For example, SKI combinators are Turing-equivalent. Regular expressions aren't. The lambda-calculus is. Context-independent grammars probably are. Most data description languages are not.
This is worthy of an HN post because a style-description language would not normally be expected to be Turing-equivalent.
When it became widely known that the C++ template system was Turing-equivalent, it was a shock to many. These meant that the C++ compiler code itself could be exploited to compute arbitrary things such as factorials, or shortest-paths in graphs using A*.
Actually, as the SKI system demonstrates, the threshold is quite low. The core needed components are just "constant" or "if" (they are often equivalent), and "repeat" (equivalent to both "iterate" and "recurse").
There is however a different correspondence between recognizing things and computing functions. If we encode a function as taking a bit string as input and generating a bit string as output, we can turn it into a series of recognition problems as follows: "recognize given S as input, whether bit i of f(S) is 1".
Given all these recognizers, we can compute the function, and given a method to compute the function, we can implement the recognizers.
That is, you can reformulate the question "is f computable" to a series of language recognition problems.
No they are not. Recursively Enumerable grammars are however. A context free grammar can't be turning complete because it can be decided with only a PDA.
Wouldn't it? XSLT, the only other style-description language in anything like common use on the web, has been Turing-complete for a while now (possibly since it started existing; I'm not familiar with the history)...
XSL-FO is the formatting part of XSL (i.e. the bit most like CSS) - I'd be pretty surprised if this was shown to be Turing-equivalent.
HTML and CSS are much more "static" in flavor (hence the OP's use of the term "descriptive") and so it might be more unexpected that they can perform universal computation.