed int}) and the outcome of said operation is
        implementation dependant (depending on the \gls{os} in use).\\
        The second example, if not checked for empty string, would lead to the
        \gls{osc} interface trying to create a possibly defective address and
        send poll messages out to it.\\
        While only some of the above mentioned scenarios could lead to a crash
        of the program under certain circumstances, left unhandled, all of them
        waste ressources, which is undesired.
        To circumvent possibly harmful input using the \gls{osc} interface, a
        set of sanity checks were implemented, that only allow for a received
        message to be processed, if all of its components fit the requirements.

      \subsubsection{Functionality and operability}
        \label{subsubsec:functionality_and_operability}

        \begin{listing}[!htb]
          \begin{mdframed}
            \inputminted[numbers=left, firstline=216, lastline=225,
            fontsize=\footnotesize]{supercollider}
            {../../ssr/supercollider/tests.scd}
          \end{mdframed}
          \caption{supercollider/tests.scd: \gls{sclang} controls a \gls{ssr}
          client (with previously polling it), creating several sources and
          moving them}
          \label{lst:ssr-tests-sclang-sources}
        \end{listing}\\

        The test described in Listing~\ref{lst:ssr-tests-sclang-sources} is a
        test for functionality, which also serves as a long-running stress test
        for the \gls{ssr}. It creates 20 sources, that are then moved around
        randomly, every 100ms, for 100 seconds, which on a Lenovo W540, with an
        Intel i7-4700MQ and 16Gb RAM created less than 50\% of \gls{cpu}
        load.\\
        Based on the above mentioned tests, the basic functionality of the
        \gls{osc} interface can be guaranteed and depending on the host's
        hardware also a maximum degree of capacity utilization can be
        estimated, when observing the \gls{ssr}'s workload towards the system,
        while using the long-running tests.\\
        It has to be mentioned, that a higher load can be observed, when using
        higher levels of verbosity (especially above \textbf{-vv}). This is
        explained by the fact, that the \gls{ssr} will print out every
        \gls{osc} message received and sent above the aforementioned verbosity
        level.

    \subsection{Future Work}
      \label{subsec:future_work}
      Several features, interesting for different use-cases, were out of scope
      for this work. They are however complementary to the \gls{osc} networking
      extension, or can be implemented on top of it and will be discussed in
      the following sections.\\

      \subsubsection{Non-renderer}
        \label{subsubsec:non-renderer}
        The \gls{ssr} features a \gls{gui}, that was in the process of being
        upgraded for \gls{qt5} at the time of writing. Future versions of the
        software could be used to also display setups of networking
        instances, instead of only displaying the locally running.\\
        The implementation could be desirable for massive multi-channel setups
        and simply switching between several (local or network-attached)
        \gls{ssr} instances alike. An additional identifier for the \textbf{-N}
        flag (see~\ref{para:server-instance}) could be used to start an
        instance in this mode.\\

        \begin{figure}[!htb]
          \centering
          \includegraphics[scale=1.0, trim = 20mm 204mm 10mm 10mm, clip]
          {ssr-client-server-clients-only-shared-output.pdf}
          \caption{A diagram displaying a \gls{ssr} client/server setup, in which
            only the clients render audio collectively (e.g.\ medium or
            large-scale \gls{wfs}). The server instance is not controlled via
            \gls{osc}, but controls its clients through it. Additionally, its
            rendering engine does not have any outputs.\\
            {\color{osc-in}\textbf{--}} \gls{osc} input
            {\color{osc-out}\textbf{--}} \gls{osc} output
            {\color{audio-in}\textbf{--}} Audio input
            {\color{audio-out}\textbf{--}} Audio output
          }
          \label{fig:ssr-client-server-clients-only-shared-output}
        \end{figure}

        The functionalities of the \gls{ssr}'s \gls{gui}, its several spatial
        audio renderers and \gls{osc} interface (amongst other parts) are
        determined by its \gls{pubsub}. For the \gls{gui} part of the software
        to display information about a networked setup, or even switch between
        several of them, it is therefore not needed or even desirable for that
        instance to render audio at all. An instance with such features could
        be imagined as a \gls{gui} only frontend.\\
        Figure~\ref{fig:ssr-client-server-clients-only-shared-output}
        illustrates a scenario, in which a server instance is used to control a
        set of \textit{n} clients, that collectively render audio (e.g.\ on a
        large scale \gls{wfs} or \gls{nfc-hoa} system). In contrast to the
        client instances, the server does not render any audio (i.e.\ has no
        outputs) and might not even need any audio input.\\
        The server in this example could also be a client, subscribed to a
        server instance in a cluster similar to the one in
        Figure~\ref{fig:ssr-external-client-server-shared-output} (i.e.\ as the
        \gls{osc} capable application controlling the server instance by using
        a \mintinline{c++}{MessageLevel} of \mintinline{c++}{SERVER} or higher).

        \begin{figure}[!htb]
          \centering
          \includegraphics[scale=1.0, trim = 20mm 204mm 10mm 10mm, clip]
          {ssr-client-server-clients-only-separate-output.pdf}
          \caption{A diagram displaying a \gls{ssr} client/server setup, in
            which only the clients render audio to separate outputs (e.g.\
            multiple \glspl{bs} renderers). The server instance is not controlled
            via \gls{osc}, but controls its clients through it. Additionally, its
            rendering engine does not have any outputs.\\
            {\color{osc-in}\textbf{--}} \gls{osc} input
            {\color{osc-out}\textbf{--}} \gls{osc} output
            {\color{audio-in}\textbf{--}} Audio input
            {\color{audio-out}\textbf{--}} Audio output
          }
          \label{fig:ssr-client-server-clients-only-separate-output}
        \end{figure}

        The example shown in
        Figure~\ref{fig:ssr-client-server-clients-only-separate-output} is
        similar to the one before, with the difference, that its \textit{n}
        clients render audio on separate systems.\\
        Said client instances could be a cluster of headphone renderers, such
        as \gls{brs} renderers, or loudspeaker renderers, such as \gls{vbap} or
        \gls{wfs}, or even a combination of both types.\\
        Analogous to the example given before, the \gls{gui} only frontend
        could also act as a client with a \mintinline{c++}{MessageLevel} of
        \mintinline{c++}{SERVER} or higher (i.e.\ again being the \gls{osc}
        capable application), controlling the server in a setup, similar to the
        one shown in
        Figure~\ref{fig:ssr-external-client-server-separate-output}.\\
        For the aforementioned use-cases to work, the \gls{gui} of the
        \gls{ssr} has to be extended to show the networking specific
        information and be able to use the \gls{osc} interface through the
        \gls{pubsub} interface (which itself would probably have to be extended
        as well). Additionally, a renderer should be conceived, that does not
        render audio, but still uses the \gls{pubsub}, so the internal
        functionality of the \gls{ssr} can be reused, leading to a relatively
        light-weight \gls{ssr} \gls{gui} variant, only for controlling
        setups.\\
        As there are many edge cases to networked setups, it seems still
        unclear, whether audio inputs would actually be needed for such a
        renderer.

        \cleardoublepage

      \subsubsection{Alien loudspeaker}
        \label{subsubsec:alien_loudspeaker}
        For the examples given in
        Figure~\ref{fig:ssr-client-server-shared-output}
        ,~\ref{fig:ssr-client-server-clients-only-shared-output}
        ,~\ref{fig:ssr-external-clients-only-shared-output}
        and~\ref{fig:ssr-external-client-server-shared-output}, which
        facilitate a set of \textit{n} clients (server instances are counted as
        clients for the point made, where applicable), used for rendering in a
        medium or large scale loudspeaker based setup, an additional type of
        loudspeaker should be conceived and implemented in the different
        renderers.\\

        \begin{listing}[!htb]
          \begin{mdframed}
            \inputminted[numbers=left, firstline=44, lastline=51,
              fontsize=\footnotesize]{c++}{../../ssr/src/loudspeaker.h}
          \end{mdframed}
          \caption{src/loudspeaker.h: \mintinline{c++}{enum model_t}}
          \label{lst:loudspeaker.h}
        \end{listing}\\

        In the setups, where rendering takes place collectively, each client
        instance currently is reproducing the complete reproduction setup. This
        means, that for a reproduction setup with several hundred loudspeakers,
        each client creates exactly that amount of \gls{jack} outputs each,
        although it would only be responsible for discretely rendering audio on
        a subset of them.\\
        To cope with this edge case, the current loudspeaker renderers would
        have to be extended to be able to distinguish between the loudspeakers
        of the current and that of other instances. This would enable them to
        create \gls{jack} outputs in the amount of the loudspeakers they are
        rendering audio for and reduce the overall processing usage.\\
        In Listing~\ref{lst:loudspeaker.h} the \mintinline{c++}{enum model_t},
        defining the loudspeaker types available to the \gls{ssr}, is extended
        to facilitate the new model type \mintinline{c++}{alien}, which could
        be used internally by the renderers to identify loudspeakers, not to
        render on.\\

        \begin{listing}[!htb]
          \begin{mdframed}
            \inputminted[numbers=left, firstline=44, lastline=62,
            fontsize=\footnotesize, breaklines]{xml}{../../ssr/data/asdf.xsd}
            \inputminted[numbers=left, firstline=250, lastline=264,
            fontsize=\footnotesize]{xml}{../../ssr/data/asdf.xsd}
            \inputminted[numbers=left, firstline=314, lastline=318,
            fontsize=\footnotesize, breaklines]{xml}{../../ssr/data/asdf.xsd}
            \inputminted[numbers=left, firstline=361, lastline=364,
            fontsize=\footnotesize, breaklines]{xml}{../../ssr/data/asdf.xsd}
            \inputminted[numbers=left, firstline=186, lastline=186,
            fontsize=\footnotesize]{xml}{../../ssr/data/asdf.xsd}
          \end{mdframed}
          \caption{data/asdf.xsd: Reproduction setup, loudspeaker, circular
            array and linear array definition in \gls{asdf}, extended by a
          \textit{hostname} attribute}
        \label{lst:asdf.xsd}
        \end{listing}\\

        Already when defining a reproduction setup for the \gls{ssr},
        host-specific loudspeakers have to be taken into account.
        Listing~\ref{lst:asdf.xsd} shows an attempt at providing a unique
        attribute for each part of the setup, that is referencing a loudspeaker
        --- the hostname or \gls{ip} address of the host --- by extending the
        \gls{asdf}.\\
        More work has to be put into implementing this feature, or rather
        improvement however, as it also requires tests in medium and large
        scale setups, to ensure a discrete rendering, as if only using one
        host.

      \subsubsection{Status messages}
        \label{subsubsec:status_messages}
        When reflecting about different use-cases for networking setups
        involving the \gls{ssr}, it became apparent, that in certain
        situations it would be desirable to be able to poll instances for
        information, involving sources, scenes and the like.\\
        One example is the implementation of a light-weight, single-purpose
        \gls{gui} (e.g.\ non-interactive display of source positions) in
        another programming language, such as \gls{python}, \gls{pd}, or
        \gls{supercollider}, while only relying on \gls{osc} for communication
        between the parts. Another would be the case, where certain aspects of
        a client or server instance should be monitored (e.g.\ \gls{cpu}
        usage). Both examples should allow a \gls{gui} (or any other
        monitoring) process to be subscribed, after the active \gls{ssr}
        instances started rendering.\\
        To be able to retrieve information from a \gls{ssr} instance, its
        \gls{pubsub} interface has to be extended and \mintinline{c++}{get}
        functions implemented --- where applicable --- to return the desired
        information. A special case of this feature is described
        in~\ref{subsubsec:scene_transfer}.

      \subsubsection{Scene transfer}
        \label{subsubsec:scene_transfer}
        The \gls{ip} interface of the \gls{ssr} implements a functionality to
        transfer all information related to a scene as a \gls{xml} formatted
        string. This is useful, if the scene information should be stored on
        the machine requesting the information, instead of on the rendering
        machine.\\
        Due to shortage of time to implement it and the original functionality
        heavily relying on the \gls{xml} associated code, the \gls{osc}
        interface still lacks this feature, which in its context could
        additionally also be used to transfer all scene information to another
        client, subscribed to a server.\\
        It could be particularly useful, if clients could for example request
        the scene currently held be their server instance, in the case where
        they are started after the server has been started, but its scene
        already being setup. The server would then send a set of instructions
        as \gls{osc} messages, needed for setting up the scene in question.\\
        For this feature to work reliably, some edge cases have to be
        considered, such as gaps in the list of source \glspl{id}: Every source
        gets a unique non-zero, non-negative \gls{id} assigned on creation.
        When a source is deleted, its \gls{id} is not assigned to a source
        anymore and will not be reused, unless the whole scene is deleted and a
        new cycle of source creation reaches a number that high.\\
        This means, if a scene with source \gls{id} gaps has to be transferred,
        the \gls{osc} messages have to be designed in such a way, that they can
        account for them, as every source message (apart from
        \textit{/source/new}) requires a valid source \gls{id} and subsequent
        calls to the server would otherwise trigger uncorrectly mapped
        operations on its clients sources.\\
        Additionally, it would be useful to be able to transfer a scene to
        another client, by request, if the caller's
        \mintinline{c++}{MessageLevel} is \mintinline{c++}{SERVER} or higher.\\
        The scene is a piece of redundant information, in a networked setup
        (whereas the reproduction setup is individual). Being able to transfer
        scenes, using \gls{osc}, would further improve usability, as the
        description files only have to be in one place.

      \subsubsection{Assigning in- and outputs on the fly}
        \label{subsubsec:assigning_in_and_outputs_on_the_fly}
        The \gls{ssr}, being a \gls{jack} client, is able to add inputs for its
        sources and outputs for its renderers according to the configuration
        variables \mintinline{yaml}{INPUT_PREFIX} and
        \mintinline{yaml}{OUTPUT_PREFIX}, as shown in
        Listing~\ref{lst:ssr.conf.example}.\\

        \begin{listing}[!htb]
          \begin{mdframed}
            \inputminted[numbers=left, firstline=44, lastline=50,
            fontsize=\footnotesize,
          breaklines]{yaml}{../../ssr/data/ssr.conf.example}
          \end{mdframed}
          \caption{data/ssr.conf.example: \gls{jack} settings in the \gls{ssr}
          configuration file}
          \label{lst:ssr.conf.example}
        \end{listing}\\

        The approach however is somewhat static, as it only allows setting up
        one predefined client during startup. This can be the system's hardware
        in- and outputs or other \gls{jack} clients. The selected input name is
        added to the prefix to connect a \gls{ssr} source to a \gls{jack}
        client output with that name. Following the configuration file example,
        given a source input name of \mintinline{yaml}{1}, the \gls{ssr} would
        connect to the \gls{jack} client port named
        \mintinline{yaml}{alsa_pcm:capture_1}.\\
        Dynamically reassigning source input or renderer output connections is
        only possible by using external tools, able to handle connections of a
        \gls{jack} session, such as QjackCtl \citep{website:qjackctl}, Patchage
        \citep{website:patchage} or aj-snapshot \citep{website:aj-snapshot}.\\
        Every \gls{jack} client is allowed to make connections to other clients
        on the same server on its own. This general feature should be harnessed
        in the case of the \gls{ssr} to allow assigning and reassigning of
        source inputs and renderer outputs on the fly and exposing this
        functionality to the \gls{osc} interface (i.e.\ by modifying and
        extending the \gls{pubsub} interface).\\
        This would make the application more dynamic and allow easier scripting
        of scenes and reproduction setups alike. This becomes especially useful
        in the case of listening experiments, as the experiments will not have
        to rely on a mixture of different scripting languages anymore (e.g.\
        see experimental setup in attachment of \citep{DmitryGrigoriev1896}).

      \subsubsection{Interpolation of moving sources}
        \label{subsubsec:interpolation_of_moving_sources}
        Using the \gls{ip} or \gls{osc} interface, it is possible to move
        sources in a scene to a new location. Unlike sWONDER
        (see~\ref{subsec:swonder}), the \gls{ssr} is not able to interpolate
        movement. When a new location for a source is requested, the movement
        is carried out instantly, whereas sWONDER is able to move a source (on
        a straight line) from one position to the next in a given time frame.\\
        A series of movements can be requested in any desired time frame, which
        means, that spatial aliasing is very likely to occur, during sets of
        requests with a short time span between them.\\
        To work around this, the \gls{ssr} should implement a comparable
        feature to the one sWONDER facilitates and apply rate limiting on the
        source positioning request, depending on a dynamically settable
        threshold value in milliseconds.\\
        Applications, such as WFSCollider (see~\ref{subsec:wfscollider}) or 3Dj
        (see~\ref{subsec:3dj_supercollider_quark}) rely on their \gls{gui} or
        rather \gls{sclang} to implement dynamic movements (e.g.\ circular or
        randomized) \citep[pp. 56-62]{manual:wfscollider}. While the creative
        process of source paths generation is clearly best placed in a visually
        interactive process, its communication with the rendering engine has to
        be high-performance and if scalable to large setups, ideally with low
        network throughput.\\
        Therefore it has to be evaluated, if implementing a set of understood
        geometrical shapes, instead of sending a high frequency of source
        positioning messages could be a more feasible solution in the case of
        the \gls{ssr}. In~\ref{subsubsec:dynamic_scene} a wider approach to
        this problem is discussed.

      \subsubsection{Dynamic scene}
        \label{subsubsec:dynamic_scene}
        When using the \gls{ssr} in a more artistic context, such as musical
        scores, or in experiment environments dealing with moving sources, this
        requires a dynamic scene.\ sWONDER (see~\ref{subsec:swonder}) has a
        scoreplayer application, that can be synced with \gls{midi}, which is
        used to record and play back scores (i.e.\ recorded source properties)
        from unvalidated \gls{xml} files.  WFSCollider has a fully integrated
        timeline, that can be used to place multiple (even concurrent) events,
        save and play them \citep[p. 10]{manual:wfscollider}.\\

        \begin{listing}[!htb]
          \begin{mdframed}
            \inputminted[numbers=left, firstline=84, lastline=98,
            fontsize=\footnotesize,
          breaklines]{xml}{../../ssr/data/asdf.xsd}
          \end{mdframed}
          \caption{data/asdf.xsd: Draft of the score element within the
          \gls{asdf} schema file}
          \label{lst:asdf.xsd-score}
        \end{listing}\\

        Unfortunately the \gls{asdf} was never properly extended for these
        purposes (see Listing~\ref{lst:asdf.xsd-score}), which is why the
        \gls{ssr} is not able to deal with dynamic scenes in a comparable
        fashion yet.\\
        Unlike sWONDER, the \gls{ssr} uses schema validated \gls{xml} only,
        whereas 3Dj and WFSCollider use a unique format or even
        \gls{supercollider} code as score files.\\
        Schema validated input is less error-prone and should generally be
        preferred over single-purpose or self-conceived formats. Therefore, it
        would be a good step to consolidate the \gls{asdf} schema part
        responsible for dynamic elements, while keeping in mind the overall
        message throughput as discussed
        in~\ref{subsubsec:interpolation_of_moving_sources} and thus enable the
        \gls{ssr} to deal with dynamic scenes efficiently.\\
        Additionally, the \gls{gui} efforts made for WFSCollider could be
        combined with a \gls{ssr} backend, as it lacks a tool for creation and
        controlling of dynamic content.

      \subsubsection{Network enabled head tracking}
        \label{subsubsec:network_enabled_headtracking}
        Due to the higher availability of sensors, microcontrollers and
        embedded systems in recent years, it has become very affordable to
        build network enabled head tracking devices in small series. Many of
        the conceived devices, such as the \gls{gpl} licensed \textit{Hedrot}
        \citep{website:hedrot}, allow for \gls{osc} communication.\\
        Using the \gls{osc} interface, such a head tracker can be added as a
        client to a \gls{ssr} instance. This would probably require
        rate-limiting the sensor output, but would enable a networked setup,
        that could prove to be cheaper, more reliable and flexible, than the
        compile-time opt-ins (i.e.\ VRPN, Polhemus Fastrak/ Patriot, InterSense
        InertiaCube3).\\
        In the specific case of setting up a large array of independent
        \gls{brs} or \gls{bs} renderers, connected to one server instance or
        application, it might however be needed to extend the messaging system
        to allow passing on of messages from one client to a server only
        towards one specific other client (a type of proxy messaging). This
        would ensure, that every renderer can be supplied with a specific
        stream of \gls{osc} messages from its assigned head tracker. Single
        (and local) renderers can be started as a server instance and clients
        can be assigned to them flexibly.

  \cleardoublepage
  \pagestyle{empty}
  \cleardoublepage
  \appendix
  \cleardoublepage
  \phantomsection
  \addcontentsline{toc}{section}{References}
  \bibliographystyle{plainnat}
  \bibliography{../bib/ssr-networking}
  \cleardoublepage
  \newpage

  \pagenumbering{Roman}
  \setcounter{page}{1}
  \pagestyle{plain}
  \printglossaries
  \newpage

  \phantomsection
  \addcontentsline{toc}{section}{\listfigurename}
  \listoffigures
  \newpage

  \phantomsection
  \addcontentsline{toc}{section}{\listlistingname}
  \listoflistings
  \newpage

  \phantomsection
  \addcontentsline{toc}{section}{\listtablename}
  \listoftables
  \newpage

  \phantomsection
  \addcontentsline{toc}{section}{Digital Ressources}
  \section*{Digital Ressources}
  \newpage

\end{document}