repoze.vhm 0.12

repoze.vhm README
=================

Overview
--------

This package provides middleware and utilities for doing virtual
hosting within a WSGI/Repoze environment. It is particularly useful
within a ``repoze.zope2`` environment, where it may be used as an
alternative to the classic `VirtualHostMonster
<http://www.zope.org/Members/4am/SiteAccess2/info>`_: method of doing
virtual hosting.

Virtual Hosting in a Nutshell
-----------------------------

"Virtual hosting" enables dynamic applications to be served from
within a larger URL namespace, independent of the physical location of
the script files used to serve the application, or the precise layout
of objects within the application. In particular, the application and
the server collaborate to generate URLs for links in the application,
such that the links preserve the "apparent" location of the
application.

The simplest case requires no effort at all: links rendered as
relative paths from within pages work nicely. However, such links
begin to be problematic quickly, e.g. when the page is serving as the
default index view for its folder, and the URL does not end in a
``/``. In that case, the browser interprets the links relative to the
folder's parent, and chaos ensues.

CGI Environment Variables
-------------------------

As used for applications running "inside" Apache (e.g., using
``mod_python``), there follwing environment variables are of interest
when doing virtual hosting:

SERVER_NAME

the name which the server believes it has.

HTTP_HOST

the apparent hostname of the server (i.e., as passed in the ``Host:``
header)

SERVER_PORT

the apparent port of the server

SCRIPT_NAME

any path prefix used by Apache to dispatch to the application (as
defined via the ``ScriptAlias`` directive).

PATH_INFO

the remainder of the path, after removing any parts used in
dispatch.

``repoze.vhm#vhm_xheaders`` WSGI Filter
---------------------------------------

When configured as WSGI middleware, this filter will convert the path
information in the environment from the "X-Vhm" headers added to the
request into the "standard" CGI environment variables outlined above.
It will also place repoze.vhm-specific environment variables into the
WSGI environment for consumption by repoze.zope2 (or another
application which chooses to use its services).

If this filter is placed into the pipeline in front of a Zope 2
application, the standard Virtual Host Monster object
(``/virtual_hosting``) may be deleted, as it is no longer necessary.
However, it does not need to be deleted; repoze.vhm will work if it is
present.

The filter requires no configuration; it can be added to any pipeline
via its egg name: ``egg:repoze.vhm#vhm_xheaders``.

``repoze.vhm#vhm_explicit`` WSGI Filter
---------------------------------------

This filter is like the repoze.vhm#vhm_xheaders filter, but instead of taking
the virtual host and/or root from the environment, they are explicitly
configured when the middleware is instantiated.

If using paste.deploy, this looks like::

[filter:vhm]
use = egg:repoze.vhm#vhm_explicit
host = http://www.example.com
root = /mysite

Both 'host' and 'root' are optional, but you probably want to specify at
least one of them.

``repoze.vhm#vhm_path`` WSGI Filter
-------------------------------

As a fallback for proxies which cannot add headers to proxied
requests, this filter implements the same path-based virtual hosting
syntax used by the Zope2 Virtual Host Monster. Because this syntax is
quite arcane (so much that there is a web-app for generating the
rewrite rules!), this filter is not recommended except for
environments which cannot be configured to add headers (e.g., Apache
has ``mod_rewrite`` enabled, but cannot be changed to enable
``mod_headers``).

When configured as WSGI middleware, this filter will convert the path
information in the environment from the classic "Zope2 virtual hosting
mungned URL" into the "standard" CGI environment variables outlined
above. It will also place repoze.vhm-specific environment variables
into the WSGI environment for consumption by repoze.zope2 (or another
application which chooses to use its services).

If this filter is placed into the pipeline in front of a Zope 2
application, the standard Virtual Host Monster object
(``/virtual_hosting``) may be deleted, as it is no longer necessary.
However, it does not need to be deleted; repoze.vhm will work if it is
present.

The filter requires no configuration; it can be added to any pipeline
via its egg name: ``egg:repoze.vhm#vhm_path``.

repoze.vhm Virtual Hosting Model
--------------------------------

This model (based on a `suggestion of Ian Bicking's ,
<http://blog.ianbicking.org/2007/08/10/defaults-inheritance/>`_),
passes virtual hosting information from the proxy / web server to the
application by adding extra headers to the proxied request:

HTTP_X_VHM_HOST

indicates the apparent URL prefix of the root of the application
(concatenating ``wsgi.url_scheme``, ``SERVER_NAME``,
``SERVER_PORT``, and ``SCRIPT_NAME`` variables; the equivalent of
Zope2's ``SERVER_URL``).

HTTP_X_VHM_ROOT

path of the object within the application which is supposed to
function as the "virtual root".

When serving an application from "within" Apache via mod_wsgi, we can
just set the environment directly::

<Directory /path/to/wsgiapp>
SetEnv HTTP_X_VHM_HOST http://www.example.com
SetEnv HTTP_X_VHM_ROOT /cms
</Directory>

If you are serving repoze.zope2 via a proxy rewrite rule, you may pass
this information by adding additional headers. E.g., a sample Apache
configuration for the example above might be::

<VirtualHost *:80>
ServerName www.example.com
RewriteEngine on
RewriteRule ^/(.*) http://localhost:8080/$1 [P,L]
RequestHeader add X-Vhm-Host http://www.example.com
RequestHeader add X-Vhm-Root /cms
</VirtualHost>

In either of the above example cases, the effect on repoze.zope2 when
repoze.vhm's filter is in the WSGI pipeline is the same: the apparent
root of ``http://www.example.com`` will be the default view of the
object that has a physical path of ``/cms``. Additionally, paths in
URLs generated by Zope will not start with ``/cms``, and the scheme
and hostname in URLs will be "http://www.example.com" as opposed to
``http://localhost:8080``.

The "vhm host" header may contain further path information as
necessary; further path information can (and will, in the case of
repoze.zope2) be respected by downstream applications to root an
application at a non-server-root path ::

<Directory /path/to/wsgiapp>
SetEnv HTTP_X_VHM_HOST http://www.example.com/further/path
SetEnv HTTP_X_VHM_ROOT /cms
</Directory>

In this case, URLs generated by Zope will begin with
``http://www.example.com/further/path``. This syntax replaces the
"inside out" virtual hosting syntax (``_vh_`` segment markers in the
URL) as described in the "Virtual Host Monster" documentation.

The "vhm host" and "vhm root" headers can be used independently (the
system will operate as you would expect in the absence of one or the
other).

``repoze.vhm`` Library API
--------------------------

Because the existing Zope 2 virtual hosting machinery does not rely on
the "standard" CGI variables, the application dispatcher needs to "fix
up" the environment to match Zope's expectations. ``repoze.vhm``
offers the following functions to aid in this fixup:

repoze.vhm.utils.setServerURL

convert the standard CGI virtual hosting environment into the form
expected by Zope2 (adding the ``SERVER_URL`` key).

repoze.vhm.utils.getVirtualRoot

return the virtual root path (``repoze.vhm.virtual_root``) as set
by the middleware.


repoze.vhm Changelog
====================

0.12 (2010-01-01)
-----------------

- Update tests of middleware to check only that ``munge`` is called correctly.

- Test ``munge`` sepearately.

- 100% test coverage.

- fixed xheaders filter to set PATH_INFO correctly
[vangheem]


0.11 (2009-08-31)
-----------------

- Add a repoze.vhm#vhm_explicit filter. This is like the vhm_xheaders
middleware, but the VHM host and/or root are set in the WSGI configuration
instead of in the request.

- Calculate a VIRTUAL_URL and put it into the environment. This is basically
the URL that the end user sees. repoze.zope2 >= 1.0.2 uses this to compute
the ACTUAL_URL request variable, for example.

0.10 (2009-08-26)
-----------------

- Apply the HTTP_HOST port number fix to the VHM Path filter as well.

0.9 (2009-07-09)
----------------

- 100% test coverage.

- ``HTTP_HOST`` parameter now includes port number if not http:80 or
https:443. Thanks to Martin Aspeli.

0.8 (2009-01-10)
----------------

- Set 'HTTP_HOST' in environ to the same value as 'SERVER_NAME', FBO apps
which need it.

0.7 (2008-05-07)
----------------

- Remove 'dependency-links=' to dist.repoze.org to prevent easy_install
from searching there inappropriately.

0.6 (2008-04-17)
----------------

- Re-added the path-segment-based filter as an option, to support scenarios
in which the reverse proxy can be configured to rewrite the URL but not
to add headers.

0.5 (2008-03-09)
----------------

- Brown bag release: I fudged the entry point for the xheaders filter.

0.4 (2008-03-09)
----------------

- Kill off path-segment-based filter (repoze.vhm.zope2). Only the
xheaders filter remains.

- Add license headers.

- The middleware now sets a 'repoze.vhm.virtual_host_base' which is
preferred by setServerUrl over 'HTTP_HOST' when present.

- Add a getVirtualRoot API.

0.3 (2007-10-25)
----------------

- Fix setServerURL method to take into account HTTP_HOST passed by
client.

0.2 (2007-09-22)
----------------

- Change repoze.vhm.zope2:setServerURL to allow Zope 2 to generate the
correct request['URL'] value when the vhm is in the pipeline.

0.1 (2007-09-21)
----------------

- Initial release.