It looks like this is for StringHTMLProperty class. 
http://roundup.hg.sourceforge.net/hgweb/roundup/roundup/file/227e7f8614d
d/roundup/cgi/templating.py#l1233

I see this architecture for the first time, and there is immediately a 
question:
1. Does it allow to implement my own replacement rules?
1.1. for example I want to add tiny bits of link ids/classes for JS
1.2. for example I want to do replacement for my own tokens like 
wiki:Stuff should go to http://wiki.python.org/Stuff

Is there any subclassing or other kind of magic for customization? It is 
completely obscure from the code. I also expected some kind of 
tokenization to occur - it's more reliable than chained regexp rules and 
probably easier to debug.

So perhaps it's worth refactoring these replacements into more generic 
utils stuff? It will be easier to test in standalone mode and perhaps 
there is already snippets in other projects that do the same.