After a request has been accepted, it should also be 'loaded'. Before being loaded the request is in an undefined state and it is unsafe to access any data associated with the request - although you may still read/write with the request. For CGI, the request's constructor calls load implicitly (this is optional behaviour), in most other situations, one of the following functions are used:

What the call does is acquire the request environment data as necessary and parse the CGI GET and HTTP_COOKIE variables. Also reads and parses form data (ie. POST data). If the form data include file uploads, these are saved to disk and information about the file is stored in the request (see basic_request<>::uploads).

CGI example

#include <boost/cgi/cgi.hpp>

namespace cgi = boost::cgi;

int main()
{

  // Delay loading the request data
  cgi::request request(cgi::parse_none);
  cgi::response response;

  // ...

  // Load the request now (including parsing stdin).
  // Passing an `error_code` argument stops exceptions being thrown.
  boost::system::error_code& ec;
  request.load(cgi::parse_all, ec);

  if (ec)
  {
    response<< "Failed to load response.";
    return cgi::commit(request, response, cgi::http::internal_server_error);
  }

  response<< "Loaded response OK.";
  return cgi::commit(request, response);
}

Synchronous FastCGI example

#include <boost/cgi/fcgi.hpp>

namespace fcgi = boost::fcgi;

int handler(fcgi::request& request)
{
    // It is undefined behaviour to use a request without loading it.
    request.load(fcgi::parse_env); // minimal parsing.
    
    // Create a Response
    fcgi::response response;
    response<< "Hello, world.";
    return fcgi::commit(request, response);
}

int main()
{
  // Create a ProtocolService
  fcgi::service service;

  // Create a RequestAcceptor
  fcgi::acceptor acceptor(service);

  int status(0);
  while (!status) {
    status = acceptor.accept(&handler);
  }

  return status;
}