I nixed that because then you've got scary unstable URLs being linked to. The URL system may change, potentially dramatically so, and they'll no longer work. (But I'm a bit of a stickler about stable URLs on articles.)
Well if http://en.wikipedia.org/XYZ redirects to http://en.wikipedia.org/wiki/XYZ , and http://en.wikipedia.org/wiki/XYZ either is the real content (or a redirect to the real content in the future event of a URL system change, so as to maintain stable URLs), then wouldn't that be okay?
The PHP manual example is particularly dangerous because they don't even issue an HTTP redirect, it just spits the content right back. Bad for caching and for linking.
I agree (and I should have said as much in the original mail). Their implementation seems wrong (bad to echo content, should redirect instead), but the principle (IMHO) of doing something nice with the input is pretty sound. i.e. I was thinking more a HTTP 30x redirect header.
All the best, Nick.