As I wrote previously, a WebSocket API is only the starting point for WebSocket-style messaging applications. Many practical challenges remain. As one Tomcat mailing list user mused recently:

it does seem to me that websockets is still not really "production-ready", (I am not talking about the Tomcat implementation per se, but more generally) ... native websockets capability in IE is only available since IE-10 and that solutions which allow this to work in lower IE versions are a bit "iffy" (relying on a diversion through Adobe's FlashPlayer e.g.). (Most of our customers are largish corporations, which are not going to update their browsers, nor open special ports in their firewalls, just to please us).

…

As you may have seen, the first milestone of Spring Framework 4.0 was already announced and with it we've released early WebSocket support. Why WebSocket matters? It enables efficient, two-way communication over the web that is essential in applications where messages need to be exchanged between client (typically browser) and server at high frequency and with low latency. Common examples include trading, games, collaboration, data visualization, and others but the range of scenarios and use cases will grow over time.

WebSocket is a very broad subject! You can watch our “Intro to WebSocket” on…

A minor maintenance release of Spring Web Flow 2.3.2 is now available via Maven and for download. See the Changelog for the list of changes.

A first milestone of Spring Web Flow 2.4 is also available through the SpringSource milestone repository. See the Changelog for the full list of changes.

Spring Web Flow samples have been separated from the distribution and into a separate Github project. In addition the booking-mvc sample has been updated to use Thymeleaf thanks to Thymeleaf's project lead Daniel Fernández.

[callout title=Update Dec 19, 2012] The final Spring Framework reference documentation contains guidance on migration as well as a complete section on Spring MVC Test. [/callout]

Last week Juergen Hoeller announced the release of Spring Framework 3.2 RC1 and Sam Brannen discussed exciting additions in its spring-test module such as support for WebApplicationContext's and upcoming plans for loading a hierarchy of contexts. Today I will continue this subject and describe another exciting spring-test addition. In 3.2 RC1 we've added first class support for testing Spring MVC applications both…

A second milestone of the Spring MVC Test project is now available through the SpringSource Artifactory repository http://repo.springsource.org/milestone.

There is much that's new. On the server side we've added support for servlet filters, basic support for Spring Security, the ability to define default request properties and expectations to re-use across every performed request, and much more. On the client side we've greatly expanded the options for request matching and response creation.

For further details and comments read this blog post.

Last updated on November 5th, 2012 (Spring MVC 3.2 RC1)

In previous blog posts I introduced the Servlet 3 based async capability in Spring MVC 3.2 and used the spring-mvc-showcase and the Spring AMQP stocks sample to demonstrate it. This post presents a chat sample where the external events are not AMQP messages but rather HTTP POST requests with chat messages. In the second part of the post, I'll switch to a distributed chat where the events are Redis notifications.

Chat is not a common requirement for web applications. However it is a good example of a requirement that can only be met with real-time notifications. It is more sensitive to time delays than email or status alerts and it is not that uncommon to chat in a browser with a friend, or with a colleague during a webinar, or with a live person on a shopping site. You can imagine other types of online collaboration.

The Sample

The spring-mvc-chat sample is available on Github. Although not the focus of this blog post, the client side uses Thymeleaf, knockout.js, and jQuery. Thymeleaf is an excellent alternative to JSPs that enables clean HTML templates with support for previews allowing a designer to double-click an HTML template and view it unlike a JSP that requires a Servlet container. knockout.js is a client-side MVC framework that's very handy for attaching behavior to HTML elements. To get an idea about it quickly, follow one of its excellent tutorials. jQuery is used for DOM scripting and Ajax requests.

ChatController

The ChatController exposes operations to get and post chat message. Here is the method to get messages:

@RequestMapping(method=RequestMethod.GET)
@ResponseBody
public DeferredResult<List<String>> getMessages(@RequestParam int messageIndex) {

  final DeferredResult<List<String>> deferredResult = new DeferredResult<List<String>>(null, Collections.emptyList());
  this.chatRequests.put(deferredResult, messageIndex);

  deferredResult.onCompletion(new Runnable() {
    @Override
    public void run() {
      chatRequests.remove(deferredResult);
    }
  });

  List<String> messages = this.chatRepository.getMessages…

Last updated on November 5th, 2012 (Spring MVC 3.2 RC1)

In my last post I discussed how to make a Spring MVC controller method asynchronous by returning a Callable which is then invoked in a separate thread by Spring MVC.

But what if async processing depended on receiving some external event in a thread not known to Spring MVC -- e.g. receiving a JMS message, an AMQP message, a Redis pub-sub notification, a Spring Integration event, and so on? I'll explore this scenario by modifying an existing sample from the Spring AMQP project.

The Sample

Spring AMQP has a stock trading sample where a QuoteController sends trade execution messages via Spring AMQP's RabbitTemplate and receives trade confirmation and price quote messages via Spring AMQP's RabbitMQ listener container in message-driven POJO style.

In the browser, the sample uses polling to display price quotes. For trades, the initial request submits the trade and a confirmation id is returned that is then used to poll for the final confirmation. I've updated the sample to take advantage of the Spring 3.2 Servlet 3 async support. The master branch has the code before and the spring-mvc-async branch has the code after the change. The images below show the effect on the frequency of price quote requests (using the Chrome developer tools):

Before the change: traditional polling

After the change: long poll

As you can see with regular polling new requests are sent very frequently (milliseconds apart) while with long polling, requests can be 5, 10, 20, or more seconds apart -- a significant reduction in the total number of requests without the loss of latency, i.e. the amount of time before a new price quote appears in the browser.

Getting Quotes

So what changes were required? From a client perspective traditional polling and long polling are indistinguishable so the HTML and JavaScript did not change. From a server perspective requests must be held up until new quotes arrive. This is how the controller processes a request for quotes:

// Class field
private Map<String, DeferredResult> suspendedTradeRequests = new ConcurrentHashMap<String, DeferredResult>();

...

@RequestMapping("/quotes")
@ResponseBody
public DeferredResult<List<Quote>> quotes(@RequestParam(required = false) Long timestamp) {

  final DeferredResult<List<Quote>> result = new DeferredResult<List<Quote>>(null, Collections.emptyList());
  this.quoteRequests.put(result, timestamp);

  result.onCompletion(new Runnable() {
    public void run() {
      quoteRequests…

Last updated on November 5th, 2012 (Spring MVC 3.2 RC1)

In previous posts I introduced the Servlet 3 based async capability in Spring MVC 3.2 and discussed techniques for real-time updates. In this post I'll go into more technical details and discuss how asynchronous processing fits into the Spring MVC request lifecycle.

As a quick reminder, you can make any existing controller method asynchronous by changing it to return a Callable. For example a controller method that returns a view name, can return Callable<String> instead. An @ResponseBody that returns an object called Person can return Callable<Person> instead. And the same is true for any other controller return value type.

A central idea is that all of what you already know about how a controller method works remains unchanged as much as possible except that the remaining processing will occur in another thread. When it comes to asynchronous execution it's important to keep things simple. As you'll see even with this seemingly simple programming model change, there is quite a bit to consider.

The spring-mvc-showcase has been updated for Spring MVC 3.2. Have a look at CallableController. Method annotations like @ResponseBody and @ResponseStatus apply to the return value from the Callable as well, as you might expect. Exceptions raised from a Callable are handled as if they were raised by the controller, in this case with an @ExceptionHandler method. And so on.

If you execute one of the CallableController methods through the "Async Requests" tab in the browser, you should see output similar to the one below:

08:25:15 [http-bio-8080-exec-10] DispatcherServlet - DispatcherServlet with name 'appServlet' processing GET request for [...]
08:25:15 [http-bio-8080-exec-10] RequestMappingHandlerMapping - Looking up handler method for path /async/callable/view
08:25:15 [http-bio-8080-exec-10] RequestMappingHandlerMapping - Returning handler method [...]
08:25:15 [http-bio-8080-exec-10] WebAsyncManager - Concurrent handling starting for GET [...]
08:25:15 [http-bio-8080-exec-10] DispatcherServlet - Leaving response open for concurrent…

Last updated on November 5th, 2012 (Spring MVC 3.2 RC1)

In my last post I introduced the new Servlet 3 based, async support in Spring MVC 3.2 and talked about long-running requests. A second very important motivation for async processing is the need for browsers to receive real-time updates. Examples include chatting in a browser, stock quotes, status updates, live sports results, and others. To be sure not all examples are equally delay-sensitive but all of them share a similar need.

In standard HTTP request-response semantics a browser initiates a request and the server sends a response, which means the server can't send new information until it has a request from the browser. Several approaches have evolved including traditional polling, long polling, and HTTP streaming and most recently we have the WebSocket protocol.

Traditional Polling

The browser keeps sending requests to check for new information and the server responds immediately each time. This fits scenarios where polling can be done at reasonably sparse intervals. For example a mail client can check for new messages every 10 minutes. It's simple and it works. However, the approach becomes inefficient when new information must be shown as soon as possible in which case polling must be very frequent.

Long Polling

The browser keeps sending requests but the server doesn't respond until it has new information to send. From a client perspective this is identical to traditional polling. From a server perspective this is very similar to a long-running request and can be scaled using the technique discussed in Part 1.

How long can the response remain open? Browsers are set to time out after 5 minutes and network intermediaries such as proxies can time out even sooner. So even if no new information arrives, a long polling request should complete regularly to allow the browser to send a new request. This IETF document recommends using a timeout value between 30 and 120 seconds but the actual value to use will likely depend on how much control you have over network intermediaries that separate the browser from server.

Long polling can dramatically reduce the number of requests required to receive information updates with low latency, especially where new information becomes available at irregular intervals. However, the more frequent the updates are the closer it gets to traditional polling.

HTTP Streaming

The browser sends a request to the server and the server responds when it has information to send. However, unlike long polling, the server keeps the response open and continues to send more updates as they arrive. The approach removes the need for polling but is also a more significant departure from typical HTTP request-response semantics. For example the client and server need to agree how to interpret the response stream so that the client will know where one update ends and another begins. Furthermore, network intermediaries can cache the response stream which thwarts the intent of the approach. This is why long polling is more commonly used today.

WebSocket Protocol

The browser sends an HTTP request to the server to switch to the WebSocket protocol and the server responds by confirming the upgrade. Thereafter browser and server can send data frames in both directions over a TCP socket.

The WebSocket protocol was designed to replace the need for polling and is specifically suited for scenarios where messages need to be exchanged between browser and server at a high frequency. The initial handshake over HTTP ensures WebSocket requests can go through firewalls. However, there are also significant challenges since a majority of deployed browsers do not support WebSockets and there are further issues with getting through network intermediaries.

WebSockets revolves around the two way exchange of text or binary messages. It leads to a significantly different approach from a RESTful, HTTP-based architecture. In fact there is a need for some another protocol on top of WebSockets, e.g. XMPP, AMQP, STOMP, or other and which one(s) will become predominant remains to be seen.

The WebSocket protocol is already standardized by the IETF while the WebSocket API is in the final stages of being standardized by W3C. A number of Java implementations have become available including servlet containers like Jetty and Tomcat. The Servlet 3.1 spec will likely support the initial WebSocket upgrade request while a separate JSR-356 will define a Java-based WebSocket API.

Coming back to Spring MVC 3.2, the Servlet 3 async feature can be used for long-running requests and also for HTTP streaming, techniques Filip Hanik referred to as "the server version of client AJAX calls". As for WebSockets, there is no support yet in Spring 3.2 but it will most likely be included in Spring 3.3. You can watch SPR-9356 for progress updates.

The next post turns to sample code and explains in more detail the new Spring MVC 3.2 feature.

Last updated on November 5th, 2012 (Spring MVC 3.2 RC1)

Overview

Spring MVC 3.2 introduces Servlet 3 based asynchronous request processing. This is the first of several blog posts covering this new capability and providing context in which to understand how and why you would use it.

The main purpose of early releases is to seek feedback. We've received plenty of it both here and in JIRA since this was first posted after the 3.2 M1 release. Thanks to everyone who gave it a try and commented! There have been numerous changes and there is still time for more feedback!

At a Glance

From a programming model perspective the new capabilities appear deceptively simple. A controller method can now return a java.util.concurrent.Callable to complete processing asynchronously. Spring MVC will then invoke the Callable in a separate thread with the help of a TaskExecutor. Here is a code snippet before:

// Before
@RequestMapping(method=RequestMethod.POST)
public String processUpload(final MultipartFile file) {
    // ...
    return "someView";
}

// After
@RequestMapping(method=RequestMethod.POST)
public Callable<String> processUpload(final MultipartFile file) {

  return new Callable<String>() {
    public Object call() throws Exception {
      // ...
      return "someView";
    }
  };
}

A controller method can also return a DeferredResult (new type in Spring MVC 3.2) to complete processing in a thread not known to Spring MVC. For example reacting to a JMS or an AMQP message, a Redis notification, and so on. Here is another code snippet:

@RequestMapping("/quotes")
@ResponseBody
public DeferredResult<String> quotes() {
  DeferredResult<String> deferredResult…