Problem is partly solved. Sorry, im not into JIRA, got to communicate via mail. Part 1. wiki_editor.js changes: -------------------------------------------- The escaping of URL's in TinyMCE seems to be part of the problem. I managed to turn it of by changing wiki_editor.js in WikiEditor.prototype.init function add: params["convert_urls"] = true; //true = default params["entity_encoding"] = 'name'; //'name' = default params["urlconverter_callback"] = 'myCustomURLConverter'; //handling URL conversion differently, //see myCustomURLConverter add a new function to wiki_editor.js: function myCustomURLConverter(url, node, on_save) { // Do some custom URL convertion url = TinyMCE.prototype.convertURL(url, node, on_save) url=unescape(url); return url; } What it does is getting the normal tinyMCE escaped URL, an unescape it. After doing so, flush the cache and the nasty changing of links is gone. The entity encoding is another interesting point. changing it to "raw" means turning it of. this way the originaly entered links are preserved. (which in my opinion would be the right way!) But this causes other Problems with IE (see below). Part 2. Problems left: -------------------------------------------- 1) With IE sometimes a js error occures - then it does not save at all (maybe a tinyMCE problem?) 2) Firefox and IE behave differently concerning URL escaping and saving. IE sometimes saves to a wrong URL if the link (in the source) contains a german umlaut (eg Ä) both, IE and Firefox use the same URL for editing, eg: WithGermanChar%C4 Firefox saves under WithGermanChar%C4, but IE saves WithGermanCharÄ -> leading to a "unchanged" WithGermanChar%C4 document. The Problem seems to be all about escaping of non standard chars. This must be an XWIKI problem, it happens with or without wysiwyg. Note: Could this be caused by charset problems? This xwiki is running on a Novell plattform. Tomcat is connected to apache via jk-connector. MyWorkaround ---------------------- 1) Enabling entity_encoding in tinyMCE - which is default anyway, but should not (in my opinion) 2) unescaping the URL's. Minor thing: PDF Export then saves files like "Sandbox_&aum.pdf" which is ugly but not a killer problem. Sascha Ludovic Dubost wrote: