Wikimedia Bugzilla is closed!

At the moment some parser functions/variables output is partially escaped and partially treated as wikitext (using wfEscapeWikiText). This breaks both parsing and the use of the output of one variable as input to another, see:

{{PAGENAME}} on [[*]]
{{urlencode:{{PAGENAME}}}} on [[&]]

The solution I propose is to add a new class of UNIQ tags to the parser. Extensions can "return $parser->armouredText( $text );" where they currently "return wfEscapeWikiText( $text )".

The parser should have two replacements for each armouredText, when calling further parser functions, it replaces it with the original; when outputting, it replaces it with the escaped version (at the same time <nowiki> is re-expanded). The ugliness is that it will require several changes to allow armouredText UNIQs to be part of links/template names/magic words.

A much hackier alternatively is an inclusion prefix or two (cf. msg: int: msgnw:) that could be added such that {{plain:PAGENAME}} returns the unescaped version, while {{armour:PAGENAME}} puts it into one of the existing nowiki UNIQs - but this really shouldn't be exposed to users.