In light of some of the recent discussions regarding the turbogears community contributing to other projects I would like some feedback on a cherryPy issue.
I have been playing with a turbogears based REST interface tp tagging: http://tasty.python-hosting.com/ and ran into a possible bug in the way cherryPy handles mapping URLS. The basic format of the REST url looks like: GET /service/<servicename>/user/<username>/item/<itemname>/ this works great for tagging my own content, such as POST /service/photo_gallery/user/aaron/item/killington_2005/tag/vacation but gets a bit ugly when dealing with tagging arbitrary content. For example lets say I want to tag the following URL: http://www.killington.com/podcast/music/ If we take that item to be one part of the url then out url to post the tags would be something like: http://localhost:9980/service/test/user/aaron/item/http%3A%2F%2Fwww.killington.com%2Fpodcast%2Fmusic%2F/tag/vacation >>> urllib.quote_plus('http://www.killington.com/podcast/music/') 'http%3A%2F%2Fwww.killington.com%2Fpodcast%2Fmusic%2F' seems ok, some potentially long URLs will be passed around, but it looks like a sane plan. Unfortunately cherryPy does not like that and returns a 404: NotFound: 404: The path '/service/test/user/aaron/item/http://www.killington.com /podcast/music/tag/fred' was not found. It looks like cherryPy unquotes the incoming URL and then splits it: around line line 232 of _cphttptools.py # Unquote the path (e.g. "/this%20path" -> "this path"). # http://www.w3.org/Protocols/rfc2616/rfc2616-sec5.html#sec5.1.2 # Note that cgi.parse_qs will decode the querystring for us. path = urllib.unquote(path) and a handful of lines later on gets the path scheme, location, p, pm, q, f = urlparse(path) path = path[len(scheme + "://" + location):] # Save original value (in case it gets modified by filters) request.path = request.originalPath = path so ignoring the fact that I could not use strict rest and go for http://localhost:9980/service/test/user/aaron?item=http%3A%2F%2Fwww.killington.com%2Fpodcast%2Fmusic%2F&tag=vacation or md5 sum the item like: http://del.icio.us/url/8b7fec48fcb35763c9f8e1a8061eb124 or maybe something simpler like soap... Should cherryPy be patched to support the type of URL encoding I need? I read the spec and I think cherry Py is actually doing the right thing but when I run tests in Apache apache seems to not treat %2F as a path separator. Any thoughts on this? 1) Can/Should I legally encode urls like this? 2) Should cherryPy be patched? Thanks, -Aaron
