It would be very helpful to disable indexing of some pages (namespaces), or even better, ranges in a page.
A ~~NOINDEX~~ element has been already suggested in the User Forum:
http://forum.dokuwiki.org/thread/4352
An even better way that solves several problems at once would be preprocessing the page markup for all ascii output such as the rel="alternate" type="text/plain" output, css feeds, search output and indexing. Such a preprocessor would remove everything between, say, <hide>...</hide> (might be also named <noascii>). The preprocessor could be extended to skip html-like tags such as <code> as well as {{...}}, [[...]], ~~...~~.
As an example, if I search for 'html code' on the dokuwiki homepage, I don't want to find all the <code> </code> <html> </html> markup, and when searching for the use of 'nolink' I want to see where it is described, not dozens of pages where it is used! Such a preprocessor would solve this.
An important application for <hide> would be wrapping it around <php>...</php>. Obviously, I don't want to have visitors search in these, and I don't want them to read my PHP code (not for copyright reasons, but for security).
We are using DokuWiki as a combination of CMS and intrawiki (it is really great for this purpose!), so we have several occurrences of php in the CMS part.