view release on metacpan or  search on metacpan

lib/Apache2/POST200.pod  view on Meta::CPAN

=head1 NAME

Apache2::POST200 - Converting code 200 responses to POST requests to 302

=head1 SYNOPSIS

  LoadModule perl_module ...
  PerlLoadModule Apache2::POST200;

  POST200Storage dbi:mysql:db=db:localhost user password
  POST200Table p200 session data
  PerlOutputFilterHandler Apache2::POST200::Filter
  <Location "/-redirect-">
    SetHandler modperl
    PerlResponseHandler Apache2::POST200::Response
  </Location>

  RewriteEngine On

  # redirect GET/HEAD requests with a matching QUERY_STRING to
  # Apache2::POST200::Response
  RewriteCond %{REQUEST_METHOD} !=POST
  RewriteCond %{QUERY_STRING} ^-redirect-[A-Za-z0-9@=-]{32}$
  RewriteRule . /-redirect- [L,PT]

  # This is needed because some applications forget the
  # action="URL" attribute in their <form>s.
  # If so we get a POST request with a QUERY_STRING appended.
  # If it matches our pattern is must be cut off (the '?' in
  # the RewriteRule).
  RewriteCond %{REQUEST_METHOD} =POST
  RewriteCond %{QUERY_STRING} ^-redirect-[A-Za-z0-9@=-]{32}$
  RewriteRule (.+) $1? [N]

  # yes, it even works on a reverse proxy.
  RewriteRule ^/proxy/(.+) http://other.host.tld/$1 [P]

  # keep mod_alias (ScriptAlias) happy
  RewriteRule . - [PT]

 or

  # This is now a forward proxy setup.

  Listen localhost:8080
  <VirtualHost localhost:8080>
    POST200Storage dbi:mysql:db=db:localhost user password
    POST200Table p200 session data
    PerlOutputFilterHandler Apache2::POST200::Filter
    <Location "/-localhost-8080-">
      SetHandler modperl
      PerlResponseHandler Apache2::POST200::Response
    </Location>

    # defined a prefix instead of "-redirect-"
    POST200Label -localhost-8080-

    RewriteEngine On

    # redirect GET/HEAD requests with a matching QUERY_STRING to
    # Apache2::POST200::Response
    RewriteCond %{REQUEST_METHOD} !=POST
    RewriteCond %{QUERY_STRING} ^-localhost-8080-[A-Za-z0-9@=-]{32}$
    RewriteRule . /-localhost-8080- [L,PT]

    # This is needed because some applications forget the
    # action="URL" attribute in their <form>s.
    # If so we get a POST request with a QUERY_STRING appended.
    # If it matches our pattern is must be cut off (the '?' in
    # the RewriteRule). Then proxy it.
    RewriteCond %{REQUEST_METHOD} =POST
    RewriteCond %{THE_REQUEST} ^\w+\s(https?://[^\s\?]+)\?-localhost-8080-[A-Za-z0-9@=-]{32}(?:\s|$)
    RewriteRule . %1? [P]

    # Proxy all other. This is an alternative to the ProxyRequests
    # statement.
    RewriteCond %{THE_REQUEST} ^\w+\s(https?://\S+)
    RewriteRule . %1 [P]
  </VirtualHost>

=head1 DESCRIPTION

A typical WEB application workflow is often similar to this:

  browser shows a form (1)
        |
        v
  user clicks submit (2)
        |
        v
  browser sends a POST request (3)
        |
        v
  server processes the form and replies with a temporary redirect (4)
        |
        v
  browser follows the redirect (5)
        |
        v
  server replies with HTTP code 200 (6)

Steps 4 and 5 are necessary to let the user reload the page shown without
having the server to reprocess the form.

With this module the workflow is shortened from the point of view of the
WEB server to this:

  browser shows a form (1)
        |
        v
  user clicks submit (2)
        |
        v
  browser sends a POST request (3)
        |
        v
  server processes the form and replies with HTTP code 200 (4)

Apache2::POST200 intercepts the server reply, stores the response
in a database and sends a temporary redirect to the browser. It also
intercepts the following request from the browser and sends the stored
reply.

=head2 How it works

This module inserts an request output filter that looks for replies for
POST requests with a HTTP code of 200. If it finds one it saves the reply
in a database and replaces the complete output with a temporary redirect
(HTTP code 302) to the same URL but with a special marked query string
appended.

When the browser follows the redirect the module recognizes the query
string and routes the request to its own response handler. The handler
then reads the saved page from the database and sends it to the browser.

Well, the request routing is actually done by a tricky translation
handler such as mod_rewrite or Apache2::Translation.

Note: the redirect must go to the same URL because some WEB application
forget the C<action> attribute in their C<E<lt>formE<gt>> definitions.

=head2 Configuration

The module itself is loaded from the Apache configuration file via a
C<PerlLoadModule> directive. It then provides a few configuration
directives of its own. All directives are allowed in server config,
virtual host and directory contexts.

=over 4

=item B<Post200Storage dsn user password>

C<Post200Storage> describes the database to be used. All 3 parameter
are passed to the DBI::connect method, see L<DBI>. User and password
can be omitted if the database supports it.

C<Post200Storage None> disables the output filter. That means replies
with a HTTP code 200 to a POST request are delivered as is.

=item B<Post200Table table key-column data-column>

C<Post200Table> describes the table to be used. The C<key> column must be
able to hold a 41-byte string of printable ascii characters. The key length
may be extented in future versions of this module but a key will always
consist of printable characters.

For best performance create an index on the C<key> column.

The C<data> column must be able to hold a variable size data block. The
maximum size can be limited using C<Post200DataBlockSize>. If
C<Post200DataBlockSize> is not used the size completely depends on your
response handlers. If possible use a BLOB type as C<data> column.

Although not used by the module it makes sense to add a 3rd column to the
table. It should be a timestamp column with the default attribute set to
C<now()>. Without it it's difficult to decide which records can be deleted.

With a MySQL database a suitable table is created by:

 create table p200 (
   session varchar(50) primary key unique not null,
   data blob,
   tm timestamp not null default 'now'
 );
 create index p200_tm_idx on p200(tm);

Deletion of expired pages is best done by a simple cron job, e.g.

 45 * * * * echo 'delete from p200 where now()-tm>3600' | mysql post200

=item B<Post200Label marker>

By means of this marker the response handler recognizes a redirected
request that it is responsible for. When the output filter generates
a query string it starts with the C<marker> as prefix.

If omitted C<-redirect-> is used.

If the module is used on a forward proxy to repair external WEB applications
choose a string here that is very likely to be used only by your proxy.

=item B<Post200Secret secret initvector>

To make sure the key provided by the browser via the query string was
generated by the filter it is encrypted. C<secret> and C<initvector>
are arbitrary strings, see L<Crypt::Blowfish>.

If omitted 2 strings are used that once came out of /dev/random on my
box.

=item B<Post200IpCheck On|Off>

With this directive set the response handler sends a page only to the
same client where the redirect was sent to. This prevents that