1 files changed, 71 insertions, 5 deletions
diff --git a/guide/internals.md b/guide/internals.md
index 431ca01..0f8adc2 100644
--- a/guide/internals.md
+++ b/guide/internals.md
@@ -4,10 +4,76 @@ Internals
 Architecture
 ------------
 
-@todo Describe.
+Cowboy is a lightweight HTTP server.
 
-Efficiency considerations
--------------------------
+It is built on top of Ranch. Please see the Ranch guide for more
+informations.
 
-@todo Mention that you need to cleanup in terminate especially if you
-used the process dictionary, started timers, started monitoring...
+It uses only one process per connection. The process where your
+code runs is the process controlling the socket. Using one process
+instead of two allows for lower memory usage.
+
+It uses binaries. Binaries are more efficient than lists for
+representing strings because they take less memory space. Processing
+performance can vary depending on the operation. Binaries are known
+for generally getting a great boost if the code is compiled natively.
+Please see the HiPE documentation for more details.
+
+Because querying for the current date and time can be expensive,
+Cowboy generates one `Date` header value every second, shares it
+to all other processes, which then simply copy it in the response.
+This allows compliance with HTTP/1.1 with no actual performance loss.
+
+One process for many requests
+-----------------------------
+
+As previously mentioned, Cowboy only use one process per connection.
+Because there can be more than one request per connection with the
+keepalive feature of HTTP/1.1, that means the same process will be
+used to handle many requests.
+
+Because of this, you are expected to make sure your process cleans
+up before terminating the handling of the current request. This may
+include cleaning up the process dictionary, timers, monitoring and
+more.
+
+Lowercase header names
+----------------------
+
+For consistency reasons it has been chosen to convert all header names
+to lowercase binary strings. This prevents the programmer from making
+case mistakes, and is possible because header names are case insensitive.
+
+This works fine for the large majority of clients. However, some badly
+implemented clients, especially ones found in corporate code or closed
+source products, may not handle header names in a case insensitive manner.
+This means that when Cowboy returns lowercase header names, these clients
+will not find the headers they are looking for.
+
+A simple way to solve this is to create an `onresponse` hook that will
+format the header names with the expected case.
+
+``` erlang
+capitalize_hook(Status, Headers, Body, Req) ->
+    Headers2 = [{cowboy_bstr:capitalize_token(N), V}
+        || {N, V} <- Headers],
+    {ok, Req2} = cowboy_req:reply(Status, Headers2, Body, Req),
+    Req2.
+```
+
+Improving performance
+---------------------
+
+By default the maximum number of active connections is set to a
+generally accepted big enough number. This is meant to prevent having
+too many processes performing potentially heavy work and slowing
+everything else down, or taking up all the memory.
+
+Disabling this feature, by setting the `{max_connections, infinity}`
+protocol option, would give you greater performance when you are
+only processing short-lived requests.
+
+Another option is to define platform-specific socket options that
+are known to improve their efficiency.
+
+Please see the Ranch guide for more information.