Actualización de acuerdo con el cambio pregunta
Posiblemente, sigue girando debido a que en combinación de ausencia de Content-Length y Connection cabeceras, navegador puede asumir que es Connection: keep-alive, por lo que continúa recibiendo los datos de su servidor para siempre. Intente enviar Connection: close, y pase Content-Length real para ver si eso ayuda.
¿No hará esto lo que usted espera? :)
#!/usr/bin/env python
# coding: utf8
import socket
MAX_PACKET = 32768
def recv_all(sock):
r'''Receive everything from `sock`, until timeout occurs, meaning sender
is exhausted, return result as string.'''
# dirty hack to simplify this stuff - you should really use zero timeout,
# deal with async socket and implement finite automata to handle incoming data
prev_timeout = sock.gettimeout()
try:
sock.settimeout(0.01)
rdata = []
while True:
try:
rdata.append(sock.recv(MAX_PACKET))
except socket.timeout:
return ''.join(rdata)
# unreachable
finally:
sock.settimeout(prev_timeout)
def normalize_line_endings(s):
r'''Convert string containing various line endings like \n, \r or \r\n,
to uniform \n.'''
return ''.join((line + '\n') for line in s.splitlines())
def run():
r'''Main loop'''
# Create TCP socket listening on 10000 port for all connections,
# with connection queue of length 1
server_sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM, \
socket.IPPROTO_TCP)
server_sock.bind(('0.0.0.0', 13000))
server_sock.listen(1)
while True:
# accept connection
client_sock, client_addr = server_sock.accept()
# headers and body are divided with \n\n (or \r\n\r\n - that's why we
# normalize endings). In real application usage, you should handle
# all variations of line endings not to screw request body
request = normalize_line_endings(recv_all(client_sock)) # hack again
request_head, request_body = request.split('\n\n', 1)
# first line is request headline, and others are headers
request_head = request_head.splitlines()
request_headline = request_head[0]
# headers have their name up to first ': '. In real world uses, they
# could duplicate, and dict drops duplicates by default, so
# be aware of this.
request_headers = dict(x.split(': ', 1) for x in request_head[1:])
# headline has form of "POST /can/i/haz/requests HTTP/1.0"
request_method, request_uri, request_proto = request_headline.split(' ', 3)
response_body = [
'<html><body><h1>Hello, world!</h1>',
'<p>This page is in location %(request_uri)r, was requested ' % locals(),
'using %(request_method)r, and with %(request_proto)r.</p>' % locals(),
'<p>Request body is %(request_body)r</p>' % locals(),
'<p>Actual set of headers received:</p>',
'<ul>',
]
for request_header_name, request_header_value in request_headers.iteritems():
response_body.append('<li><b>%r</b> == %r</li>' % (request_header_name, \
request_header_value))
response_body.append('</ul></body></html>')
response_body_raw = ''.join(response_body)
# Clearly state that connection will be closed after this response,
# and specify length of response body
response_headers = {
'Content-Type': 'text/html; encoding=utf8',
'Content-Length': len(response_body_raw),
'Connection': 'close',
}
response_headers_raw = ''.join('%s: %s\n' % (k, v) for k, v in \
response_headers.iteritems())
# Reply as HTTP/1.1 server, saying "HTTP OK" (code 200).
response_proto = 'HTTP/1.1'
response_status = '200'
response_status_text = 'OK' # this can be random
# sending all this stuff
client_sock.send('%s %s %s' % (response_proto, response_status, \
response_status_text))
client_sock.send(response_headers_raw)
client_sock.send('\n') # to separate headers from body
client_sock.send(response_body_raw)
# and closing connection, as we stated before
client_sock.close()
run()
Para una descripción más detallada, véase description of HTTP protocol.
Sí, pero ... ¿el navegador simplemente "gira" y no se muestra nada? – antonpug
He actualizado el ejemplo de código para trabajar 100% garantizado :) Espero que encuentre útiles los principios básicos del manejo de encabezados, pero le recomiendo que no confíe en este tipo de código e implemente el analizador de solicitudes HTTP con todas las funciones. – toriningen
Bueno, has hecho una respuesta bastante completa ... aunque creo que la razón por la que todavía estaba girando (hasta el tiempo de espera) es que estaba esperando un doble "\ n". Pero al menos, su ejemplo de código es un buen fragmento para tener debajo del gancho, por las dudas;) – zmo