The str_replace call removes occurrences of target="_blank" everywhere. This means that a blog post that explains the benefits of, say, this transcoding library has no easy way to write:

[[ Very cool, this plug-in removes target="_blank" from content to suppress popups. ]]

The real solution would be to parse content as a DOM tree, but that may fail if content is not valid XML.

Another workaround is to make sure the matched string contains a "<" character, because "<" cannot appear in regular content (it has to be escaped as "&lt;").

But... as defined in the XML spec, there are exceptions to the rule:
 http://www.w3.org/TR/xml/#syntax

"<" may appear in:
 1. comments
 2. CDATA sections
 3. Processing instructions (but I'm not too worried about that possibility as processing instructions are usually found near the top of the document).

Can we come up with a way to account for these exceptions while sticking to a simple search and replace paradigm?