content: Reorganize to better suit ananke theme

14 files changed, 1574 insertions, 0 deletions

diff --git a/content/docs/_index.md b/content/docs/_index.md
new file mode 100644
index 0000000..f2a842a
--- /dev/null
+++ b/content/docs/_index.md
@@ -0,0 +1,6 @@
+---
+title: Documentation
+date:  2019-06-22
+type:  page
+draft: false
+---
diff --git a/content/docs/compopt.html b/content/docs/compopt.html
new file mode 100644
index 0000000..4c4459c
--- /dev/null
+++ b/content/docs/compopt.html
@@ -0,0 +1,435 @@
+---
+title: "Compilation options"
+date:  2019-06-22
+draft: false
+---
+
+<p>
+  Below is a list of the compile-time configuration options for
+  Ouroboros. These can be set using
+</p>
+<pre><code>$ cmake -D&lt;option&gt;=&lt;value&gt; ..</code></pre>
+<p>or using</p>
+<pre><code>ccmake .</code></pre>
+<p>
+  Options will only show up in ccmake if they are relevant for
+  your system configuration. The default value for each option
+  is <u>underlined</u>. Boolean values will print as ON/OFF in
+  ccmake instead of True/False.
+</p>
+<table>
+  <tr>
+    <th>Option</th>
+    <th>Description</th>
+    <th>Values</th>
+  </tr>
+  <tr>
+    <th colspan="3">Compilation options</th>
+  </tr>
+  <tr>
+    <td>CMAKE_BUILD_TYPE</td>
+    <td>
+      Set the build type for Ouroboros. Debug builds will add some
+      extra logging. The debug build can further enable the
+      address sanitizer (ASan) thread sanitizer (TSan) and leak
+      sanitizer (LSan) options.
+    </td>
+    <td>
+      <u>Release</u>, Debug, DebugASan, DebugTSan, DebugLSan
+    </td>
+  </tr>
+  <tr>
+    <td>CMAKE_INSTALL_PREFIX</td>
+    <td>
+      Set a path prefix in order to install Ouroboros in a
+      sandboxed environment. Default is a system-wide install.
+    </td>
+    <td>
+      &lt;path&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>DISABLE_SWIG</td>
+    <td>
+      Disable SWIG support.
+    </td>
+    <td>
+      True, <u>False</u>
+    </td>
+  </tr>
+  <tr>
+    <th colspan="3">Library options</th>
+  <tr>
+  <tr>
+    <td>DISABLE_FUSE</td>
+    <td>
+      Disable FUSE support, removing the virtual filesystem under
+      &lt;FUSE_PREFIX&gt;.
+    </td>
+    <td>
+      True, <u>False</u>
+    </td>
+  </tr>
+  <tr>
+    <td>FUSE_PREFIX</td>
+    <td>
+      Set the path where the fuse system should be
+      mounted. Default is /tmp/ouroboros.
+    </td>
+    <td>
+      &lt;path&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>DISABLE_LIBGCRYPT</td>
+    <td>
+      Disable support for using the libgcrypt library for
+      cryptographically secure random number generation and
+      hashing.
+    </td>
+    <td>
+      True, <u>False</u>
+    </td>
+  </tr>
+  <tr>
+    <td>DISABLE_OPENSSL</td>
+    <td>
+      Disable support for the libssl library for cryptographic
+      random number generation and hashing.
+    </td>
+    <td>
+      True, <u>False</u>
+    </td>
+  </tr>
+  <tr>
+    <td>DISABLE_ROBUST_MUTEXES</td>
+    <td>
+      Disable
+      <a href="http://pubs.opengroup.org/onlinepubs/9699919799/functions/pthread_mutexattr_getrobust.html">
+	robust mutex
+      </a>
+      support. Without robust mutex support, Ouroboros may lock up
+      if processes are killed using SIGKILL.
+    </td>
+    <td>
+      True, <u>False</u>
+    </td>
+  </tr>
+  <tr>
+    <td>PTHREAD_COND_CLOCK</td>
+    <td>
+      Set the
+      <a href="http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/time.h.html">
+	clock type
+      </a>
+      to use for timeouts for pthread condition variables. Default
+      on Linux/FreeBSD: CLOCK_MONOTONIC.  Default on OS X:
+      CLOCK_REALTIME.
+    </td>
+    <td>
+      &lt;clock_id_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <th colspan="3">Shared memory system options</th>
+  <tr>
+  <tr>
+    <td>SHM_PREFIX</td>
+    <td>
+      Set a prefix for the shared memory filenames. The mandatory
+      leading
+      <a href="http://pubs.opengroup.org/onlinepubs/9699919799/functions/shm_open.html">
+	slash
+      </a>
+      is added by the build system. Default is "ouroboros".
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>SHM_BUFFER_SIZE</td>
+    <td>
+      Set the maximum total number of packet blocks Ouroboros
+      can buffer at any point in time. Must be a power of 2.
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>SHM_RDRB_BLOCK_SIZE</td>
+    <td>
+      Set the size of a packet block. Default: page size of the
+      system.
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>SHM_RDRB_MULTI-BLOCK</td>
+    <td>
+      Allow packets that are larger than a single packet block.
+    </td>
+    <td>
+      <u>True</u>, False
+    </td>
+  </tr>
+  <tr>
+    <td>DU_BUFF_HEADSPACE</td>
+    <td>
+      Set the amount of space to allow for the addition of
+      protocol headers when a new packet buffer is passed to the
+      system. Default: 128 bytes.
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>DU_BUFF_TAILSPACE</td>
+    <td>
+      Set the amount of space to allow for the addition of
+      protocol tail information (CRCs) when a new packet buffer
+      is passed to the system. Default: 32 bytes.
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <th colspan="3">IRMd options</th>
+  </tr>
+  <tr>
+    <td>SYS_MAX_FLOWS</td>
+    <td>
+      The maximum number of flows this Ouroboros system can
+      allocate. Default: 10240.
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>SOCKET_TIMEOUT</td>
+    <td>
+      The IRMd sends commands to IPCPs over UNIX sockets. This
+      sets the timeout for such commands in milliseconds. Some
+      commands can be set independently. Default: 1000.
+    </td>
+    <td>
+      &lt;time_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>BOOTSTRAP_TIMEOUT</td>
+    <td>
+      Timeout for the IRMd to wait for a response to a bootstrap
+      command from an IPCP in milliseconds. Default: 5000.
+    </td>
+    <td>
+      &lt;time_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>ENROLL_TIMEOUT</td>
+    <td>
+      Timeout for the IRMd to wait for a response to an enroll
+      command from an IPCP in milliseconds. Default: 60000.
+    </td>
+    <td>
+      &lt;time_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>CONNECT_TIMEOUT</td>
+    <td>
+      Timeout for the IRMd to wait for a response to a connect
+      command from an IPCP in milliseconds. Default: 5000.
+    </td>
+    <td>
+      &lt;time_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>REG_TIMEOUT</td>
+    <td>
+      Timeout for the IRMd to wait for a response to a register
+      command from an IPCP in milliseconds. Default: 3000.
+    </td>
+    <td>
+      &lt;time_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>QUERY_TIMEOUT</td>
+    <td>
+      Timeout for the IRMd to wait for a response to a query
+      command from an IPCP in milliseconds. Default: 3000.
+    </td>
+    <td>
+      &lt;time_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>IRMD_MIN_THREADS</td>
+    <td>
+      The minimum number of threads in the threadpool the IRMd
+      keeps waiting for commands. Default: 8.
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>IRMD_ADD_THREADS</td>
+    <td>
+      The number of threads the IRMd will create if the current
+      available threadpool is lower than
+      IRMD_MIN_THREADS. Default: 8.
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <th colspan="3">IPCP options</th>
+  </tr>
+  <tr>
+    <td>DISABLE_RAPTOR</td>
+    <td>
+      Disable support for the raptor NetFPGA implementation.
+    </td>
+    <td>
+      True, <u>False</u>
+    </td>
+  </tr>
+  <tr>
+    <td>DISABLE_BPF</td>
+    <td>
+      Disable support for the Berkeley Packet Filter device
+      interface for the Ethernet LLC layer. If no suitable
+      interface is found, the LLC layer will not be built.
+    </td>
+    <td>
+      True, <u>False</u>
+    </td>
+  </tr>
+  <tr>
+    <td>DISABLE_NETMAP</td>
+    <td>
+      Disable <a href="http://info.iet.unipi.it/~luigi/netmap/">netmap</a>
+      support for the Ethernet LLC layer. If no suitable interface
+      is found, the LLC layer will not be built.
+    </td>
+    <td>
+      True, <u>False</u>
+    </td>
+  </tr>
+  <tr>
+    <td>DISABLE_RAW_SOCKETS</td>
+    <td>
+      Disable raw sockets support for the Ethernet LLC layer. If
+      no suitable interface is found,the LLC layer will not be
+      built.
+    </td>
+    <td>
+      True, <u>False</u>
+    </td>
+  </tr>
+  <tr>
+    <td>DISABLE_DDNS</td>
+    <td>
+      Disable Dynamic Domain Name System support for the UDP
+      layer.
+    </td>
+    <td>
+      True, <u>False</u>
+    </td>
+  </tr>
+  <tr>
+    <td>IPCP_SCHED_THR_MUL</td>
+    <td>
+      The number of scheduler threads an IPCP runs per QoS
+      cube. Default is 2.
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>IPCP_QOS_CUBE_BE_PRIORITY</td>
+    <td>
+      Priority for the best effort qos cube scheduler
+      thread. This is mapped to a system value. Scheduler
+      threads have at least half the system max priority value.
+    </td>
+    <td>
+      <u>0</u>..99
+    </td>
+  </tr>
+  <tr>
+    <td>IPCP_QOS_CUBE_VIDEO_PRIORITY</td>
+    <td>
+      Priority for the video qos cube scheduler thread. This is
+      mapped to a system value. Scheduler threads have at least
+      half the system max priority value.
+    </td>
+    <td>
+      0..<u>90</u>..99
+    </td>
+  </tr>
+  <tr>
+    <td>IPCP_QOS_CUBE_VOICE_PRIORITY</td>
+    <td>
+      Priority for the voice qos cube scheduler thread. This is
+      mapped to a system value. Scheduler threads have at least
+      half the system max priority value.
+    </td>
+    <td>
+      0..<u>99</u>
+    </td>
+  </tr>
+  <tr>
+    <td>IPCP_FLOW_STATS</td>
+    <td>
+      Enable statistics for the data transfer component.
+    </td>
+    <td>
+      True, <u>False</u>
+    </td>
+  </tr>
+
+  <tr>
+    <td>PFT_SIZE</td>
+    <td>
+      The forwarding table in the normal IPCP uses a
+      hashtable. This sets the size of this hash table. Default: 4096.
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>IPCP_MIN_THREADS</td>
+    <td>
+      The minimum number of threads in the threadpool the IPCP
+      keeps waiting for commands. Default: 4.
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+  <tr>
+    <td>IPCP_ADD_THREADS</td>
+    <td>
+      The number of threads the IPCP will create if the current
+      available threadpool is lower than
+      IPCP_MIN_THREADS. Default:4.
+    </td>
+    <td>
+      &lt;size_t&gt;
+    </td>
+  </tr>
+</table>
diff --git a/content/docs/development/_index.md b/content/docs/development/_index.md
new file mode 100644
index 0000000..0a5257f
--- /dev/null
+++ b/content/docs/development/_index.md
@@ -0,0 +1,6 @@
+---
+title: Development
+date:  2019-06-22
+#description: Ouroboros development blog
+draft: false
+---
diff --git a/content/docs/documentation.md b/content/docs/documentation.md
new file mode 100644
index 0000000..f48e03c
--- /dev/null
+++ b/content/docs/documentation.md
@@ -0,0 +1,41 @@
+---
+title: "Documentation"
+date:  2019-06-22
+type:  page
+draft: false
+---
+
+# Getting started
+
+* [Requirements](/requirements/)
+* [Download Ouroboros](/download/)
+* [Installing Ouroboros](/install/)
+* [Compilation options](/compopt/)
+
+# User tutorials
+
+These tutorials will be kept up-to-date for the latest version of
+Ouroboros. Check the version that is installed on your system using:
+
+```
+$ irmd --version
+```
+
+The output shown in the tutorials uses a [*debug*](/compopt) build
+of Ouroboros, with FUSE installed and IPCP\_FLOW\_STATS enabled to show
+some additional details of what is happening.
+
+* [Tutorial 1: Local test](/tutorial-1/)
+* [Tutorial 2: Adding a layer](/tutorial-2/)
+* [Tutorial 3: IPCP statistics](/tutorial-3/)
+* [Tutorial 4: Connecting two machines over Ethernet](/tutorial-4/)
+
+# Developer tutorials
+
+* [Developer tutorial 1: Writing your first Ouroboros C program](/dev-tut-1/)
+
+# Extra info
+
+* [Manual pages](/manuals/)
+* [Frequently Asked Questions (FAQ)](/faq/)
+* [Performance tests](/performance/)
diff --git a/content/docs/faq.md b/content/docs/faq.md
new file mode 100644
index 0000000..b3ac687
--- /dev/null
+++ b/content/docs/faq.md
@@ -0,0 +1,122 @@
+---
+title: "Frequently Asked Questions (FAQ)"
+date:  2019-06-22
+draft: false
+---
+
+Got a question that is not listed here? Just pop it on our IRC channel
+or mailing list and we will be happy to answer it!
+
+[What is Ouroboros?](#what)\
+[Is Ouroboros the same as the Recursive InterNetwork Architecture
+(RINA)?](#rina)\
+[How can I use Ouroboros right now?](#deploy)\
+[What are the benefits of Ouroboros?](#benefits)\
+[How do you manage the namespaces?](#namespaces)\
+
+### <a name="what">What is Ouroboros?</a>
+
+Ouroboros is a packet-based IPC mechanism. It allows programs to
+communicate by sending messages, and provides a very simple API to do
+so. At its core, it's an implementation of a recursive network
+architecture. It can run next to, or over, common network technologies
+such as Ethernet and IP.
+
+[[back to top](#top)]
+
+### <a name="rina">Is Ouroboros the same as the Recursive InterNetwork Architecture (RINA)?</a>
+
+No. Ouroboros is a recursive network, and is born as part of our
+research into RINA networks. Without the pioneering work of John Day and
+others on RINA, Ouroboros would not exist. We consider the RINA model an
+elegant way to think about distributed applications and networks.
+
+However, there are major architectural differences between Ouroboros and
+RINA. The most important difference is the location of the "transport
+functions" which are related to connection management, such as
+fragmentation, packet ordering and automated repeat request (ARQ). RINA
+places these functions in special applications called IPCPs that form
+layers known as Distributed IPC Facilities (DIFs) as part of a protocol
+called EFCP. This allows a RINA DIF to provide an *IPC service* to the
+layer on top.
+
+Ouroboros has those functions in *every* application. The benefit of
+this approach is that it is possible to multi-home applications in
+different networks, and still have a reliable connection. It is also
+more resilient since every connection is - at least in theory -
+recoverable unless the application itself crashes. So, Ouroboros IPCPs
+form a layer that only provides *IPC resources*. The application does
+its connection management, which is implemented in the Ouroboros
+library. This architectural difference impact the components and
+protocols that underly the network, which are all different from RINA.
+
+This change has a major impact on other components and protocols. We are
+preparing a research paper on Ouroboros that will contain all these
+details and more.
+
+[[back to top](#top)]
+
+### <a name="deploy">How can I use Ouroboros right now?</a>
+
+At this point, Ouroboros is a useable prototype. You can use it to build
+small deployments for personal use. There is no global Ouroboros network
+yet, but if you're interested in helping us set that up, contact us on
+our channel or mailing list.
+
+[[back to top](#top)]
+
+### <a name="benefits">What are the benefits of Ouroboros?</a>
+
+We get this question a lot, and there is no single simple answer to
+it.  Its benefits are those of a RINA network and more. In general, if
+two systems provide the same service, simpler systems tend to be the
+more robust and reliable ones. This is why we designed Ouroboros the
+way we did. It has a bunch of small improvements over current networks
+which may not look like anything game-changing by themselves, but do
+add up. The reaction we usually get when demonstrating Ouroboros, is
+that it makes everything really really easy.
+
+Some benefits are improved anonymity as we do not send source addresses
+in our data transfer packets. This also prevents all kinds of swerve and
+amplification attacks. The packet structures are not fixed (as the
+number of layers is not fixed), so there is no fast way to decode a
+packet when captured "raw" on the wire. It also makes Deep Packet
+Inspection harder to do. By attaching names to data transfer components
+(so there can be multiple of these to form an "address"), we can
+significantly reduce routing table sizes.
+
+The API is very simple and universal, so we can run applications as
+close to the hardware as possible to reduce latency. Currently it
+requires quite some work from the application programmer to create
+programs that run directly over Ethernet or over UDP or over TCP. With
+the Ouroboros API, the application doesn't need to be changed. Even if
+somebody comes up with a different transmission technology, the
+application will never need to be modified to run over it.
+
+Ouroboros also makes it easy to run different instances of the same
+application on the same server and load-balance them. In IP networks
+this requires at least some NAT trickery (since each application is tied
+to an interface:port). For instance, it takes no effort at all to run
+three different webserver implementations and load-balance flows between
+them for resiliency and seamless attack mitigation.
+
+The architecture still needs to be evaluated at scale. Ultimately, the
+only way to get the numbers, are to get a large (pre-)production
+deployment with real users.
+
+[[back to top](#top)]
+
+### <a name="namespaces">How do you manage the namespaces?</a>
+
+Ouroboros uses names that are attached to programs and processes. The
+layer API always uses hashes and the network maps hashes to addresses
+for location. This function is similar to a DNS lookup. The current
+implementation uses a DHT for that function in the ipcp-normal (the
+ipcp-udp uses a DynDNS server, the eth-llc and eth-dix use a local
+database with broadcast queries).
+
+But this leaves the question how we assign names. Currently this is
+ad-hoc, but eventually we will need an organized way for a global
+namespace so that application names are unique. If we want to avoid a
+central authority like ICANN, a distributed ledger would be a viable
+technology to implement this, similar to, for instance, namecoin.
diff --git a/content/docs/manuals.md b/content/docs/manuals.md
new file mode 100644
index 0000000..4ce3bce
--- /dev/null
+++ b/content/docs/manuals.md
@@ -0,0 +1,18 @@
+---
+title: "Manuals"
+date:  2019-06-22
+draft: false
+---
+
+These are the man pages for ouroboros. If ouroboros is installed on your
+system, you can also access them using "man".
+
+For general use of Ouroboros, refer to the [Ouroboros User
+Manual](/man/man8/ouroboros.8.html).
+
+For use of the API, refer to the [Ouroboros Programmer's
+Manual](/man/man3/flow_alloc.3.html).
+
+The man section also contains a
+[tutorial](man/man7/ouroboros-tutorial.7.html) and a
+[glossary](man/man7/ouroboros-glossary.7.html).
diff --git a/content/docs/performance.md b/content/docs/performance.md
new file mode 100644
index 0000000..5cd2dc0
--- /dev/null
+++ b/content/docs/performance.md
@@ -0,0 +1,74 @@
+---
+title: "Performance tests"
+date:  2019-06-22
+draft: false
+---
+
+Below you will find some measurements on the performance of Ouroboros.
+
+### Local IPC performance test
+
+This test uses the *oping* tool to measure round trip time. This tools
+generates traffic from a single thread. The server has a single thread
+that handles ping requests and sends responses.
+
+```
+$ oping -n oping -i 0 -s <sdu size>
+```
+
+The figure below shows the round-trip-time (rtt) in milliseconds (ms)
+for IPC over a local layer for different packet sizes, measured on an
+Intel Core i7 4500U (2 cores @ 2.4GHz). For small payloads (up to 1500
+bytes), the rtt is quite stable at around 30 µs. This will mostly depend
+on CPU frequency and to a lesser extent the OS scheduler.
+
+![Ouroboros local rtt](/images/avgrttlocal.png)
+
+This test uses the *ocbr* tool to measure goodput between a sender and
+receiver. The sender generates traffic from a single thread. The
+receiver handles traffic from a single thread. The performance will
+heavily depend on your system's memory layout (cache sizes etc). This
+test was run on a Dell XPS13 9333 (2013 model).
+
+```
+$ ocbr -n ocbr -f -s <sdu size>
+```
+
+![Ouroboros local pps](/images/throughputlocalpps.png)
+
+The goodput (Mb/s) is shown below:
+
+![ouroboros local mbits](/images/goodputlocalmbits.png)
+
+### Ethernet + Normal test
+
+This connects 2 machines over a Gb LAN using the eth-dix and a normal
+layer. The oping server is registered in the dix as oping.dix and in the
+normal as oping.normal. The machines (dual-socket Intel Xeon E5520) are
+connected over a non-blocking switch.
+
+Latency test:
+
+ICMP ping:
+
+```
+--- 192.168.1.2 ping statistics ---
+1000 packets transmitted, 1000 received, 0% packet loss, time 65ms
+rtt min/avg/max/mdev = 0.046/0.049/0.083/0.002 ms, ipg/ewma 0.065/0.049 ms
+```
+
+oping over eth-dix:
+
+```
+--- oping.dix ping statistics ---
+1000 SDUs transmitted, 1000 received, 0% packet loss, time: 66.142 ms
+rtt min/avg/max/mdev = 0.098/0.112/0.290/0.010 ms
+```
+
+oping over eth-normal:
+
+```
+--- oping.normal ping statistics ---
+1000 SDUs transmitted, 1000 received, 0% packet loss, time: 71.532 ms
+rtt min/avg/max/mdev = 0.143/0.180/0.373/0.020 ms
+```
\ No newline at end of file
diff --git a/content/docs/quickstart.md b/content/docs/quickstart.md
new file mode 100644
index 0000000..a1bb44b
--- /dev/null
+++ b/content/docs/quickstart.md
@@ -0,0 +1,11 @@
+---
+title: "Quick Start"
+linktitle: "Quick Start"
+date:  2019-06-22
+type:  page
+draft: false
+description: "Quick Start Guide"
+---
+
+
+This quickstart guide is under construction.
\ No newline at end of file
diff --git a/content/docs/tutorials/_index.md b/content/docs/tutorials/_index.md
new file mode 100644
index 0000000..b35d0b8
--- /dev/null
+++ b/content/docs/tutorials/_index.md
@@ -0,0 +1,5 @@
+---
+title: "Ouroboros Tutorials"
+date:  2019-06-22
+draft: false
+---
\ No newline at end of file
diff --git a/content/docs/tutorials/dev-tut-1.md b/content/docs/tutorials/dev-tut-1.md
new file mode 100644
index 0000000..ceac8b6
--- /dev/null
+++ b/content/docs/tutorials/dev-tut-1.md
@@ -0,0 +1,73 @@
+---
+title: "Developer tutorial 1: Writing your first Ouroboros C program"
+draft: false
+---
+
+This tutorial will guide you to write your first ouroboros program. It
+will use the basic Ouroboros IPC Application Programming Interface. It
+will has a client and a server that send a small message from the client
+to the server.
+
+We will explain how to connect two applications. The server application
+uses the flow_accept() call to accept incoming connections and the
+client uses the flow_alloc() call to connect to the server. The
+flow_accept and flow_alloc call have the following definitions:
+
+```
+int flow_accept(qosspec_t * qs, const struct timespec * timeo);
+int flow_alloc(const char * dst, qosspec_t * qs, const struct
+timespec * timeo);
+```
+
+On the server side, the flow_accept() call is a blocking call that will
+wait for an incoming flow from a client. On the client side, the
+flow_alloc() call is a blocking call that allocates a flow to *dst*.
+Both calls return an non-negative integer number describing a "flow
+descriptor", which is very similar to a file descriptor. On error, they
+will return a negative error code. (See the [man
+page](/man/man3/flow_alloc.html) for all details). If the *timeo*
+parameter supplied is NULL, the calls will block indefinitely, otherwise
+flow_alloc() will return -ETIMEDOUT when the time interval provided by
+*timeo* expires. We are working on implementing non-blocking versions if
+the provided *timeo* is 0.
+
+After the flow is allocated, the flow_read() and flow_write() calls
+are used to read from the flow descriptor. They operate just like the
+read() and write() POSIX calls. The default behaviour is that these
+calls will block. To release the resource, the flow can be deallocated
+using flow_dealloc.
+
+```
+ssize_t flow_write(int fd, const void * buf, size_t count);
+ssize_t flow_read(int fd, void * buf, size_t count); int
+flow_dealloc(int fd);
+```
+
+So a very simple application would just need a couple of lines of code
+for both the server and the client:
+
+```
+/* server side */
+char msg[BUF_LEN];
+int fd = flow_accept(NULL, NULL);
+flow_read(fd, msg, BUF_LEN);
+flow_dealloc(fd);
+
+/* client side */
+char * msg = "message";
+int fd = flow_alloc("server", NULL, NULL);
+flow_write(fd, msg, strlen(msg));
+flow_dealloc(fd);
+```
+
+The full code for an example is the
+[oecho](/cgit/ouroboros/tree/src/tools/oecho/oecho.c)
+application in the tools directory.
+
+To compile your C program from the command line, you have to link
+-lourobos-dev. For instance, in the Ouroboros repository, you can do
+
+```
+cd src/tools/oecho
+gcc -louroboros-dev oecho.c -o oecho
+```
\ No newline at end of file
diff --git a/content/docs/tutorials/tutorial-1.md b/content/docs/tutorials/tutorial-1.md
new file mode 100644
index 0000000..d1ac3c6
--- /dev/null
+++ b/content/docs/tutorials/tutorial-1.md
@@ -0,0 +1,153 @@
+---
+title: "Tutorial 1: local test"
+draft: false
+---
+
+This tutorial runs through the basics of Ouroboros. Here, we will see
+the general use of two core components of Ouroboros, the IPC Resource
+Manager daemon (IRMd) and an IPC Process (IPCP).
+
+![Tutorial 1 setup](/images/ouroboros_tut1_overview.png)
+
+We will start the IRMd, create a local IPCP, start a ping server and
+connect a client. This will involve **binding (1)** that server to a
+name and **registering (2)** that name into the local layer. After that
+the client will be able to **allocate a flow (3)** to that name for
+which the server will respond.
+
+We recommend to open 3 terminal windows for this tutorial. In the first
+window, start the IRMd (as a superuser) in stdout mode. The output shows
+the process id (pid) of the IRMd, which will be different on your
+machine.
+
+```
+$ sudo irmd --stdout
+==02301== irmd(II): Ouroboros IPC Resource Manager daemon started\...
+```
+
+The type of IPCP we will create is a "local" IPCP. The local IPCP is a
+kind of loopback interface that is native to Ouroboros. It implements
+all the functions that the Ouroboros API provides, but only for a local
+scope. The IPCP create function will instantiate a new local IPC
+process, which in our case has pid 2324. The "ipcp create" command
+merely creates the IPCP. At this point it is not a part of a layer. We
+will also need to bootstrap this IPCP in a layer, we will name it
+"local_layer". As a shortcut, the bootstrap command will
+automatically create an IPCP if no IPCP by than name exists, so in this
+case, the IPCP create command is optional. In the second terminal, enter
+the commands:
+
+```
+$ irm ipcp create type local name local_ipcp
+$ irm ipcp bootstrap type local name local_ipcp layer local_layer
+```
+
+The IRMd and ipcpd output in the first terminal reads:
+
+```
+==02301== irmd(II): Created IPCP 2324.
+==02324== ipcpd-local(II): Bootstrapped local IPCP with pid 2324.
+==02301== irmd(II): Bootstrapped IPCP 2324 in layer local_layer.
+```
+
+From the third terminal window, let's start our oping application in
+server mode ("oping --help" shows oping command line parameters):
+
+```
+$ oping --listen
+Ouroboros ping server started.
+```
+
+The IRMd will notice that an oping server with pid 10539 has started:
+
+```
+==02301== irmd(DB): New instance (10539) of oping added.
+==02301== irmd(DB): This process accepts flows for:
+```
+
+The server application is not yet reachable by clients. Next we will
+bind the server to a name and register that name in the
+"local_layer". The name for the server can be chosen at will, let's
+take "oping_server". In the second terminal window, execute:
+
+```
+$ irm bind proc 2337 name oping_server
+$ irm register name oping_server layer local_layer
+```
+
+The IRMd and IPCPd in terminal one will now acknowledge that the name is
+bound and registered:
+
+```
+==02301== irmd(II): Bound process 2337 to name oping_server.
+==02324== ipcpd-local(II): Registered 4721372d.
+==02301== irmd(II): Registered oping_server in local_layer as
+4721372d.
+```
+
+Ouroboros registers name not in plaintext but using a (configurable)
+hashing algorithm. The default hash is a 256 bit SHA3 hash. The output
+in the logs is truncated to the first 4 bytes in a HEX notation.
+
+Now that we have bound and registered our server, we can connect from
+the client. In the second terminal window, start an oping client with
+destination oping_server and it will begin pinging:
+
+```
+$ oping -n oping_server -c 5
+Pinging oping_server with 64 bytes of data:
+
+64 bytes from oping_server: seq=0 time=0.694 ms
+64 bytes from oping_server: seq=1 time=0.364 ms
+64 bytes from oping_server: seq=2 time=0.190 ms
+64 bytes from oping_server: seq=3 time=0.269 ms
+64 bytes from oping_server: seq=4 time=0.351 ms
+
+--- oping_server ping statistics ---
+5 SDUs transmitted, 5 received, 0% packet loss, time: 5001.744 ms
+rtt min/avg/max/mdev = 0.190/0.374/0.694/0.192 ms
+```
+
+The server will acknowledge that it has a new flow connected on flow
+descriptor 64, which will time out a few seconds after the oping client
+stops sending:
+
+```
+New flow 64.
+Flow 64 timed out.
+```
+
+The IRMd and IPCP logs provide some additional output detailing the flow
+allocation process:
+
+```
+==02324== ipcpd-local(DB): Allocating flow to 4721372d on fd 64.
+==02301== irmd(DB): Flow req arrived from IPCP 2324 for 4721372d.
+==02301== irmd(II): Flow request arrived for oping_server.
+==02324== ipcpd-local(II): Pending local allocation request on fd 64.
+==02301== irmd(II): Flow on port_id 0 allocated.
+==02324== ipcpd-local(II): Flow allocation completed, fds (64, 65).
+==02301== irmd(II): Flow on port_id 1 allocated.
+==02301== irmd(DB): New instance (2337) of oping added.
+==02301== irmd(DB): This process accepts flows for:
+==02301== irmd(DB): oping_server
+```
+
+First, the IPCPd shows that it will allocate a flow towards a
+destination hash "4721372d" (truncated). The IRMd logs that IPCPd 2324
+(our local IPCPd) requests a flow towards any process that is listening
+for "4721372d", and resolves it to "oping_server", as that is a
+process that is bound to that name. At this point, the local IPCPd has a
+pending flow on the client side. Since this is the first port_id in the
+system, it has port_id 0. The server will accept the flow and the other
+end of the flow gets port_id 1. The local IPCPd sees that the flow
+allocation is completed. Internally it sees the endpoints as flow
+descriptors 64 and 65 which map to port_id 0 and port_id 1. The IPCP
+cannot directly access port_ids, they are assigned and managed by the
+IRMd. After it has accepted the flow, the oping server enters
+flow_accept() again. The IRMd notices the instance and reports that it
+accepts flows for "oping_server".
+
+This concludes this first short tutorial. All running processes can be
+terminated by issuing a Ctrl-C command in their respective terminals or
+you can continue with the next tutorial.
diff --git a/content/docs/tutorials/tutorial-2.md b/content/docs/tutorials/tutorial-2.md
new file mode 100644
index 0000000..392a659
--- /dev/null
+++ b/content/docs/tutorials/tutorial-2.md
@@ -0,0 +1,297 @@
+---
+title: "Tutorial 2: Adding a layer"
+draft: false
+---
+
+In this tutorial we will add a *normal layer* on top of the local layer.
+Make sure you have a local layer running. The network will look like
+this:
+
+![Tutorial 2 setup](/images/ouroboros_tut2_overview.png)
+
+Let's start adding the normal layer. We will first bootstrap a normal
+IPCP, with name "normal_1" into the "normal_layer" (using default
+options). In terminal 2, type:
+
+```
+$ irm ipcp bootstrap type normal name normal_1 layer normal_layer
+```
+
+The IRMd and IPCP will report the bootstrap:
+
+```
+==02301== irmd(II): Created IPCP 4363.
+==04363== normal-ipcp(DB): IPCP got address 465922905.
+==04363== directory(DB): Bootstrapping directory.
+==04363== directory(II): Directory bootstrapped.
+==04363== normal-ipcp(DB): Bootstrapped in layer normal_layer.
+==02301== irmd(II): Bootstrapped IPCP 4363 in layer normal_layer.
+==02301== irmd(DB): New instance (4363) of ipcpd-normal added.
+==02301== irmd(DB): This process accepts flows for:
+```
+
+The new IPCP has pid 4363. It also generated an *address* for itself,
+465922905. Then it bootstrapped a directory. The directory will map
+registered names to an address or a set of addresses. In the normal DHT
+the current default (and only option) for the directory is a Distributed
+Hash Table (DHT) based on the Kademlia protocol, similar to the DHT used
+in the mainline BitTorrent as specified by the
+[BEP5](http://www.bittorrent.org/beps/bep_0005.html). This DHT will use
+the hash algorithm specified for the layer (default is 256-bit SHA3)
+instead of the SHA1 algorithm used by Kademlia. Just like any
+Ouroboros-capable process, the IRMd will notice a new instance of the
+normal IPCP. We will now bind this IPCP to some names and register them
+in the local_layer:
+
+```
+$ irm bind ipcp normal_1 name normal_1
+$ irm bind ipcp normal_1 name normal_layer
+$ irm register name normal_1 layer local_layer
+$ irm register name normal_layer layer local_layer
+```
+
+The "irm bind ipcp" call is a shorthand for the "irm bind proc" call
+that uses the ipcp name instead of the pid for convenience. Note that
+we have bound the same process to two different names. This is to
+allow enrollment using a layer name (anycast) instead of a specific
+ipcp_name. The IRMd and local IPCP should log the following, just as
+in tutorial 1:
+
+```
+==02301== irmd(II): Bound process 4363 to name normal_1.
+==02301== irmd(II): Bound process 4363 to name normal_layer.
+==02324== ipcpd-local(II): Registered e9504761.
+==02301== irmd(II): Registered normal_1 in local_layer as e9504761.
+==02324== ipcpd-local(II): Registered f40ee0f0.
+==02301== irmd(II): Registered normal_layer in local_layer as
+f40ee0f0.
+```
+
+We will now create a second IPCP and enroll it in the normal_layer.
+Like the "irm ipcp bootstrap command", the "irm ipcp enroll" command
+will create the IPCP if an IPCP with that name does not yet exist in the
+system. An "autobind" option is a shorthand for binding the IPCP to
+the IPCP name and the layer name.
+
+```
+$ irm ipcp enroll name normal_2 layer normal_layer autobind
+```
+
+The activity is shown by the output of the IRMd and the IPCPs. Let's
+break it down. First, the new normal IPCP is created and bound to its
+process name:
+
+```
+==02301== irmd(II): Created IPCP 13569.
+==02301== irmd(II): Bound process 13569 to name normal_2.
+```
+
+Next, that IPCP will *enroll* with an existing member of the layer
+"normal_layer". To do that it first allocates a flow over the local
+layer:
+
+```
+==02324== ipcpd-local(DB): Allocating flow to f40ee0f0 on fd 64.
+==02301== irmd(DB): Flow req arrived from IPCP 2324 for f40ee0f0.
+==02301== irmd(II): Flow request arrived for normal_layer.
+==02324== ipcpd-local(II): Pending local allocation request on fd 64.
+==02301== irmd(II): Flow on port_id 0 allocated.
+==02324== ipcpd-local(II): Flow allocation completed, fds (64, 65).
+==02301== irmd(II): Flow on port_id 1 allocated.
+```
+
+Over this flow, it connects to the enrollment component of the normal_1
+IPCP. It sends some information that it will speak the Ouroboros
+Enrollment Protocol (OEP). Then it will receive boot information from
+normal_1 (the configuration of the layer that was provided when we
+bootstrapped the normal_1 process), such as the hash it will use for
+the directory. It signals normal_1 that it got the information so that
+normal_1 knows this was successful. It will also get an address. After
+enrollment is complete, both normal_1 and normal_2 will be ready to
+accept incoming flows:
+
+```
+==13569== connection-manager(DB): Sending cacep info for protocol OEP to
+fd 64.
+==13569== enrollment(DB): Getting boot information.
+==02301== irmd(DB): New instance (4363) of ipcpd-normal added.
+==02301== irmd(DB): This process accepts flows for:
+==02301== irmd(DB): normal_layer
+==02301== irmd(DB): normal_1
+==04363== enrollment(DB): Enrolling a new neighbor.
+==04363== enrollment(DB): Sending enrollment info (49 bytes).
+==13569== enrollment(DB): Received enrollment info (49 bytes).
+==13569== normal-ipcp(DB): IPCP got address 416743497.
+==04363== enrollment(DB): Neighbor enrollment successful.
+==02301== irmd(DB): New instance (13569) of ipcpd-normal added.
+==02301== irmd(DB): This process accepts flows for:
+==02301== irmd(DB): normal_2
+```
+
+Now that the member is enrolled, normal_1 and normal_2 will deallocate
+the flow over which it enrolled and signal the IRMd that the enrollment
+was successful:
+
+```
+==02301== irmd(DB): Partial deallocation of port_id 0 by process
+13569.
+==02301== irmd(DB): Partial deallocation of port_id 1 by process 4363.
+==02301== irmd(II): Completed deallocation of port_id 0 by process
+2324.
+==02301== irmd(II): Completed deallocation of port_id 1 by process
+2324.
+==02324== ipcpd-local(II): Flow with fd 64 deallocated.
+==02324== ipcpd-local(II): Flow with fd 65 deallocated.
+==13569== normal-ipcp(II): Enrolled with normal_layer.
+==02301== irmd(II): Enrolled IPCP 13569 in layer normal_layer.
+```
+
+Now that normal_2 is a full member of the layer, the irm tool will
+complete the autobind option and bind normal_2 to the name
+"normal_layer" so it can also enroll new members.
+
+```
+==02301== irmd(II): Bound process 13569 to name normal_layer.
+```
+
+![Tutorial 2 after enrolment](/images/ouroboros_tut2_enrolled.png)
+
+At this point, have two enrolled members of the normal_layer. What we
+need to do next is connect them. We will need a *management flow*, for
+the management network, which is used to distribute point-to-point
+information (such as routing information) and a *data transfer flow*
+over which the layer will forward traffic coming either from higher
+layers or internal components (such as the DHT and flow allocator). They
+can be established in any order, but it is recommended to create the
+management network first to achieve the minimal setup times for the
+network layer:
+
+```
+$ irm ipcp connect name normal_2 dst normal_1 comp mgmt
+$ irm ipcp connect name normal_2 dst normal_1 comp dt
+```
+
+The IPCP and IRMd log the flow and connection establishment:
+
+```
+==02301== irmd(DB): Connecting Management to normal_1.
+==02324== ipcpd-local(DB): Allocating flow to e9504761 on fd 64.
+==02301== irmd(DB): Flow req arrived from IPCP 2324 for e9504761.
+==02301== irmd(II): Flow request arrived for normal_1.
+==02324== ipcpd-local(II): Pending local allocation request on fd 64.
+==02301== irmd(II): Flow on port_id 0 allocated.
+==02324== ipcpd-local(II): Flow allocation completed, fds (64, 65).
+==02301== irmd(II): Flow on port_id 1 allocated.
+==13569== connection-manager(DB): Sending cacep info for protocol LSP to
+fd 64.
+==04363== link-state-routing(DB): Type mgmt neighbor 416743497 added.
+==02301== irmd(DB): New instance (4363) of ipcpd-normal added.
+==02301== irmd(DB): This process accepts flows for:
+==02301== irmd(DB): normal_layer
+==02301== irmd(DB): normal_1
+==13569== link-state-routing(DB): Type mgmt neighbor 465922905 added.
+==02301== irmd(II): Established Management connection between IPCP 13569
+and normal_1.
+```
+
+The IPCPs established a management flow between the link-state routing
+components (currently that is the only component that needs a management
+flow). The output is similar for the data transfer flow, however,
+creating a data transfer flow triggers some additional activity:
+
+```
+==02301== irmd(DB): Connecting Data Transfer to normal_1.
+==02324== ipcpd-local(DB): Allocating flow to e9504761 on fd 66.
+==02301== irmd(DB): Flow req arrived from IPCP 2324 for e9504761.
+==02301== irmd(II): Flow request arrived for normal_1.
+==02324== ipcpd-local(II): Pending local allocation request on fd 66.
+==02301== irmd(II): Flow on port_id 2 allocated.
+==02324== ipcpd-local(II): Flow allocation completed, fds (66, 67).
+==02301== irmd(II): Flow on port_id 3 allocated.
+==13569== connection-manager(DB): Sending cacep info for protocol dtp to
+fd 65.
+==04363== dt(DB): Added fd 65 to SDU scheduler.
+==04363== link-state-routing(DB): Type dt neighbor 416743497 added.
+==02301== irmd(DB): New instance (4363) of ipcpd-normal added.
+==02301== irmd(DB): This process accepts flows for:
+==02301== irmd(DB): normal_layer
+==02301== irmd(DB): normal_1
+==13569== dt(DB): Added fd 65 to SDU scheduler.
+==13569== link-state-routing(DB): Type dt neighbor 465922905 added.
+==13569== dt(DB): Could not get nhop for addr 465922905.
+==02301== irmd(II): Established Data Transfer connection between IPCP
+13569 and normal_1.
+==13569== dt(DB): Could not get nhop for addr 465922905.
+==13569== dht(DB): Enrollment of DHT completed.
+```
+
+First, the data transfer flow is added to the SDU scheduler. Next, the
+neighbor's address is added to the link-state database and a Link-State
+Update message is broadcast over the management network. Finally, if the
+DHT is not yet enrolled, it will try to do so when it detects a new data
+transfer flow. Since this is the first data transfer flow in the
+network, the DHT will try to enroll. It may take some time for the
+routing entry to get inserted to the forwarding table, so the DHT
+re-tries a couple of times (this is the "could not get nhop" message
+in the debug log).
+
+Our oping server is not registered yet in the normal layer. Let's
+register it in the normal layer as well, and connect the client:
+
+```
+$ irm r n oping_server layer normal_layer
+$ oping -n oping_server -c 5
+```
+
+The IRMd and IPCP will log:
+
+```
+==02301== irmd(II): Registered oping_server in normal_layer as
+465bac77.
+==02301== irmd(II): Registered oping_server in normal_layer as
+465bac77.
+==02324== ipcpd-local(DB): Allocating flow to 4721372d on fd 68.
+==02301== irmd(DB): Flow req arrived from IPCP 2324 for 4721372d.
+==02301== irmd(II): Flow request arrived for oping_server.
+==02324== ipcpd-local(II): Pending local allocation request on fd 68.
+==02301== irmd(II): Flow on port_id 4 allocated.
+==02324== ipcpd-local(II): Flow allocation completed, fds (68, 69).
+==02301== irmd(II): Flow on port_id 5 allocated.
+==02301== irmd(DB): New instance (2337) of oping added.
+==02301== irmd(DB): This process accepts flows for:
+==02301== irmd(DB): oping_server
+==02301== irmd(DB): Partial deallocation of port_id 4 by process 749.
+==02301== irmd(II): Completed deallocation of port_id 4 by process
+2324.
+==02324== ipcpd-local(II): Flow with fd 68 deallocated.
+==02301== irmd(DB): Dead process removed: 749.
+==02301== irmd(DB): Partial deallocation of port_id 5 by process 2337.
+==02301== irmd(II): Completed deallocation of port_id 5 by process
+2324.
+==02324== ipcpd-local(II): Flow with fd 69 deallocated.
+```
+
+The client connected over the local layer instead of the normal layer.
+This is because the IRMd prefers the local layer. If we unregister the
+name from the local layer, the client will connect over the normal
+layer:
+
+```
+$ irm unregister name oping_server layer local_layer
+$ oping -n oping_server -c 5
+```
+
+As shown by the logs (the normal IPCP doesn't log the flow allocation):
+
+```
+==02301== irmd(DB): Flow req arrived from IPCP 13569 for 465bac77.
+==02301== irmd(II): Flow request arrived for oping_server.
+==02301== irmd(II): Flow on port_id 5 allocated.
+==02301== irmd(II): Flow on port_id 4 allocated.
+==02301== irmd(DB): New instance (2337) of oping added.
+==02301== irmd(DB): This process accepts flows for:
+==02301== irmd(DB): oping_server
+```
+
+This concludes tutorial 2. You can shut down everything or continue with
+tutorial 3.
diff --git a/content/docs/tutorials/tutorial-3.md b/content/docs/tutorials/tutorial-3.md
new file mode 100644
index 0000000..2dd0645
--- /dev/null
+++ b/content/docs/tutorials/tutorial-3.md
@@ -0,0 +1,210 @@
+---
+title: "Tutorial 3: IPCP statistics"
+draft: false
+---
+
+For this tutorial, you should have a local layer, a normal layer and a
+ping server registered in the normal layer. You will need to have the
+FUSE libraries installed and Ouroboros compiled with FUSE support. We
+will show you how to get some statistics from the network layer which is
+exported by the IPCPs at /tmp/ouroboros (this mountpoint can be set at
+compile time):
+
+```
+$ tree /tmp/ouroboros
+/tmp/ouroboros/
+|-- ipcpd-normal.13569
+|   |-- dt
+|   |   |-- 0
+|   |   |-- 1
+|   |   `-- 65
+|   `-- lsdb
+|       |-- 416743497.465922905
+|       |-- 465922905.416743497
+|       |-- dt.465922905
+|       `-- mgmt.465922905
+`-- ipcpd-normal.4363
+    |-- dt
+    |   |-- 0
+    |   |-- 1
+    |   `-- 65
+    `-- lsdb
+        |-- 416743497.465922905
+        |-- 465922905.416743497
+        |-- dt.416743497
+        `-- mgmt.416743497
+
+6 directories, 14 files
+```
+
+There are two filesystems, one for each normal IPCP. Currently, it shows
+information for two components: data transfer and the link-state
+database. The data transfer component lists flows on known flow
+descriptors. The flow allocator component will usually be on fd 0 and
+the directory (DHT). There is a single (N-1) data transfer flow on fd 65
+that the IPCPs can use to send data (these fd's will usually not be the
+same). The routing component sees that data transfer flow as two
+unidirectional links. It has a management flow and data transfer flow to
+its neighbor. Let's have a look at the data transfer flow in the
+network:
+
+```
+$ cat /tmp/ouroboros/ipcpd-normal.13569/dt/65
+Flow established at:       2018-03-07 18:47:43
+Endpoint address:                    465922905
+Queued packets (rx):                         0
+Queued packets (tx):                         0
+
+Qos cube   0:
+ sent (packets):                             4
+ sent (bytes):                             268
+ rcvd (packets):                             3
+ rcvd (bytes):                             298
+ local sent (packets):                       4
+ local sent (bytes):                       268
+ local rcvd (packets):                       3
+ local rcvd (bytes):                       298
+ dropped ttl (packets):                      0
+ dropped ttl (bytes):                        0
+ failed writes (packets):                    0
+ failed writes (bytes):                      0
+ failed nhop (packets):                      0
+ failed nhop (bytes):                        0
+
+<no traffic on other qos cubes>
+```
+
+The above output shows the statistics for the data transfer component of
+the IPCP that enrolled into the layer. It shows the time the flow was
+established, the endpoint address and the number of packets that are in
+the incoming and outgoing queues. Then it lists packet statistics per
+QoS cube. It sent 4 packets, and received 3 packets. All the packets
+came from local sources (internal components of the IPCP) and were
+delivered to local destinations. Let's have a look where they went.
+
+```
+$ cat /tmp/ouroboros/ipcpd-normal.13569/dt/1
+Flow established at:       2018-03-07 18:47:43
+Endpoint address:               flow-allocator
+Queued packets (rx):                         0
+Queued packets (tx):                         0
+
+<no packets on this flow>
+```
+
+There is no traffic on fd 0, which is the flow allocator component. This
+will only be used when higher layer applications will use this normal
+layer. Let's have a look at fd 1.
+
+```
+$ cat /tmp/ouroboros/ipcpd-normal.13569/dt/1
+Flow established at:       2018-03-07 18:47:43
+Endpoint address:                          dht
+Queued packets (rx):                         0
+Queued packets (tx):                         0
+
+Qos cube   0:
+ sent (packets):                             3
+ sent (bytes):                             298
+ rcvd (packets):                             0
+ rcvd (bytes):                               0
+ local sent (packets):                       0
+ local sent (bytes):                         0
+ local rcvd (packets):                       6
+ local rcvd (bytes):                       312
+ dropped ttl (packets):                      0
+ dropped ttl (bytes):                        0
+ failed writes (packets):                    0
+ failed writes (bytes):                      0
+ failed nhop (packets):                      2
+ failed nhop (bytes):                       44
+
+<no traffic on other qos cubes>
+```
+
+The traffic for the directory (DHT) is on fd1. Take note that this is
+from the perspective of the data transfer component. The data transfer
+component sent 3 packets to the DHT, these are the 3 packets it received
+from the data transfer flow. The data transfer component received 6
+packets from the DHT. It only sent 4 on fd 65. 2 packets failed because
+of nhop. This is because the forwarding table was being updated from the
+routing table. Let's send some traffic to the oping server.
+
+```
+$ oping -n oping_server -i 0
+Pinging oping_server with 64 bytes of data:
+
+64 bytes from oping_server: seq=0 time=0.547 ms
+...
+64 bytes from oping_server: seq=999 time=0.184 ms
+
+--- oping_server ping statistics ---
+1000 SDUs transmitted, 1000 received, 0% packet loss, time: 106.538 ms
+rtt min/avg/max/mdev = 0.151/0.299/2.269/0.230 ms
+```
+
+This sent 1000 packets to the server. Let's have a look at the flow
+allocator:
+
+```
+$ cat /tmp/ouroboros/ipcpd-normal.13569/dt/0
+Flow established at:       2018-03-07 18:47:43
+Endpoint address:               flow-allocator
+Queued packets (rx):                         0
+Queued packets (tx):                         0
+
+Qos cube   0:
+ sent (packets):                             1
+ sent (bytes):                              59
+ rcvd (packets):                             0
+ rcvd (bytes):                               0
+ local sent (packets):                       0
+ local sent (bytes):                         0
+ local rcvd (packets):                       1
+ local rcvd (bytes):                        51
+ dropped ttl (packets):                      0
+ dropped ttl (bytes):                        0
+ failed writes (packets):                    0
+ failed writes (bytes):                      0
+ failed nhop (packets):                      0
+ failed nhop (bytes):                        0
+
+<no traffic on other qos cubes>
+```
+
+The flow allocator has sent and received a message: a request and a
+response for the flow allocation between the oping client and server.
+The data transfer flow will also have additional traffic:
+
+```
+$ cat /tmp/ouroboros/ipcpd-normal.13569/dt/65
+Flow established at:       2018-03-07 18:47:43
+Endpoint address:                    465922905
+Queued packets (rx):                         0
+Queued packets (tx):                         0
+
+Qos cube   0:
+ sent (packets):                          1013
+ sent (bytes):                           85171
+ rcvd (packets):                          1014
+ rcvd (bytes):                           85373
+ local sent (packets):                      13
+ local sent (bytes):                      1171
+ local rcvd (packets):                      14
+ local rcvd (bytes):                      1373
+ dropped ttl (packets):                      0
+ dropped ttl (bytes):                        0
+ failed writes (packets):                    0
+ failed writes (bytes):                      0
+ failed nhop (packets):                      0
+ failed nhop (bytes):                        0
+```
+
+This shows the traffic from the oping application. The additional
+traffic (the oping sent 1000, the flow allocator 1 and the DHT
+previously sent 3) is additional DHT traffic (the DHT periodically
+updates). Also note that the traffic reported on the link includes the
+FRCT and data transfer headers which in the default configuration is 20
+bytes per packet.
+
+This concludes tutorial 3.
diff --git a/content/docs/tutorials/tutorial-4.md b/content/docs/tutorials/tutorial-4.md
new file mode 100644
index 0000000..fd7db3a
--- /dev/null
+++ b/content/docs/tutorials/tutorial-4.md
@@ -0,0 +1,123 @@
+---
+title: "Tutorial 4: Connecting two machines over Ethernet"
+draft: false
+---
+
+In this tutorial we will connect two machines over an Ethernet network
+using the eth-llc or eth-dix IPCPs. The eth-llc use of the IEEE 802.2
+Link Layer Control (LLC) service type 1 frame header. The eth-dix IPCP
+uses DIX (DEC, Intel, Xerox) Ethernet, also known as Ethernet II. Both
+provide a connectionless packet service with unacknowledged delivery.
+
+Make sure that you have an Ouroboros IRM daemon running on both
+machines:
+
+```
+$ sudo irmd --stdout
+```
+
+The eth-llc and eth-dix IPCPs attach to an ethernet interface, which is
+specified by its device name. The device name can be found in a number
+of ways, we'll use the "ip" command here:
+
+```
+$ ip a
+1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
+group default qlen 1
+link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
+...
+2: ens3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast
+state UP group default qlen 1000
+link/ether fa:16:3e:42:00:38 brd ff:ff:ff:ff:ff:ff
+...
+3: ens6: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast
+state UP group default qlen 1000
+link/ether fa:16:3e:00:76:c2 brd ff:ff:ff:ff:ff:ff
+...
+```
+
+The output of this command differs between operating systems and
+distributions. The interface we need to use in our setup is "ens3" on
+both machines, but for you it may be something like "eth0" or
+"enp0s7f1" if you are on a wired LAN, or something like "wlan0" or
+"wlp2s0" if you are on a Wi-Fi network. For Wi-Fi networks, we
+recommend using the eth-dix.
+
+Usually the interface you will use is the one that has an IP address for
+your LAN set. Note that you do not need to have an IP address for this
+tutorial, but do make sure the interface is UP.
+
+Now that we know the interfaces to connect to the network with, let's
+start the eth-llc/eth-dix IPCPs. The eth-llc/eth-dix layers don't have
+an enrollment phase, all eth-llc IPCPs that are connected to the same
+Ethernet, will be part of the layer. For eth-dix IPCPs the layers can be
+separated by ethertype. The eth-llc and eth-dix IPCPs can only be
+bootstrapped, so care must be taken by to provide the same hash
+algorithm to all eth-llc and eth-dix IPCPs that should be in the same
+network. We use the default (256-bit SHA3) for the hash and 0xa000 for
+the Ethertype for the DIX IPCP. For our setup, it's the exact same
+command on both machines. You will likely need to set a different
+interface name on each machine. The irm tool allows abbreviated commands
+(it is modelled after the "ip" command), which we show here (both
+commands do the same):
+
+```
+node0: $ irm ipcp bootstrap type eth-llc name llc layer eth dev ens3
+node1: $ irm i b t eth-llc n llc l eth if ens3
+```
+
+Both IRM daemons should acknowledge the creation of the IPCP:
+
+```
+==26504== irmd(II): Ouroboros IPC Resource Manager daemon started...
+==26504== irmd(II): Created IPCP 27317.
+==27317== ipcpd/eth-llc(II): Using raw socket device.
+==27317== ipcpd/eth-llc(DB): Bootstrapped IPCP over Ethernet with LLC
+with pid 27317.
+==26504== irmd(II): Bootstrapped IPCP 27317 in layer eth.
+```
+
+If it failed, you may have mistyped the interface name, or your system
+may not have a valid raw packet API. We are using GNU/Linux machines, so
+the IPCP announces that it is using a [raw
+socket](http://man7.org/linux/man-pages/man2/socket.2.html) device. On
+OS X, the default is a [Berkeley Packet Filter
+(BPF)](http://www.manpages.info/macosx/bpf.4.html) device, and on
+FreeBSD, the default is a
+[netmap](http://info.iet.unipi.it/~luigi/netmap/) device. See the
+[compilation options](/compopt) for more information on choosing the
+raw packet API.
+
+The Ethernet layer is ready to use. We will now create a normal layer
+on top of it, just like we did over the local layer in the second
+tutorial. We are showing some different ways of entering these
+commands on the two machines:
+
+```
+node0:
+$ irm ipcp bootstrap type normal name normal_0 layer normal_layer
+$ irm bind ipcp normal_0 name normal_0
+$ irm b i normal_0 n normal_layer
+$ irm register name normal_layer layer eth
+$ irm r n normal_0 l eth
+node1:
+$ irm ipcp enroll name normal_1 layer normal_layer autobind
+$ irm r n normal_layer l eth
+$ irm r n normal_1 l eth
+```
+
+The IPCPs should acknowledge the enrollment in their logs:
+
+```
+node0:
+==27452== enrollment(DB): Enrolling a new neighbor.
+==27452== enrollment(DB): Sending enrollment info (47 bytes).
+==27452== enrollment(DB): Neighbor enrollment successful.
+node1:
+==27720== enrollment(DB): Getting boot information.
+==27720== enrollment(DB): Received enrollment info (47 bytes).
+```
+
+You can now continue to set up a management flow and data transfer
+flow for the normal layer, like in tutorial 2. This concludes the
+fourth tutorial.