- The Advantages of Servlets Over "Traditional" CGI
- Basic Servlet Structure
- The Servlet Life Cycle
- The Client Request: Form Data
- The Client Request: HTTP Request Headers
- The Servlet Equivalent of the Standard CGI Variables
- The Server Response: HTTP Status Codes
- The Server Response: HTTP Response Headers
- Cookies
- Session Tracking
2.7 The Server Response: HTTP Status Codes
When a Web server responds to a request from a browser or other Web client, the response typically consists of a status line, some response headers, a blank line, and the document. Here is a minimal example:
HTTP/1.1 200 OK Content-Type: text/plain Hello World
The status line consists of the HTTP version (HTTP/1.1 in the example above), a status code (an integer; 200 in the example), and a very short message corresponding to the status code (OK in the example). In most cases, all of the headers are optional except for Content-Type, which specifies the MIME type of the document that follows. Although most responses contain a document, some don't. For example, responses to HEAD requests should never include a document, and a variety of status codes essentially indicate failure and either don't include a document or include only a short error-message document.
Servlets can perform a variety of important tasks by manipulating the status line and the response headers. For example, they can forward the user to other sites; indicate that the attached document is an image, Adobe Acrobat file, or HTML file; tell the user that a password is required to access the document; and so forth. This section briefly summarizes the most important status codes and what can be accomplished with them; see Chapter 6 of Core Servlets and JavaServer Pages (in PDF at http://www.moreservlets.com) for more details. The following section discusses the response headers.
Specifying Status Codes
As just described, the HTTP response status line consists of an HTTP version, a status code, and an associated message. Since the message is directly associated with the status code and the HTTP version is determined by the server, all a servlet needs to do is to set the status code. A code of 200 is set automatically, so servlets don't usually need to specify a status code at all. When they do set a code, they do so with the setStatus method of HttpServletResponse. If your response includes a special status code and a document, be sure to call setStatus before actually returning any of the content with the PrintWriter. That's because an HTTP response consists of the status line, one or more headers, a blank line, and the actual document, in that order. As discussed in Section 2.2 (Basic Servlet Structure), servlets do not necessarily buffer the document (version 2.1 servlets never do so), so you have to either set the status code before first using the PrintWriter or carefully check that the buffer hasn't been flushed and content actually sent to the browser.
Core Approach
Set status codes before sending any document content to the client.
The setStatus method takes an int (the status code) as an argument, but instead of using explicit numbers, for clarity and reliability use the constants defined in HttpServletResponse. The name of each constant is derived from the standard HTTP 1.1 message for each constant, all upper case with a prefix of SC (for Status Code) and spaces changed to underscores. Thus, since the message for 404 is Not Found, the equivalent constant in HttpServletResponse is SC_NOT_FOUND. There are two exceptions, however. The constant for code 302 is derived from the HTTP 1.0 message (Moved Temporarily), not the HTTP 1.1 message (Found), and the constant for code 307 (Temporary Redirect) is missing altogether.
Although the general method of setting status codes is simply to call response.setStatus(int), there are two common cases where a shortcut method in HttpServletResponse is provided. Just be aware that both of these methods throw IOException, whereas setStatus does not.
public void sendError(int code, String message) The sendError method sends a status code (usually 404) along with a short message that is automatically formatted inside an HTML document and sent to the client.
public void sendRedirect(String url) The sendRedirect method generates a 302 response along with a Location header giving the URL of the new document. With servlets version 2.1, this must be an absolute URL. In version 2.2 and 2.3, either an absolute or a relative URL is permitted; the system automatically translates relative URLs into absolute ones before putting them in the Location header.
Setting a status code does not necessarily mean that you don't need to return a document. For example, although most servers automatically generate a small File Not Found message for 404 responses, a servlet might want to customize this response. Again, remember that if you do send output, you have to call setStatus or sendError first.
HTTP 1.1 Status Codes
In this subsection I describe the most important status codes available for use in servlets talking to HTTP 1.1 clients, along with the standard message associated with each code. A good understanding of these codes can dramatically increase the capabilities of your servlets, so you should at least skim the descriptions to see what options are at your disposal. You can come back for details when you are ready to make use of some of the capabilities.
The complete HTTP 1.1 specification is given in RFC 2616. In general, you can access RFCs online by going to http://www.rfc-editor.org/ and following the links to the latest RFC archive sites, but since this one came from the World Wide Web Consortium, you can just go to http://www.w3.org/Protocols/. Codes that are new in HTTP 1.1 are noted, since some browsers support only HTTP 1.0. You should only send the new codes to clients that support HTTP 1.1, as verified by checking request.getRequestProtocol.
The rest of this section describes the specific status codes available in HTTP 1.1. These codes fall into five general categories:
100199 Codes in the 100s are informational, indicating that the client should respond with some other action.
200299 Values in the 200s signify that the request was successful.
300399 Values in the 300s are used for files that have moved and usually include a Location header indicating the new address.
400499 Values in the 400s indicate an error by the client.
500599 Codes in the 500s signify an error by the server.
The constants in HttpServletResponse that represent the various codes are derived from the standard messages associated with the codes. In servlets, you usually refer to status codes only by means of these constants. For example, you would use response.setStatus(response.SC_NO_CONTENT) rather than response.setStatus(204), since the latter is unclear to readers and is prone to typographical errors. However, you should note that servers are allowed to vary the messages slightly, and clients pay attention only to the numeric value. So, for example, you might see a server return a status line of HTTP/1.1 200 Document Fol_lows instead of HTTP/1.1 200 OK.
100 (Continue)
If the server receives an Expect request header with a value of 100-con_tinue, it means that the client is asking if it can send an attached document in a follow-up request. In such a case, the server should either respond with status 100 (SC_CONTINUE) to tell the client to go ahead or use 417 (Expectation Failed) to tell the browser it won't accept the document. This status code is new in HTTP 1.1.
200 (OK)
A value of 200 (SC_OK) means that everything is fine. The document follows for GET and POST requests. This status is the default for servlets; if you don't use setStatus, you'll get 200.
202 (Accepted)
A value of 202 (SC_ACCEPTED) tells the client that the request is being acted upon, but processing is not yet complete.
204 (No Content)
A status code of 204 (SC_NO_CONTENT) stipulates that the browser should continue to display the previous document because no new document is available. This behavior is useful if the user periodically reloads a page by pressing the Reload button, and you can determine that the previous page is already up-to-date.
205 (Reset Content)
A value of 205 (SC_RESET_CONTENT) means that there is no new document, but the browser should reset the document view. This status code instructs browsers to clear form fields. It is new in HTTP 1.1.
301 (Moved Permanently)
The 301 (SC_MOVED_PERMANENTLY) status indicates that the requested document is elsewhere; the new URL for the document is given in the Location response header. Browsers should automatically follow the link to the new URL.
302 (Found)
This value is similar to 301, except that in principle the URL given by the Location header should be interpreted as a temporary replacement, not a permanent one. In practice, most browsers treat 301 and 302 identically. Note: In HTTP 1.0, the message was Moved Temp_orarily instead of Found, and the constant in HttpServletRe_sponse is SC_MOVED_TEMPORARILY, not the expected SC_FOUND.
Core Note
The constant representing 302 is SC_MOVED_TEMPORARILY, not SC_FOUND.
Status code 302 is useful because browsers automatically follow the reference to the new URL given in the Location response header. It is so useful, in fact, that there is a special method for it, sendRedirect. Using response.sendRedirect(url) has a couple of advantages over using response.setStatus(response.SC_MOVED_TEMPORARILY) and response.setHeader("Location", url). First, it is shorter and easier. Second, with sendRedirect, the servlet automatically builds a page containing the link to show to older browsers that don't automatically follow redirects. Finally, with version 2.2 and 2.3 of servlets, sendRedirect can handle relative URLs, automatically translating them into absolute ones.
Technically, browsers are only supposed to automatically follow the redirection if the original request was GET. For details, see the discussion of the 307 status code.
303 (See Other)
The 303 (SC_SEE_OTHER) status is similar to 301 and 302, except that if the original request was POST, the new document (given in the Location header) should be retrieved with GET. This code is new in HTTP 1.1.
304 (Not Modified)
When a client has a cached document, it can perform a conditional request by supplying an If-Modified-Since header to indicate that it wants the document only if it has been changed since the specified date. A value of 304 (SC_NOT_MODIFIED) means that the cached version is up-to-date and the client should use it. Otherwise, the server should return the requested document with the normal (200) status code. Servlets normally should not set this status code directly. Instead, they should implement the getLastModified method and let the default service method handle conditional requests based upon this modification date. For an example, see Section 2.8 of Core Servlets and JavaServer Pages.
307 (Temporary Redirect)
The rules for how a browser should handle a 307 status are identical to those for 302. The 307 value was added to HTTP 1.1 since many browsers erroneously follow the redirection on a 302 response even if the original message is a POST. Browsers are supposed to follow the redirection of a POST request only when they receive a 303 response status. This new status is intended to be unambiguously clear: follow redirected GET and POST requests in the case of 303 responses; follow redirected GET but not POST requests in the case of 307 responses. Note: For some reason there is no constant in HttpServletResponse corresponding to this status code, so you have to use 307 explicitly. This status code is new in HTTP 1.1.
400 (Bad Request)
A 400 (SC_BAD_REQUEST) status indicates bad syntax in the client request.
401 (Unauthorized)
A value of 401 (SC_UNAUTHORIZED) signifies that the client tried to access a password-protected page without proper identifying information in the Authorization header. The response must include a WWW-Authenticate header.
403 (Forbidden)
A status code of 403 (SC_FORBIDDEN) means that the server refuses to supply the resource, regardless of authorization. This status is often the result of bad file or directory permissions on the server.
404 (Not Found)
The infamous 404 (SC_NOT_FOUND) status tells the client that no resource could be found at that address. This value is the standard "no such page" response. It is such a common and useful response that there is a special method for it in the HttpServletResponse class: sendError("message"). The advantage of sendError over set_Status is that, with sendError, the server automatically generates an error page showing the error message. 404 errors need not merely say "Sorry, the page cannot be found." Instead, they can give information on why the page couldn't be found or supply search boxes or alternative places to look. The sites at http://www.microsoft.com and http://www.ibm.com have particularly good examples of useful error pages. In fact, there is an entire site dedicated to the good, the bad, the ugly, and the bizarre in 404 error messages: http://www.plinko.net/404/. I find http://www.plinko.net/404/category.asp?Category=Funny particularly amusing.
Unfortunately, however, the default behavior of Internet Explorer 5 is to ignore the error page you send back and to display its own, even though doing so contradicts the HTTP specification. To turn off this setting, you can go to the Tools menu, select Internet Options, choose the Advanced tab, and make sure "Show friendly HTTP error messages" box is not checked. Unfortunately, however, few users are aware of this setting, so this "feature" prevents most users of Internet Explorer version 5 from seeing any informative messages you return. Other major browsers and version 4 of Internet Explorer properly display server-generated error pages.
Core Warning
By default, Internet Explorer version 5 improperly ignores server-generated error pages.
To make matters worse, some versions of Tomcat 3 fail to properly handle strings that are passed to sendError. So, if you are using Tomcat 3, you may need to generate 404 error messages by hand. Fortunately, it is relatively uncommon for individual servlets to build their own 404 error pages. A more common approach is to set up error pages for each Web application; see Section 5.8 (Designating Pages to Handle Errors) for details. Tomcat correctly handles these pages.
Core Warning
Some versions of Tomcat 3.x fail to properly display strings that are supplied to sendError.
405 (Method Not Allowed)
A 405 (SC_METHOD_NOT_ALLOWED) value indicates that the request method (GET, POST, HEAD, PUT, DELETE, etc.) was not allowed for this particular resource. This status code is new in HTTP 1.1.
415 (Unsupported Media Type)
A value of 415 (SC_UNSUPPORTED_MEDIA_TYPE) means that the request had an attached document of a type the server doesn't know how to handle. This status code is new in HTTP 1.1.
417 (Expectation Failed)
If the server receives an Expect request header with a value of 100-con_tinue, it means that the client is asking if it can send an attached document in a follow-up request. In such a case, the server should either respond with this status (417) to tell the browser it won't accept the document or use 100 (SC_CONTINUE) to tell the client to go ahead. This status code is new in HTTP 1.1.
500 (Internal Server Error)
500 (SC_INTERNAL_SERVER_ERROR) is the generic "server is confused" status code. It often results from CGI programs or (heaven forbid!) servlets that crash or return improperly formatted headers.
501 (Not Implemented)
The 501 (SC_NOT_IMPLEMENTED) status notifies the client that the server doesn't support the functionality to fulfill the request. It is used, for example, when the client issues a command like PUT that the server doesn't support.
503 (Service Unavailable)
A status code of 503 (SC_SERVICE_UNAVAILABLE) signifies that the server cannot respond because of maintenance or overloading. For example, a servlet might return this header if some thread or database connection pool is currently full. The server can supply a Retry-After header to tell the client when to try again.
505 (HTTP Version Not Supported)
The 505 (SC_HTTP_VERSION_NOT_SUPPORTED) code means that the server doesn't support the version of HTTP named in the request line. This status code is new in HTTP 1.1.
A Front End to Various Search Engines
Listing 2.12 presents an example that makes use of the two most common status codes other than 200 (OK): 302 (Found) and 404 (Not Found). The 302 code is set by the shorthand sendRedirect method of HttpServletResponse, and 404 is specified by sendError.
In this application, an HTML form (see Figure 210 and the source code in Listing 2.14) first displays a page that lets the user specify a search string, the number of results to show per page, and the search engine to use. When the form is submitted, the servlet extracts those three parameters, constructs a URL with the parameters embedded in a way appropriate to the search engine selected (see the SearchSpec class of Listing 2.13), and redirects the user to that URL (see Figure 211). If the user fails to choose a search engine or specify search terms, an error page informs the client of this fact (but see warnings under the 404 status code in the previous subsection).
Listing 2.12 SearchEngines.java
package moreservlets; import java.io.*; import javax.servlet.*; import javax.servlet.http.*; import java.net.*; /** Servlet that takes a search string, number of results per * page, and a search engine name, sending the query to * that search engine. Illustrates manipulating * the response status line. It sends a 302 response * (via sendRedirect) if it gets a known search engine, * and sends a 404 response (via sendError) otherwise. */ public class SearchEngines extends HttpServlet { public void doGet(HttpServletRequest request, HttpServletResponse response) throws ServletException, IOException { String searchString = request.getParameter("searchString"); if ((searchString == null) || (searchString.length() == 0)) { reportProblem(response, "Missing search string."); return; } // The URLEncoder changes spaces to "+" signs and other // non-alphanumeric characters to "%XY", where XY is the // hex value of the ASCII (or ISO Latin-1) character. // Browsers always URL-encode form values, so the // getParameter method decodes automatically. But since // we're just passing this on to another server, we need to // re-encode it. searchString = URLEncoder.encode(searchString); String numResults = request.getParameter("numResults"); if ((numResults == null) || (numResults.equals("0")) || (numResults.length() == 0)) { numResults = "10"; } String searchEngine = request.getParameter("searchEngine"); if (searchEngine == null) { reportProblem(response, "Missing search engine name."); return; } SearchSpec[] commonSpecs = SearchSpec.getCommonSpecs(); for(int i=0; i<commonSpecs.length; i++) { SearchSpec searchSpec = commonSpecs[i]; if (searchSpec.getName().equals(searchEngine)) { String url = searchSpec.makeURL(searchString, numResults); response.sendRedirect(url); return; } } reportProblem(response, "Unrecognized search engine."); } private void reportProblem(HttpServletResponse response, String message) throws IOException { response.sendError(response.SC_NOT_FOUND, "<H2>" + message + "</H2>"); } public void doPost(HttpServletRequest request, HttpServletResponse response) throws ServletException, IOException { doGet(request, response); } }
Listing 2.13 SearchSpec.java
package moreservlets; /** Small class that encapsulates how to construct a * search string for a particular search engine. */ public class SearchSpec { private String name, baseURL, numResultsSuffix; private static SearchSpec[] commonSpecs = { new SearchSpec ("google", "http://www.google.com/search?q=", "&num="), new SearchSpec ("altavista", "http://www.altavista.com/sites/search/web?q=", "&nbq="), new SearchSpec ("lycos", "http://lycospro.lycos.com/cgi-bin/" + "pursuit?query=", "&maxhits="), new SearchSpec ("hotbot", "http://www.hotbot.com/?MT=", "&DC=") }; public SearchSpec(String name, String baseURL, String numResultsSuffix) { this.name = name; this.baseURL = baseURL; this.numResultsSuffix = numResultsSuffix; } public String makeURL(String searchString, String numResults) { return(baseURL + searchString + numResultsSuffix + numResults); } public String getName() { return(name); } public static SearchSpec[] getCommonSpecs() { return(commonSpecs); } }
Figure 210 Front end to the SearchEngines servlet. See Listing 2.14 for the HTML source code.
Figure 211 Result of the SearchEngines servlet when the form of Figure 210 is submitted.
Figure 212 Result of the SearchEngines servlet when a form that has no search string is submitted. This result is for JRun 3.1; results can vary slightly among servers and will omit the "Missing search string" message in most Tomcat versions.
Listing 2.14 SearchEngines.html
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN"> <HTML> <HEAD> <TITLE>Searching the Web</TITLE> </HEAD> <BODY BGCOLOR="#FDF5E6"> <H1 ALIGN="CENTER">Searching the Web</H1> <FORM ACTION="/servlet/moreservlets.SearchEngines"> <CENTER> Search String: <INPUT TYPE="TEXT" NAME="searchString"><BR> Results to Show Per Page: <INPUT TYPE="TEXT" NAME="numResults" VALUE=10 SIZE=3><BR> <INPUT TYPE="RADIO" NAME="searchEngine" VALUE="google"> Google | <INPUT TYPE="RADIO" NAME="searchEngine" VALUE="altavista"> AltaVista | <INPUT TYPE="RADIO" NAME="searchEngine" VALUE="lycos"> Lycos | <INPUT TYPE="RADIO" NAME="searchEngine" VALUE="hotbot"> HotBot <BR> <INPUT TYPE="SUBMIT" VALUE="Search"> </CENTER> </FORM> </BODY> </HTML>