I was reading How to tell if you're a New Zealander (it turns out I'm not one), and I noticed an extremely small typo: “If you're a women”. Obviously it should read woman, but the typo is so small it's not worth sending in a fix to the author, Mark Rosenfelder. The problem is that the effort required to fix the problem is comparatively much larger than the very small inconvenience that readers face of having to fix the typo in their heads.
Could we make the problem easier to fix? I think so. If I downloaded this file locally and fixed the typo myself, then took a diff of it, it'd come out as something like this:
$ diff -Naur orig.nztest.html nztest.html --- orig.nztest.html 2006-12-01 15:08:23.000000000 +0000 +++ nztest.html 2006-12-13 17:55:34.000000000 +0000 @@ -91,7 +91,7 @@ <li>You call everyone by their first name, from the Prime Minister down. -<li>If you're a women, you certainly wouldn't go to the beach topless. +<li>If you're a woman, you certainly wouldn't go to the beach topless. <font color="#808080"><li>You seriously expect to be able to transact business, or deal with the government, without paying bribes.</font>
I could send this off to Rosenfelder, but this wouldn't save much time: patching the file would take about as long as editing it in vi or emacs. Moreover, it's taken about two or three minutes just to make the patch, whereas reporting the URI and quoting the fragment which is broken would only take say thirty seconds.
So the idea is: automate the diff system. These little niggling typos occur all the time on Wikipedia, and sometimes I fix them even though it's a bit of a pain having to sign in and so on. Pedants will go to great lengths to correct typos; authors won't. When I feel like an author, I won't even correct my own typos.
I envisage some bit of Ajax Bookmarklet wizardry that will let you fix very simple typos using some kind of DOM interface, and then will prepare a textual diff and send it off to $location. The problem is that it would only work on pages where you can figure out a $location. It might be some HTTP POST form, or an email address, or whatever, but then something on the author's side should make it very easy to integrate the errors. This is rather like Annotea, only for a very specific purpose, so it might be possible to leverage off of the centralised annotation database principle.
Sean B. Palmer, 2006-12-13