removed the auto-starting of the session when a previous session exists as the session is now always auto-started

Commits
-------

b4f7ce21 Fixed typo
cb3fd7d5 Changed in phpunit.xml.dist and added testsfor lazy url matcher

Discussion
----------

Changed in phpunit.xml.dist and added testsfor lazy url matcher

* Changed in phpunit.xml.dist make sure that tests will covers only code in src directory.
* Added lazy url matcher

Tests passed: [![Build Status](https://secure.travis-ci.org/l3l0/Silex.png?branch=add-lazy-url-matcher-test)](http://travis-ci.org/l3l0/Silex)

---------------------------------------------------------------------------

by igorw at 2012-06-30T12:45:18Z

Nice 👍

Also, how about adding this to .travis.yml:

    script: phpunit --coverage-text

---------------------------------------------------------------------------

by l3l0 at 2012-06-30T14:39:59Z

@igorw Ok I tried but error appear for php 5.3 cause try to cover all code under src even traits: http://travis-ci.org/#!/l3l0/Silex/jobs/1743756

I removed "--coverage-text" in next commit 1201230

Commits
-------

32ca5b51 Do not care about `mailer` service changes
efd3757d Avoid Swift Mailer initialization if possible

Discussion
----------

Avoid Swift Mailer initialization if possible

The finish() filter flushing spool messages after each request
caused Swift Mailer to be initialized and all services in this
provider to be instantiated.

Doing all this work takes time and is potentially unnecessary,
since only a small number of all requests queue messages for sending.

The time it takes to initialize swiftmailer and all services is between
1 and 2ms on the 2 machines I tested on.
This seems like a big price to pay for just including a provider
that does nothing for most requests.

Until now we tried to detect `mailer` service changes,
and always flushed messages for them, just to be safe.

To keep things simple, we don't do this anymore.
If `mailer` gets replaced by another service, flushing will
fail as expected. People should not replace it.

The finish() filter flushing spool messages after each request
caused Swift Mailer to be initialized and all services in this
provider to be instantiated.

Doing all this work takes time and is potentially unnecessary,
since only a small number of all requests queue messages for sending.

The time it takes to initialize swiftmailer and all services is between
1 and 2ms on the 2 machines I tested on.
This seems like a big price to pay for just including a provider
that does nothing for most requests.

Commits
-------

2a3bb653 Tie configs to firewall name to ensure correct exception listener is used

Discussion
----------

Tie configs to firewall name to ensure correct exception listener is used

Without this, the last declared firewall's exception listener is used for all firewalls

Commits
-------

14da2873 Typo in security docs

Discussion
----------

Typo in security docs

Commits
-------

7ef0e3a4 Correct key name in example

Discussion
----------

Correct key name in WSSE example docs

Commits
-------

33061c8f Recommend composer in README

Discussion
----------

Recommend composer in README

Side note: Please update http://silex.sensiolabs.org/download to use 1.0.* instead of dev-master.

Commits
-------

e408ca41 Pulled standard auth mechanisms in to extendable factories

Discussion
----------

Pull standard auth mechanisms in to extendable factories

I'm not really too sure about this implementation myself, but it got me where I needed to be, in that I can define a factory for a custom authentication provider and list the type in the firewall config, something like this:

``` php
<?php

$app['security.firewalls'] = array(
    'api' => array(
        'pattern' => '^/api',
        'wsse' => true,
        'anonymous' => true,
    ),
);

$app['security.authentication.factory.wsse'] = $app->protect(function ($name, $options) use ($app) {
    $app['security.entry_point.wsse.'.$name] = new WsseEntryPoint(...);
    return array(new WsseListener(...), 'wsse');
});

```

I can't see any particular use case for overriding the standard factories (particularly because the _proto functions are there to override) but they are there if someone comes up with one, and it seemed to make things neater in the block building the firewall map.

I've got enough here so that I can move on with my project, if this approach is good enough, I can refine it as necessary, if not, any guidance would be appreciated and I'll work on something better.