Content-Base: http://a/b/c/d;p?q=1/2the relative URLs should be resolved as shown below. In this test page, I am particularly interested in testing whether "/" in query information is or is not treated as part of the path hierarchy.
I will need your help testing the examples on multiple browsers. What you need to do is point to the example anchor and compare it to the resolved URL in your browser (most browsers have a feature by which you can see the resolved URL at the bottom of the window/screen when the anchor is active).
We have decided that this behavior is due to an oversight in the original libwww implementation, and it is better to correct the oversight in future parsers than it is to make a nonsensical standard. A note has been added to the URI draft to account for the differences in implementations. This should have no impact on current practice since unescaped "/" is rarely (if ever) used within the query part of a URL, and query parts themselves are rarely used with relative URLs.
RESULTS from g = http://a/b/c/g [R,X,1,2,3,4] ./g = http://a/b/c/g [R,X,1,2,3,4] g/ = http://a/b/c/g/ [R,X,1,2,3,4] /g = http://a/g [R,X,1,2,3,4] //g = http://g [R,X,1,2,3,4] g?y = http://a/b/c/g?y [R,X,1,2,3,4] g?y/./x = http://a/b/c/g?y/./x [R,X] http://a/b/c/g?y/x [1,2,3,4] g?y/../x = http://a/b/c/g?y/../x [R,X] http://a/b/c/x [1,2,3,4] g#s = http://a/b/c/g#s [R,X,1,2,3,4] g#s/./x = http://a/b/c/g#s/./x [R,X,2,3,4] http://a/b/c/g#s/x  g#s/../x = http://a/b/c/g#s/../x [R,X,2,3,4] http://a/b/c/x  ./ = http://a/b/c/ [R,X,1,2,3,4] ../ = http://a/b/ [R,X,1,2,3,4] ../g = http://a/b/g [R,X,1,2,3,4] ../../ = http://a/ [R,X,1,2,3,4] ../../g = http://a/g [R,X,1,2,3,4]