views:

1143

answers:

3

I have an API that is dependent on certain state information between requests. As an easy first version of the code, I am simply using PHP session's to store the state information instead of something more advanced (APC, memcache, DB). Throughout my initial testing in a web browser, everything worked perfectly. However, it seems that when clients try to connect through non-browser methods such as Curl or wget, the state information is not being preserved.

Will a PHP session only be created if a browser is requesting the page? I am explicitly starting the session with session_start() as well as naming it before hand with session_name().

An added note. I learned that one of the major problems I was having was that I was naming the session instead of setting the session id via session_id($id); My intention in using session_name() was to retrieve the same session that was previously created, and the correct way to do this is by setting the session_id not the session_name.

It seems that session information will be persisted on the server as noted below (THANK YOU). But to keep this you must pass the session id, or, as in my case, any other id that would uniquely identify the user. Use this id as the session_id and your sessions will function as expected.

+18  A: 

Session Cookies

Remember that HTTP is stateless, so sessions are tracked on your server, but the client has to identify itself with each request. When you declare session_start(), your browser is usually setting a cookie (the "PHP Session Id"), and then identifying itself by sending the cookie value with each request. When a script is called using a request with a session value, then the session_start() function will try to look up the session. To prove this to yourself, notice that sessions die when you clear your cookies.. many will die even as soon as you quit the browser, if the cookie is a "session" cookie (a temporary one). You mentioned that you're naming the session.. take a look in your browser cookies and see if you can find a cookie with the same name.

All of this is to say that cookies are playing an active role in your sessions, so if the client doesn't support cookies, then you can't do a session the way you're currently doing it.. at least not for those alternative clients. A session will be created on the server; the question is whether or not the client is participating.

If cookies aren't an option for your client, you're going to have to find another way to pass a session id to the server. This can be done in the query string, for example, although it's a considered a bit less private to send a session id in this way.

mysite.com?PHPSESSID=10alksdjfq9e

How do to this specifically may vary with your version of PHP, but it's basically just a configuration. If the proper runtime options are set, PHP will transparently add the session id as a query parameter to links on the page (same-source only, of course). You can find the specifics for setting that up on the PHP website.

Sidenote: Years ago, this was a common problem when attempting to implement a session. Cookies were newer and many people were turning off the cookie support in their browsers because of purported security concerns.

Sidenote: @Uberfuzzy makes a good point- Using sessions with curl or wget is actually possible. The problem is that it's less automatic. A user might dump header values into a file and use the values on future requests. curl has some "cookie awareness" flags, which allow you to handle this more easily, but you still must explicitly do it. Then again, you could use this to your advantage. If curl is available on your alternative client, then you can plausibly make the call yourself, using the cookie awareness flags. Refer to the curl manual.

keparo
> but if the client doesn't support cookies (like curl or wget)actually. both support some way of storing/loading a cookie file
Uberfuzzy
Good catch. Thanks.
keparo
+1  A: 

If you call session_start(), then a session will be created if the client isn't in an existing one. If the client doesn't support (or is configured to ignore) the cookies or querystring mechanism used to maintain the session, a new session will be created on every request.

This may bloat your session storage mechanism with unused sessions.

It might be a better idea to only call session_start() when you have something to store in the session (e.g. user login, or something else that robots aren't likely to do), if you feel this is likely to be a problem.

MarkR
A: 

Will a PHP session only be created if a browser is requesting the page?

Short answer: Yes. Sessions were created specifically to solve the HTTP stateless problem by leveraging browser features. APC, memcached, DB, etc. don't matter. Those are just storage methods for the session, and will suffer from the same problem.

Longer answer: The concept of sessions were created to account for the fact that HTTP is a stateless protocol, and it turns out that state's pretty important for a wide variety of software applications.

The most common way of implementing sessions is with cookies. PHP sends the session ID in a cookie, and the browser sends the cookie with the session ID back. This ID is used on the server to find whatever information you've stored in the session. PHP has the capacity to include and read a session ID at the end of a URLs, but this assumes that users will navigate to pages on your site/application by clicking links that include a generated session ID.

In your specific case, it is possible to use cookies with curl (and possibly wget). Curl is a web browser, just one without a GUI. If it's the command line curl program you're using (as opposed to the C library, PHP extension,etc.) read up on the following options

-b/--cookie
-c/--cookie-jar
-j/--junk-session-cookies
Alan Storm