phil ringnalda

Lisa (and to a lesser extent Brad) had a bit of trouble with their various and sundry RSS feeds today, thanks to the way that the new version of encode_xml in MT 2.6 wraps things in a CDATA section rather than using entity encoding, combined with the way that some readers (including but not just limited to Radio) don’t handle RSS with an element that has some CDATA, then some entity escaped PCDATA, then some more CDATA, then maybe a bit more PCDATA.

If you also have an RSS template that puts more than one MT tag into an RSS element, you’ll need to make the same fix they have, taking the encode_xml=”1″ filter out of the MT tags, and adding your own CDATA tag in the template, and even if you are them, you need a plugin to make your feed truly safe, so don’t stop reading yet.

If you had a template with something like

<description><$MTEntryBody encode_xml="1"$>
<$MTEntryMore encode_xml="1"$></description>

then your first step is to add your own CDATA escaping in the template, and stop MT from trying, with

<description><![CDATA[<$MTEntryBody$>
<$MTEntryMore$>]]></description>

which will work just fine for the most part* until the moment that you are casually discussing the use of CDATA escaping and forget to entity encode the > when you say “and ends with ]]>” (or, much more likely, someone who thinks they are funny sticks an unencoded one into your comments, and thus into your comment feed). At that point, your CDATA section ends, and everything after that is invalid XML, right up until the now-invalid extra CDATA end tag. To avoid that, you need a global filter that will just escape that one tag, converting ]]> to ]]&gt;, which is why I wrote my most trivial plugin yet, available here. It adds a global filter, cdata_escape, so that if you have <$MTCommentBody cdata_escape=”1″$>, any ]]> in CommentBody will be escaped, and since that’s too tiny a function for even me to release on it’s own, also adds a filter named smartest_encode_xml** which does exactly what encode_xml used to do: apostrophes, quotes, less-than and greater-than are converted to their XML entity equivalent, and nothing more. Use it when you know damned well that nothing else needs to be escaped, and you don’t want to bother with a call to encode_xml with its check for HTML before it escapes one apostrophe.

* Putting encoded HTML in description is a bad, bad thing that you shouldn’t do, you should put it in <content:encoded> instead: see ten thousand discussions all over the web. Also, if you really must put encoded HTML in the description element, using CDATA rather than entity encoding will break any reader that uses regular expressions rather than a real XML parser, unless it’s very new and thus knows to expect the possibility of CDATA escaping. But too bad for them; I know you’re going to just do whatever you want anyway. Bad you.

** A jab, meant in the kindest way and the best of humor, at Timothy Appnel’s mt-xml-smart plugin which became the current MT encode_xml code.