Most ColdFusion applications that require users to be authenticated follow the pattern laid out in the official ColdFusion documentation and in the ColdFusion MX Developer's Guide.

This HTML forms-based method is perfectly serviceable, but as applications become more complicated it's not uncommon to run across issues with session expiration and deep linking/bookmarking. This article describes an alternative method of implementing user authentication that's based on the mature and time-tested authentication scheme laid out in the original HTTP specification. Without making the application significantly more complex, it can route around many of the most common limitations of traditional forms-based authentication.

An Overview of Forms-Based Authentication
To start off, let's take a look at some typical ColdFusion code that uses forms-based authentication from the point-of-view of a Web browser. A typical application might have code in the Application.cfm file that looks something like this:

<cfif NOT IsDefined("SESSION.login")>
    <cfif NOT IsDefined("FORM.login")>
      <cfinclude template="loginForm.cfm">
      <cfabort>
    <cfelse>
      <!--- Process login information --->
    </cfif>
</cfif>

Let's look at the series of requests and responses that occur between a Web browser (WB) and a server (S) when the user requests a page that's protected by this kind of authentication code:

WB: Please send me the page called /index.cfm.
S: OK, here's the page called /index.cfm. (The server sends a page containing a form with text fields for user name and password and a submit button.)
WB: (After the user fills in the form and clicks the submit button) Here's a POST to the page called /index.cfm.
S: OK, I processed the submission and here's the page called /index.cfm.

Notice how the server claimed the two pages it sent were both /index.cfm? That's an example of one problem with forms-based authentication: it misrepresents the resources identified by URLs. To illustrate another problem, take a look at what happens later, after the user takes a break from using the application and the session expires:

WB: Please send me the page called /administrator/addUser.cfm.
S: OK, here's the page called /administrator/addUser.cfm. (The server sends a page containing a form with text fields for user name and password and a submit button.)

Since the session expired, the server responded to the request by providing the login form again. Even if the form provides some information about why the user needs to log in again, the user may become confused, and at the very least is inconvenienced.

There are other ways to implement forms-based authentication (such as using <cflocation> to redirect the browser to the login form instead of using <cfinclude> to send it in place of the requested page), but the basic issues are the same:

1. Forms-based authentication misrepresents the resource requested by the browser. Depending on the implementation, it claims the login form is the resource that was requested, or it sometimes claims that the resource requested is located somewhere else (that "somewhere else" being the location of the login form).
2. It forces the user to understand technical issues like session expiration.
3. Since it's dependent on session cookies, it's not easily scalable to cluster environments and doesn't work if the browser has cookies disabled.
4. Since it's dependent on the end user to read the login form and understand that it's not the resource the user requested, it's not compatible with alternative user agents

More Detail: URLs and HTTP Status Codes
What's all this talk about "misrepresenting the resource," you say? It boils down to URLs and status codes.

Each time the browser requests a page, it specifies the exact resource (page) it's requesting. This resource is identified by the URL and a basic tenet of HTTP is that two requests for the same resource (assuming there's no form submission happening) should return exactly the same result (this is closely related to the property of HTTP requests called "idempotency," the principle that GET requests shouldn't change the state of the resource being requested). When the server sometimes sends a login form and sometimes sends the actual page, it's breaking this one-to-one correspondence between the URL and the resource.

Each time the server responds to a request, it sends a status code that very succinctly tells the Web browser the "gist" of its response. Most of the time (when it's serving ColdFusion pages at least), the server sends the status code 200 ("OK"), which means "this page I'm sending you is exactly what you asked for." If instead of executing the page the user requested, the server is sends a login form, it's lying to the browser when it says it found what the user asked for.

When the ColdFusion server processes a <cflocation> tag, it sends the status code 302 ("Found"), which means "the page you're looking for is temporarily located somewhere else and I'm sending you its location." If the address the server sends isn't actually the page the user requested, but a login form, it's lying again.

These distinctions may seem pedantic. But every Web browser since NCSA Mosaic has been raised speaking HTTP as its native language and there's a surprising amount of potential to be leveraged in modern browsers when you respect the true meaning of the terms in that language. And sometimes when you ignore the rules of HTTP you get burned - a notable example of this was when Google released a beta of its Web Accelerator and lots of developers who had ignored the principle of idempotency found that their Web sites were breaking.

Basic Authentication to the Rescue
All the way back at the dawn of the Web (May 1996, to be specific), when Tim Berners-Lee codified the rules of the Hypertext Transfer Protocol (HTTP/1.0), he and his co-authors considered the problem of how to secure resources on the Web and their answer to the problem has come to be known as HTTP Basic Authentication. With Basic authentication, the initial dialogue between Web browser and server looks more like this:

WB: Please send me the page called /index.cfm.
S: Sorry, that page is part of an application called "MyApplication." You can't see that page without providing a user name and password. Here's a page to tell the user why he can't see the page he asked for.
WB: (Prompts the user for a user name and password.) Please send me the page called /index.cfm - and here are the user's credentials to access that page.
S: OK, I processed the credentials and here's the page called /index.cfm.

Doesn't that look like a much more productive exchange than when the server was telling all sorts of lies about the resources it was sending to the browser? In fact it is more productive and here are a few reasons why:

1. HTTP status codes and URLs are used exactly as they were intended.
2. Once the browser knows the user's credentials for the application, it can continue to provide them when needed without prompting the user again (more on this below). Depending on your application, this might mean the end of session expiration!
3. It's not dependent on keeping login credentials in the session scope, so it even works when browsers have cookies disabled. (Your application may still need to use session-scope variables for other purposes.)
4. It allows alternative clients (such as command-line and automated user agents) to access your application, unlocking the potential for new and exciting uses of the data in your application.

Web Server-Integrated Basic Authentication
Basic authentication doesn't have to be implemented in application code. All the most common Web servers (including Apache and Microsoft's Internet Information Server) allow the administrator to protect resources with Basic authentication by setting some configuration parameters. In some cases, this may be sufficient, but if you want to validate the user login information against an existing user database (the one built into your application, for example), you'll almost certainly find it easier to keep the authentication code integrated with the rest of your application.