3
****************************
4
Socket Programming HOWTO
5
****************************
7
:Author: Gordon McMillan
12
Sockets are used nearly everywhere, but are one of the most severely
13
misunderstood technologies around. This is a 10,000 foot overview of sockets.
14
It's not really a tutorial - you'll still have work to do in getting things
15
operational. It doesn't cover the fine points (and there are a lot of them), but
16
I hope it will give you enough background to begin using them decently.
22
I'm only going to talk about INET (i.e. IPv4) sockets, but they account for at least 99% of
23
the sockets in use. And I'll only talk about STREAM (i.e. TCP) sockets - unless you really
24
know what you're doing (in which case this HOWTO isn't for you!), you'll get
25
better behavior and performance from a STREAM socket than anything else. I will
26
try to clear up the mystery of what a socket is, as well as some hints on how to
27
work with blocking and non-blocking sockets. But I'll start by talking about
28
blocking sockets. You'll need to know how they work before dealing with
31
Part of the trouble with understanding these things is that "socket" can mean a
32
number of subtly different things, depending on context. So first, let's make a
33
distinction between a "client" socket - an endpoint of a conversation, and a
34
"server" socket, which is more like a switchboard operator. The client
35
application (your browser, for example) uses "client" sockets exclusively; the
36
web server it's talking to uses both "server" sockets and "client" sockets.
42
Of the various forms of :abbr:`IPC (Inter Process Communication)`,
43
sockets are by far the most popular. On any given platform, there are
44
likely to be other forms of IPC that are faster, but for
45
cross-platform communication, sockets are about the only game in town.
47
They were invented in Berkeley as part of the BSD flavor of Unix. They spread
48
like wildfire with the Internet. With good reason --- the combination of sockets
49
with INET makes talking to arbitrary machines around the world unbelievably easy
50
(at least compared to other schemes).
56
Roughly speaking, when you clicked on the link that brought you to this page,
57
your browser did something like the following::
59
# create an INET, STREAMing socket
60
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
61
# now connect to the web server on port 80 - the normal http port
62
s.connect(("www.python.org", 80))
64
When the ``connect`` completes, the socket ``s`` can be used to send
65
in a request for the text of the page. The same socket will read the
66
reply, and then be destroyed. That's right, destroyed. Client sockets
67
are normally only used for one exchange (or a small set of sequential
70
What happens in the web server is a bit more complex. First, the web server
71
creates a "server socket"::
73
# create an INET, STREAMing socket
74
serversocket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
75
# bind the socket to a public host, and a well-known port
76
serversocket.bind((socket.gethostname(), 80))
77
# become a server socket
78
serversocket.listen(5)
80
A couple things to notice: we used ``socket.gethostname()`` so that the socket
81
would be visible to the outside world. If we had used ``s.bind(('localhost',
82
80))`` or ``s.bind(('127.0.0.1', 80))`` we would still have a "server" socket,
83
but one that was only visible within the same machine. ``s.bind(('', 80))``
84
specifies that the socket is reachable by any address the machine happens to
87
A second thing to note: low number ports are usually reserved for "well known"
88
services (HTTP, SNMP etc). If you're playing around, use a nice high number (4
91
Finally, the argument to ``listen`` tells the socket library that we want it to
92
queue up as many as 5 connect requests (the normal max) before refusing outside
93
connections. If the rest of the code is written properly, that should be plenty.
95
Now that we have a "server" socket, listening on port 80, we can enter the
96
mainloop of the web server::
99
# accept connections from outside
100
(clientsocket, address) = serversocket.accept()
101
# now do something with the clientsocket
102
# in this case, we'll pretend this is a threaded server
103
ct = client_thread(clientsocket)
106
There's actually 3 general ways in which this loop could work - dispatching a
107
thread to handle ``clientsocket``, create a new process to handle
108
``clientsocket``, or restructure this app to use non-blocking sockets, and
109
mulitplex between our "server" socket and any active ``clientsocket``\ s using
110
``select``. More about that later. The important thing to understand now is
111
this: this is *all* a "server" socket does. It doesn't send any data. It doesn't
112
receive any data. It just produces "client" sockets. Each ``clientsocket`` is
113
created in response to some *other* "client" socket doing a ``connect()`` to the
114
host and port we're bound to. As soon as we've created that ``clientsocket``, we
115
go back to listening for more connections. The two "clients" are free to chat it
116
up - they are using some dynamically allocated port which will be recycled when
117
the conversation ends.
123
If you need fast IPC between two processes on one machine, you should look into
124
pipes or shared memory. If you do decide to use AF_INET sockets, bind the
125
"server" socket to ``'localhost'``. On most platforms, this will take a
126
shortcut around a couple of layers of network code and be quite a bit faster.
129
The :mod:`multiprocessing` integrates cross-platform IPC into a higher-level
136
The first thing to note, is that the web browser's "client" socket and the web
137
server's "client" socket are identical beasts. That is, this is a "peer to peer"
138
conversation. Or to put it another way, *as the designer, you will have to
139
decide what the rules of etiquette are for a conversation*. Normally, the
140
``connect``\ ing socket starts the conversation, by sending in a request, or
141
perhaps a signon. But that's a design decision - it's not a rule of sockets.
143
Now there are two sets of verbs to use for communication. You can use ``send``
144
and ``recv``, or you can transform your client socket into a file-like beast and
145
use ``read`` and ``write``. The latter is the way Java presents its sockets.
146
I'm not going to talk about it here, except to warn you that you need to use
147
``flush`` on sockets. These are buffered "files", and a common mistake is to
148
``write`` something, and then ``read`` for a reply. Without a ``flush`` in
149
there, you may wait forever for the reply, because the request may still be in
152
Now we come to the major stumbling block of sockets - ``send`` and ``recv`` operate
153
on the network buffers. They do not necessarily handle all the bytes you hand
154
them (or expect from them), because their major focus is handling the network
155
buffers. In general, they return when the associated network buffers have been
156
filled (``send``) or emptied (``recv``). They then tell you how many bytes they
157
handled. It is *your* responsibility to call them again until your message has
158
been completely dealt with.
160
When a ``recv`` returns 0 bytes, it means the other side has closed (or is in
161
the process of closing) the connection. You will not receive any more data on
162
this connection. Ever. You may be able to send data successfully; I'll talk
163
more about this later.
165
A protocol like HTTP uses a socket for only one transfer. The client sends a
166
request, then reads a reply. That's it. The socket is discarded. This means that
167
a client can detect the end of the reply by receiving 0 bytes.
169
But if you plan to reuse your socket for further transfers, you need to realize
170
that *there is no* :abbr:`EOT (End of Transfer)` *on a socket.* I repeat: if a socket
171
``send`` or ``recv`` returns after handling 0 bytes, the connection has been
172
broken. If the connection has *not* been broken, you may wait on a ``recv``
173
forever, because the socket will *not* tell you that there's nothing more to
174
read (for now). Now if you think about that a bit, you'll come to realize a
175
fundamental truth of sockets: *messages must either be fixed length* (yuck), *or
176
be delimited* (shrug), *or indicate how long they are* (much better), *or end by
177
shutting down the connection*. The choice is entirely yours, (but some ways are
178
righter than others).
180
Assuming you don't want to end the connection, the simplest solution is a fixed
184
"""demonstration class only
185
- coded for clarity, not efficiency
188
def __init__(self, sock=None):
190
self.sock = socket.socket(
191
socket.AF_INET, socket.SOCK_STREAM)
195
def connect(self, host, port):
196
self.sock.connect((host, port))
198
def mysend(self, msg):
200
while totalsent < MSGLEN:
201
sent = self.sock.send(msg[totalsent:])
203
raise RuntimeError("socket connection broken")
204
totalsent = totalsent + sent
209
while bytes_recd < MSGLEN:
210
chunk = self.sock.recv(min(MSGLEN - bytes_recd, 2048))
212
raise RuntimeError("socket connection broken")
214
bytes_recd = bytes_recd + len(chunk)
215
return b''.join(chunks)
217
The sending code here is usable for almost any messaging scheme - in Python you
218
send strings, and you can use ``len()`` to determine its length (even if it has
219
embedded ``\0`` characters). It's mostly the receiving code that gets more
220
complex. (And in C, it's not much worse, except you can't use ``strlen`` if the
221
message has embedded ``\0``\ s.)
223
The easiest enhancement is to make the first character of the message an
224
indicator of message type, and have the type determine the length. Now you have
225
two ``recv``\ s - the first to get (at least) that first character so you can
226
look up the length, and the second in a loop to get the rest. If you decide to
227
go the delimited route, you'll be receiving in some arbitrary chunk size, (4096
228
or 8192 is frequently a good match for network buffer sizes), and scanning what
229
you've received for a delimiter.
231
One complication to be aware of: if your conversational protocol allows multiple
232
messages to be sent back to back (without some kind of reply), and you pass
233
``recv`` an arbitrary chunk size, you may end up reading the start of a
234
following message. You'll need to put that aside and hold onto it, until it's
237
Prefixing the message with its length (say, as 5 numeric characters) gets more
238
complex, because (believe it or not), you may not get all 5 characters in one
239
``recv``. In playing around, you'll get away with it; but in high network loads,
240
your code will very quickly break unless you use two ``recv`` loops - the first
241
to determine the length, the second to get the data part of the message. Nasty.
242
This is also when you'll discover that ``send`` does not always manage to get
243
rid of everything in one pass. And despite having read this, you will eventually
246
In the interests of space, building your character, (and preserving my
247
competitive position), these enhancements are left as an exercise for the
248
reader. Lets move on to cleaning up.
254
It is perfectly possible to send binary data over a socket. The major problem is
255
that not all machines use the same formats for binary data. For example, a
256
Motorola chip will represent a 16 bit integer with the value 1 as the two hex
257
bytes 00 01. Intel and DEC, however, are byte-reversed - that same 1 is 01 00.
258
Socket libraries have calls for converting 16 and 32 bit integers - ``ntohl,
259
htonl, ntohs, htons`` where "n" means *network* and "h" means *host*, "s" means
260
*short* and "l" means *long*. Where network order is host order, these do
261
nothing, but where the machine is byte-reversed, these swap the bytes around
264
In these days of 32 bit machines, the ascii representation of binary data is
265
frequently smaller than the binary representation. That's because a surprising
266
amount of the time, all those longs have the value 0, or maybe 1. The string "0"
267
would be two bytes, while binary is four. Of course, this doesn't fit well with
268
fixed-length messages. Decisions, decisions.
274
Strictly speaking, you're supposed to use ``shutdown`` on a socket before you
275
``close`` it. The ``shutdown`` is an advisory to the socket at the other end.
276
Depending on the argument you pass it, it can mean "I'm not going to send
277
anymore, but I'll still listen", or "I'm not listening, good riddance!". Most
278
socket libraries, however, are so used to programmers neglecting to use this
279
piece of etiquette that normally a ``close`` is the same as ``shutdown();
280
close()``. So in most situations, an explicit ``shutdown`` is not needed.
282
One way to use ``shutdown`` effectively is in an HTTP-like exchange. The client
283
sends a request and then does a ``shutdown(1)``. This tells the server "This
284
client is done sending, but can still receive." The server can detect "EOF" by
285
a receive of 0 bytes. It can assume it has the complete request. The server
286
sends a reply. If the ``send`` completes successfully then, indeed, the client
289
Python takes the automatic shutdown a step further, and says that when a socket
290
is garbage collected, it will automatically do a ``close`` if it's needed. But
291
relying on this is a very bad habit. If your socket just disappears without
292
doing a ``close``, the socket at the other end may hang indefinitely, thinking
293
you're just being slow. *Please* ``close`` your sockets when you're done.
299
Probably the worst thing about using blocking sockets is what happens when the
300
other side comes down hard (without doing a ``close``). Your socket is likely to
301
hang. TCP is a reliable protocol, and it will wait a long, long time
302
before giving up on a connection. If you're using threads, the entire thread is
303
essentially dead. There's not much you can do about it. As long as you aren't
304
doing something dumb, like holding a lock while doing a blocking read, the
305
thread isn't really consuming much in the way of resources. Do *not* try to kill
306
the thread - part of the reason that threads are more efficient than processes
307
is that they avoid the overhead associated with the automatic recycling of
308
resources. In other words, if you do manage to kill the thread, your whole
309
process is likely to be screwed up.
315
If you've understood the preceding, you already know most of what you need to
316
know about the mechanics of using sockets. You'll still use the same calls, in
317
much the same ways. It's just that, if you do it right, your app will be almost
320
In Python, you use ``socket.setblocking(0)`` to make it non-blocking. In C, it's
321
more complex, (for one thing, you'll need to choose between the BSD flavor
322
``O_NONBLOCK`` and the almost indistinguishable Posix flavor ``O_NDELAY``, which
323
is completely different from ``TCP_NODELAY``), but it's the exact same idea. You
324
do this after creating the socket, but before using it. (Actually, if you're
325
nuts, you can switch back and forth.)
327
The major mechanical difference is that ``send``, ``recv``, ``connect`` and
328
``accept`` can return without having done anything. You have (of course) a
329
number of choices. You can check return code and error codes and generally drive
330
yourself crazy. If you don't believe me, try it sometime. Your app will grow
331
large, buggy and suck CPU. So let's skip the brain-dead solutions and do it
336
In C, coding ``select`` is fairly complex. In Python, it's a piece of cake, but
337
it's close enough to the C version that if you understand ``select`` in Python,
338
you'll have little trouble with it in C::
340
ready_to_read, ready_to_write, in_error = \
347
You pass ``select`` three lists: the first contains all sockets that you might
348
want to try reading; the second all the sockets you might want to try writing
349
to, and the last (normally left empty) those that you want to check for errors.
350
You should note that a socket can go into more than one list. The ``select``
351
call is blocking, but you can give it a timeout. This is generally a sensible
352
thing to do - give it a nice long timeout (say a minute) unless you have good
353
reason to do otherwise.
355
In return, you will get three lists. They contain the sockets that are actually
356
readable, writable and in error. Each of these lists is a subset (possibly
357
empty) of the corresponding list you passed in.
359
If a socket is in the output readable list, you can be
360
as-close-to-certain-as-we-ever-get-in-this-business that a ``recv`` on that
361
socket will return *something*. Same idea for the writable list. You'll be able
362
to send *something*. Maybe not all you want to, but *something* is better than
363
nothing. (Actually, any reasonably healthy socket will return as writable - it
364
just means outbound network buffer space is available.)
366
If you have a "server" socket, put it in the potential_readers list. If it comes
367
out in the readable list, your ``accept`` will (almost certainly) work. If you
368
have created a new socket to ``connect`` to someone else, put it in the
369
potential_writers list. If it shows up in the writable list, you have a decent
370
chance that it has connected.
372
Actually, ``select`` can be handy even with blocking sockets. It's one way of
373
determining whether you will block - the socket returns as readable when there's
374
something in the buffers. However, this still doesn't help with the problem of
375
determining whether the other end is done, or just busy with something else.
377
**Portability alert**: On Unix, ``select`` works both with the sockets and
378
files. Don't try this on Windows. On Windows, ``select`` works with sockets
379
only. Also note that in C, many of the more advanced socket options are done
380
differently on Windows. In fact, on Windows I usually use threads (which work
381
very, very well) with my sockets.