HTTPServer

The HTTPServer intrinsic class implements an HTTP network server. HTTP is the protocol that Web browsers use to communicate with Web sites, so this class lets you write a TADS program that can be accessed via a Web browser, which in turn lets you use a Web browser as the user interface for your game. It also makes it possible to create a multi-user applications with TADS, since the server is designed to handle multiple simultaneous connections.

Headers and library files

To use the HTTPServer class, you must #include <httpsrv.h> in your program. In addition, we recommend that you add the library file tadsnet.t to your build (by adding it to your project .t3m file), since this file defines some helper classes often used with HTTPServer.

Starting a server

Setting up a Web server is easy: you simply create an HTTPServer object, specifying the network address that clients use to connect. Once you create the HTTPServer object, it runs autonomously, handling network connections in the background while your program runs.

The constructor looks like this:

local server = new HTTPServer(hostname, port, uploadLimit);

The hostname argument gives the network address of the server. This is the address on which the server will accept incoming connections, so it's the address that clients will use to connect to the server. The port is an optional integer value giving the port number where the server will listen for connections. port can be omitted entirely, or given as nil, which means that you want the operating system to select a port number for you. We'll go into more detail on the host name and port number below.

uploadLimit is an optional size limit, in bytes, for each uploaded request body. If this is nil or is omitted, there's no size limit. Some HTTP requests can include uploaded content; for example, a POST request includes a message body that describes the HTML form data being posted, and can also include uploaded file attachments. The protocol itself doesn't have any limit to these uploaded messages, so if you don't set a limit via this argument, the client will be able to upload arbitrarily large objects. Uploads consume memory and disk space on the server, so it's a good idea to set a size limit when possible, to prevent errant or malicious clients from overwhelming the server with very large uploads.

If the server is started successfully, the new operator returns a new HTTPServer object representing the server. If an error occurs, it throws a NetException error.

Selecting the address and port number

The host name and port number determine the network address where your server will accept connections.

For the standard TADS Web UI setup, simply use getLaunchHostAddr() as the host name and nil as the port number. This creates a listener on a free port on the same address that the client used to connect to the external Web server that launched the program, or simply on "localhost" if the user launched the program locally as a stand-alone application. We know that this address is reachable from the client because the client had to connect to it to trigger the program launch in the first place.

If you're creating some other kind of application rather than using the TADS Web UI conventions, you might have reasons to choose some other host address and port number.

The host name can be given either as the human-readable name for the computer, or as a decimal IP address, in the form '1.2.3.4'. The host name for the server must always be a name or address that's associated with the same machine where the program is running: in other words, you can't run the program on one machine and start the server on another machine just by entering the other machine's address. The name or address must also already be configured in the operating system: you can't create new host names for the computer just by starting a server on an invented name.

Given all these restrictions, it might seem redundant to make you specify the host name. After all, if you can only use an existing, pre-configured name, why can't the system just look up that pre-configured name for you? The reason is that there might be more than one valid host name and address. The host address isn't actually associated with the computer itself; it's associated with a network adapter, which is the hardware that physically connects the computer to the network. Most modern computers can handle multiple network adapters at once; each adapter has its own network address, so a computer with multiple adapters attached will have multiple network addresses. So the reason that HTTPServer makes you specify the host address is that doing so gives you control over which address to use when multiple adapters are present.

There are three main ways of selecting a custom host address:

• You can use the default name or IP address for the computer, as configured in the operating system. This is the most common case, because most computers only have one network address, meaning there's only the one address to choose from anyway. You can get the default host name with the getHostName() function. You can also get the default IP address with the getLocalIP() function.
• In some cases, you'll only want the program to accept local connections - that is, clients running on the same machine. You might want to do this for testing purposes, or if you're setting up the program as a local service only and you don't want external clients to connect for security reasons. In this case, you can use nil or 'localhost' as the host name. The only connections possible will be from the local machine, connecting to the address "http://localhost:port/".
• If you're running your program on a big network server that has multiple network adapters, you might actually want to run a separate server on each adapter, or at least choose the suitable adapter for this particular server. In this case, you'll want to manually configure your program. One way of doing this is to specifically write the program for a particular host, in which case you'd just hard-code the host name right into the new HTTPServer() call. More likely, though, you'd set up the program to get the host name from some external source, such as the command-line parameters or an external configuration file. This scenario is only for sophisticated users, obviously, so we'll assume you can figure it out on your own if this applies to you.

The port number is how the networking system tells apart different servers running on the same computer. Each server has its own port number: a given port can only be used by only one server at a time.

From the client's perspective, the port number is part of the HTTP URL. For example, to connect to port 9155 on IP address 1.2.3.4, you'd enter the URL "http://1.2.3.4:9155/". You probably haven't that ":port" portion very often in everyday Web browsing, because most ordinary Web servers run on the standard HTTP port, port 80. When you don't include a ":port" element in the address, the browser simply assumes ":80" by default.

There are two ways to select a port number. One is to use a "well known" port. With this approach, you select the port number in advance, when you write your program, choosing a number in the range 0 to 1023. You can then tell users that to connect to your server, they will always use that number you select. The difficulty of this approach is that a given port number can only be used by one server at a time on a given machine, so if you choose a port that some other program happens to be using, one or the other program won't be able to use the port. You can reduce the chances of this by consulting a list of commonly used ports (for example, see the Wikipedia list) and choosing an unused port - but there's still no guarantee, because these lists obviously only include prominent, widely-used software.

The other approach is to let the operating system assign an available port, by specifying nil as the port parameter. This guarantees that you'll get a port, because the OS will always choose an available port. The complication is that the port number will be different each time you run your program, because the OS assigns port numbers arbitrarily. This means that you can't just tell your users in advance which port to use: you'll need some way to communicate the port number to users each time you run the server program.

Handling server requests

The HTTPServer object handles all of the low-level networking details automatically for you, running in separate background threads. However, your program has full control over the high-level services provided by the server. This means that it's up to you to respond to each client request.

HTTP is a simple protocol in which clients connect to the server and send requests; the server answers each request with a reply. Each request/reply pair is essentially a stand-alone unit of conversation.

The HTTPServer object handles each incoming request by creating an HTTPRequest object, and placing this in the network event queue. Your program receives these by calling the getNetEvent() function, which waits for and returns the next network event. The HTTPRequest object provides methods that let you retrieve the request information and send the reply. The basic structure of a TADS program that serves HTTP requests is as follows:

#include <tads.h>
#include <tadsnet.h>
#include <httpsrv.h>
#include <httpreq.h>

main(args)
{
    // start the server
    local server = new HTTPServer(getLocalIP(), nil);
    "HTTP Server listening on port <<server.getPortNum()>>\n";

    // process requests
    for (;;)
    {
        local evt = getNetEvent();
        if (evt.evType == NetEvRequest && evt.evRequest.ofKind(HTTPRequest))
        {
            // send the reply
            evt.evRequest.sendReply('It worked! You asked for <<
               evt.evRequest.getQuery()>>.', 'text/plain', 200);
        }
    }

    // shut down the server
    server.shutdown();
}

This is obviously missing a few things you'd want in a real server (it doesn't have any error handling, for example), but it will actually work as written. You can type it in and run it, and it'll serve pages to your browser. To expand this into a real server, you'd add code to inspect the request parameters to determine what the client is asking for, process the request accordingly, and send a suitable reply. You'd also want to add some error handling so that the server doesn't abort at the first lost network connection. (HTTP is by design a stateless protocol, so dropped connections are common, and shouldn't faze a server.)

As you can see, the design makes it quite easy to write the skeleton of a server, while providing you with essentially unlimited flexibility. The TADS infrastructure doesn't make any assumptions about what the client will request or what the reply will look like - that's all up to you.

HTTPServer methods

getAddress()

getIPAddress()

getPortNum()

shutdown(wait?)

Save, restore, undo

HTTPServer objects are inherently transient. This is because they're associated with live operating system resources (threads and network sockets) that can't be saved in a file. This means that any object properties that refer to HTTPServer objects will be nil when you restore a previously saved session. Also, HTTPServer objects are not affected by Undo operations.

Automatic shutdown

HTTPServer objects work like any others with respect to garbage collection: if an HTTPServer object becomes unreachable, the system will delete it during the next garbage collection pass. When this happens, the object automatically shuts down the listener and any session threads started by the server. This means that you have to ensure that each HTTPServer instance remains reachable for as long as you want the server to continue running. You can accomplish this by storing a reference to the HTTPServer object in a named object's property, for example.

If the server has posted any requests to the network event queue, and the garbage collector deletes the server object after those events have been posted but before you retrieve them via getNetEvent(), the events will not be deleted from the queue - they'll be returned as normal from getNetEvent(). If you use the getServer() method on one of these request objects, it will return nil, since the server object no longer exists. If you send a response to the request, it will fail with a network exception.

Network safety levels

The VM enforces a user-specified security setting called the "network safety level". The user interface for setting the level varies by interpreter; for command-line interpreters, for example, it's controlled by the -ns option when starting the program.

The HTTPServer class is subject to the "server" component of the network safety level. The VM enforces the safety level when you attempt to create an HTTPServer object with operator new. If the safety level doesn't allow the type of server you're trying to create, the new operator throws a NetSafetyException instead of creating the HTTPServer object.

When the server component of the net safety level is set to 0 ("minimum safety"), network access is unrestricted. When the safety level is 1, an HTTPServer can only be created if the host name is "localhost" or "127.0.0.1"; in other words, the server can only be created with local access, from applications on the same machine only. Higher safety levels prohibit creating an HTTPServer at all.