Using web.ctx makes this possible. First a little architecture: web.ctx is based on the class threadeddict a.k.a. ThreadedDict. This class creates a dictionary-like object that has attributes specific to the thread process id. This is nice because it lets us use a dictionary-like object when many users are accessing the system simultaneously, and the object will only have the data for the given HTTP request (no data is shared so the object is thread-safe).

web.ctx holds variables for each request that contain specific information to each request such as the client environment variable. Assuming you want to determine what the referring page was for a user accessing a page:

Example
class example:
    def GET(self):
        referer = web.ctx.env.get('HTTP_REFERER', 'http://google.com')
        raise web.seeother(referer)
This code uses web.ctx.env to access the HTTP_REFERER environment variable. If there isn't one, it defaults to google.com. Finally, it redirects the user to the page they came from.

web.ctx is also useful because it can be set by a loadhook. Session data, for example, is set each time a request is handled and the data is stored in web.ctx. Since web.ctx is thread-safe, you can use the session data as if it were a regular python object.