Added information regarding restrictions

1 files changed, 20 insertions, 19 deletions

diff --git a/README.md b/README.md
index dac882b..32d8a56 100644
--- a/README.md
+++ b/README.md
@@ -1,32 +1,33 @@
-QWebSockets
-===========
-Introduction
-------------
+### Introduction
 QWebSockets is a pure Qt implementation of WebSockets - both client and server.
-It is implemented as a source code module (.pri file), that can easily be embedded into existing Qt projects. It has no other dependencies that Qt.
+It is implemented as a Qt source code module (.pri file), that can easily be embedded into existing Qt projects. It has no other dependencies that Qt.
 
-Features
---------
+### Features
 * Text and binary sockets
 * Frame-based and message-based signals
 * Works through proxies
+* Strict Unicode checking
 
+### Restrictions
+Non-characters (according [Unicode Standard 6.2](http://www.unicode.org/versions/Unicode6.2.0/)) are rejected by QWebSockets, **even if the UTF-8 sequence is valid**.  
+##### Rationale
+The WebSocket specification is talking about _Valid UTF-8 codes and sequences_. Strictly speaking, UTF-xx encodings are reversible. That means, that the [66 non-character codes](http://www.unicode.org/faq/private_use.html#noncharacters) (including `U+FFFE` and `U+FFFF`), are valid UTF-8, and hence are perfectly acceptable within WebSocket text messages.  
+According to the Unicode standard, they SHOULD NOT be used in information interchange, but a [recent corrigendum](http://www.unicode.org/versions/corrigendum9.html) clarifies that non-characters CAN be exchanged.  
+However, non-characters are for internal use, and hence, they are implementation specific (e.g. non-characters can be used to carry meta-information). _They have to be interpreted._  
+When used with `QString`, they are replaced with `U+FFFD - REPLACEMENT CHARACTER`, and rendered - non-standard - as a question mark (this is the `QString` rendering of the non-character `U+FFFD`: �). Browsers keep the control characters untouched (this is the browser rendering of the non-character `U+FDD0`: ).  
+  
+With QWebSockets, text messages are just that: a collection of human-readable characters. Text messages never have to be interpreted to be rendered correctly. In case, you still want to do implementation specific trickery, use binary messages instead. Indeed, if non-characters were allowed in text messages, then every text message has to be parsed, character-per-character, to find out if it contains special control codes, or a protocol should be devised that indicates whether the message contains that kind of control codes. We keep it simple: text is text and nothing more.
 
-Requirements
-------------
+### Requirements
 Qt 5.x
 
-Compliance
-----------
-QWebSockets is compliant with [RFC6455](http://datatracker.ietf.org/doc/rfc6455/?include_text=1) and has been tested with the [Autobahn Testsuite](http://autobahn.ws/testsuite).
-Only tests with **invalid UTF-8 sequences** do **not** pass from the Autobahn Testsuite.
+### Compliance
+QWebSockets is compliant with [RFC6455](http://datatracker.ietf.org/doc/rfc6455/?include_text=1) and has been tested with the [Autobahn Testsuite](http://autobahn.ws/testsuite).  
+Only tests with **invalid Unicode code points** do **not** pass from the Autobahn Testsuite (see [Restrictions](#Restrictions)). 
 
-Missing Features
-----------------
+### Missing Features
 * WSS protocol
-* Checks on valid UTF-8 sequences
 * Extensions and sub-protocols
 
-License
--------
-This code is licensed under LGPL v3.
\ No newline at end of file
+### License
+This code is licensed under LGPL v3.