|Previous Contents Index DocHome Next|
|iPlanet Web Server, Enterprise Edition NSAPI Programmer's Guide|
Appendix E HyperText Transfer Protocol
The HyperText Transfer Protocol (HTTP) is a protocol (a set of rules that describes how information is exchanged) that allows a client (such as a web browser) and a web server to communicate with each other.
HTTP is based on a request/response model. The browser opens a connection to the server and sends a request to the server.
The server processes the request and generates a response which it sends to the browser. The server then closes the connection.
This appendix provides a short introduction to a few HTTP basics. For more information on HTTP, see the IETF home page at:
This appendix has the following sections:
iPlanet Web Server 6.0 supports HTTP 1.1. Previous versions of the server supported HTTP 1.0. The server is conditionally compliant with the HTTP 1.1 proposed standard, as approved by the Internet Engineering Steering Group (IESG) and the Internet Engineering Task Force (IETF) HTTP working group.
For more information on the criteria for being conditionally compliant, see the Hypertext Transfer ProtocolHTTP/1.1 specification (RFC 2068) at:
A request from a browser to a server includes the following information:
Request Method, URI, and Protocol Version
A browser can request information using a number of methods. The commonly used methods include the following:
GETRequests the specified resource (such as a document or image)
The browser can send headers to the server. Most are optional. Some commonly used request headers are shown in Table E-1.
If the browser has made a POST or PUT request, it sends data after the blank line following the request headers. If the browser sends a GET or HEAD request, there is no data to send.
The server's response includes the following:
HTTP Protocol Version, Status Code, and Reason Phrase
The server sends back a status code, which is a three-digit numeric code. The five categories of status codes are:
100-199 a provisional response.Some common status codes are shown in Table E-2.
Use a local copy. If a browser already has a page in its cache, and the page is requested again, some browsers (such as Netscape Navigator) relay to the web server the "last-modified" timestamp on the browser's cached copy. If the copy on the server is not newer than the browser's copy, the server returns a 304 code instead of returning the page, reducing unnecessary network traffic. This is not an error.
If the client starts a request but does not complete it within the keep-alive timeout configured in the server, then this reponse will be sent and the connection closed. The request can be repeated with another open connection.
The client submitted a POST request with chunked-encoding, which is of variable length. However, the resource or application on the server requires a fixed length - a "content-length" header to be present. This code tells the client to resubmit its request with content-length.
The response headers contain information about the server and the response data. Common response headers are shown in Table E-3.
The server sends a blank line after the last header. It then sends the response data such as an image or an HTML page.
Buffered streams improve the efficiency of network I/O (for example the exchange of HTTP requests and responses) especially for dynamic content generation. Buffered streams are implemented as transparent NSPR I/O layers, which means even existing NSAPI modules can use them without any change.
The buffered streams layer adds following features to the iPlanet Web Server:
Enhanced keep-alive support: When the response is smaller than the buffer size, the buffering layer generates the content-length header so that client can detect the end of the response and re-use the connection for subsequent requests.The improved connection handling and response length header generation provided by buffered streams also addresses the HTTP 1.1 protocol compliance issues where absence of the response length headers is regarded as a category 1 failure. In previous Enterprise Server versions it was the responsibility of the dynamic content generation programs to send the length headers. If a CGI script did not generate the content-length header, the server had to close the connection to indicate the end of the response, breaking the keep-alive mechanism. However, it is often very inconvenient to keep track of response length in CGI scripts or servlets, and as an application platform provider, the web server is expected to handle such low-level protocol issues.
Response length determination: If the buffering layer cannot determine the length of the response, it uses HTTP 1.1 chunked encoding instead of the content-length header to convey the delineation information. If the client only understands HTTP 1.0, the server must close the connection to indicate the end of the response.
Output buffering has been built in to the functions that transmit data, such as net_write (see Chapter 5 "NSAPI Function Reference."). You can specify the following Service SAF parameters that affect stream buffering, which are described in detail in Chapter 3 "Predefined SAFs and the Request Handling Process."
Note The UseOutputStreamSize parameter can be set to zero in the obj.conf file to disable output stream buffering. For the magnus.conf file, setting UseOutputStreamSize to zero has no effect.
To override the default behavior when invoking an SAF that uses one of the functions net_read or netbuf_grab, you can specify the value of the parameter in obj.conf, for example:
Service fn="my-service-saf" type=perf UseOutputStreamSize=8192
Previous Contents Index DocHome Next
Copyright © 2001 Sun Microsystems, Inc. Some preexisting portions Copyright © 2001 Netscape Communications Corp. All rights reserved.
Last Updated May 15, 2001