Hi,
  This is my first problem with anything Python :-) and my first issue.

  Doing in the following:

  urlparse.urljoin( 'http://site.com/', '../../../../path/' )
  'http://site.com/../../../../path/'

  urlparse.urljoin( 'http://site.com/', '/path/../path/.././path/./' )
  'http://site.com/path/../path/.././path/./'

These URLs are normalized to http://site.com/path/ in both Firefox and
Google (the google spider would follow these OK)

  I think the documentation could be improved to point at the
posixpath.py normpath function and how it solves the above. I blogged a
how to:

http://teethgrinder.co.uk/blog/Normalize-URL-path-python/

I hope my bug report is OK. Thanks for all the code :-)

johng@neutralize.com

Just try it this way.
>>> print urlparse.urljoin('http://site.com/', 'path/../path/.././path/./')
http://site.com/path/
>>>

The difference is the inital '/' in the second argument.
Human interpretation is:
Go to http://site.com/ and 1) go to path directory 2) go to one-level
above (/../) which results in site.com again 3) go to path directory 4)
go to one-level above (..) (results site.com )5) Stay in the same
directory (.) 6) goto path 7) stay there (.) 
Final result is http://www.site.com/path/

When you start the path with a '/'
>>> print urlparse.urljoin('http://site.com/', '/path/../path/.././path/./')
http://site.com/path/../path/.././path/./

The RFC (1808) suggests the following.
urlparse.urljoin('http://a/b/c/d','/./g') = <URL:http://a/./g>
The argument is taken as a complete path for the server.


The way to use this would be, this way:

>>> print urlparse.urljoin('http://site.com/', 'path/../path/.././path/./')
http://site.com/path/
>>>

This is not a bug and can be closed.

Btw, Thank you for the exciting report monk.e.boy. :-)
There are many hidden in urlparse,urllib*. I hope you will have fun time
finding them (and fixing them too :)

And one general comment. If the bug is valid, Python official
Documentation cannot be made to  reference a blog site. Instead, a patch
to fix the python doc would itself be welcome.

Not a bug...