On Wed, Jul 23, 2008 at 12:35 PM, Gregory Maxwell gmaxwell@gmail.com wrote:
Having to read some enormous page every page-load wouldn't be good. It would be better to do the right thing on average per-namespace then use something in the pages to control exceptions.
The ability to noindex namespaces is already in MediaWiki and could be turned on at will. See $wgNamespaceRobotPolicies. $wgArticleRobotPolicies is also relevant, although not really useful unless we have a very short fixed list of exceptions that can be maintained by sysadmins (which we don't).
It wouldn't be too hard to fix bug 8068 and add a __NOINDEX__ keyword, though. We would of course want an __INDEX__ keyword as well. (Probably we can leave out __FOLLOW__ and __NOFOLLOW__ as options. Using the namespace default for all pages should be fine there.) Maybe I'll make that my project for today.