> >     # location: '/app.cgi'
> >     SCRIPT_NAME: ''
> >     PATH_INFO: '/app.cgi'
>
> Ugh.  I want Apache + app.cgi to provide
> SCRIPT_NAME: '/app.cgi'
> and
> PATH_INFO: ''
> In this case.

That would make a lot more sense.

> If you use:
>     ScriptAlias /foo 
> do you get the same results?

Yes, for the request '/foo' it gives:
    SCRIPT_NAME: ''
    PATH_INFO: '/foo'


> I think you should not be using SCRIPT_FILENAME at all, since it is not
> part of the CGI 1.1 specification.

My thinking there is that no harm can be done, except that if it is not present
then nothing will change.

> In your case, it looks like this method could be:
>
>      def process (self, stdin, env):
>          if env.get('PATH_INFO') == '':
>              env['PATH_INFO'] = env.get('SCRIPT_NAME')
>              env['SCRIPT_NAME'] = ''
>          hit = Hit(stdin, env)
>          self.process_hit(hit)
>          return hit
>
> This hack also frustrates me though.  The cgi-caller really should
> provide a SCRIPT_NAME that makes sense.

Hmmn, in my case PATH_INFO is never an empty string, so this will never match.
Or did you mean to switch the 2 vars here? I.e. this works in this root case
(but will clearly not in the general case) :

    def process (self, stdin, env):
        if env.get('SCRIPT_NAME') == '':
            env['SCRIPT_NAME'] = env.get('PATH_INFO')
            env['PATH_INFO'] = ''
        hit = Hit(stdin, env)
        self.process_hit(hit)
        return hit

>
> Another idea:
>
> Add this to your root directory:
>
>       def _q_lookup(self, component):
>           redirect('')
>
> Wouldn't this redirect all unknown paths, like '/app.cgi', to '/'?

I tried this, after removing my overriden publisher.process():

    def _q_lookup (self, component):
        r = super(MySiteDirectory, self)._q_lookup(component)
        if r is not None:
            return r
        get_publisher().redirect('')

When requesting '/app.cgi' this gives a redirect loop... I guess the redirect
should be:

        env = get_publisher().get_hit().get_request().environ
        print '%s:%s:%s' % (component, env['SCRIPT_NAME'],
                env.get('PATH_INFO'))
        get_publisher().redirect(env['PATH_INFO']+'/')

Interestingly, when requesting '/app.cgi' the above prints:
app.cgi::/app.cgi
This must simply be because the result of the PATH_INFO is being screwed here.

In the end I am leaning towards overriding process_inputs() instead... as this
will also redirect correctly. The _q_lookup() trick has the advantage of being
executed only when needed, not on every hit, but i find it more convoluted.

    # Handle case when site is deployed under a script, but the requested
    # url ends with precisely the SCRIPT_NAME (with nothing trailing it).
    # This erroneously (on Apache2) gives an empty string SCRIPT_NAME and
    # a PATH_INFO set to the actual script_name -- that will break any
    # dynamically calculated relative url paths.
    # Assumption here is that if a SCRIPT_FILENAME is set, then it must
    # end with the (non-empty) SCRIPT_NAME.
    def process_inputs (self):
        env = self.get_hit().get_request().environ
        if env.get('SCRIPT_FILENAME'):
            if not env.get('SCRIPT_NAME') and env.get('PATH_INFO'):
                if env['SCRIPT_FILENAME'].endswith(env['PATH_INFO']):
                    self.redirect(env['PATH_INFO'] + '/')
        super(Publisher, self).process_inputs()


Yes, it relies on SCRIPT_FILENAME... but it is a hack a fix servers returning an
incorrect SCRIPT_NAME in this special situation. Maybe servers that do not
supply the SCRIPT_FILENAME would anyway return the correct SCRIPT_NAME ?!?

Related question: is the current check in process() still needed? WHich
situations (depoyment scenario, server, os?) trigger it to be executed? The
long comment for it does not seem to indicate that...

mario