On Dec 23, 2010, at 9:29 AM, Jonathan Lundell wrote: > > On Dec 23, 2010, at 9:09 AM, mr.freeze wrote: >> >> Given this url: >> URL(args=1,vars=dict(v='yes'),anchor='end') >> It produces: >> http://server/app/default/index/1#end?v=yes >> Problem: >> request.vars is empty >> >> This url: >> http://server/app/default/index/1?v=yes#end >> request.vars is populated and the anchor works >> >> Shouldn't the anchor be at the end of the query string? I'm running >> the latest trunk. > > Yes; that's an error, per RFC 3986. > > URI = scheme ":" hier-part [ "?" query ] [ "#" fragment ] > > > Not only that, but '?' is legal in a fragment identifier, as is '=', so in > your first example, "end?v=yes" is itself the entire (legal) anchor. > > I wouldn't be terribly surprised if some browsers treated this > inconsistently, but we should be adhering to the RFC here. > > The fix looks easy; change: > > if anchor: > other += '#' + urllib.quote(str(anchor)) > if vars: > other += '?%s' % m_vars > > to: > > if vars: > other += '?%s' % m_vars > if anchor: > other += '#' + urllib.quote(str(anchor)) > > ...and fix the doctest. Though I'm not sure that urllib.quote is right for an > anchor.
Following up: as I read the RFC, the quoting doesn't hurt.
