WebSocket frame decoding in Python

Question

I've tried to implement WebSocket frame unmasking algorithm (based on this: How can I send and receive WebSocket messages on the server side?) on the server side. Here's what I've got:

def decode(data):
    frame = bytearray(data)

    length = frame[1] & 127

    indexFirstMask = 2
    if length == 126:
        indexFirstMask = 4
    elif length == 127:
        indexFirstMask = 10

    indexFirstDataByte = indexFirstMask + 4
    mask = frame[indexFirstMask:indexFirstDataByte]

    i = indexFirstDataByte
    j = 0
    decoded = []
    while i < len(frame):
        decoded.append(frame[i] ^ mask[j%4])
        i += 1
        j += 1

    print decoded

    return "".join(chr(byte) for byte in decoded)

However I get very strange results. On JavaScript side:

w = new WebSocket("ws://localhost:2000");
w.send("test");
w.send("test");
w.send("test");
w.send("test");

produces on the server side:

[42, 73, 45, 46, 1, 0]
[42, 1, 98, 0, 0]
[2, 97, 0, 0]
[2, 97, 0, 0]

More calls to w.send("test"); produces [2, 97, 0, 0]. Also first two arrays have length >4 (number of characters in word test). And none of these gets converted to word test. It seeems that I must be doing something wrong in my decoding code. What's causing that? Any help?

EDIT Have look at raw frames:

[193, 134, 48, 166, 232, 11, 26, 239, 197, 37, 49, 166]
[193, 133, 57, 161, 169, 218, 19, 160, 203, 218, 57]
[193, 132, 150, 97, 124, 54, 148, 0, 124, 54]
[193, 132, 163, 26, 102, 249, 161, 123, 102, 249]
[193, 132, 238, 212, 210, 156, 236, 181, 210, 156]

Of course these numbers are a bit random (due to masking), but note that second byte (which is supposed to represent the length of payload) is 134, then 133 and then always 132. Also first two frames are longer then other frames. What's going on here?

Please have a look at this answer it might help you http://stackoverflow.com/a/9778823/1193863 — naren, Oct 08 '13 at 16:05

score 2 · Accepted Answer · answered Apr 01 '13 at 09:32

I've just looked at this frames one more time and I've realized that the first byte:

193[DEC] == 11000001[BIN]

which according to the specification means that RSV1 is used (the second bit is 1). This seems to mean that the extension is used and indeed in my handshake code I've found the following line:

ext = req.headers.get("Sec-WebSocket-Extensions", "").strip()
if ext:
    res += "Sec-WebSocket-Extensions: %s\r\n" % ext

and in Chrome I can see:

Sec-WebSocket-Extensions:x-webkit-deflate-frame

when I do the handshake.

I'm not sure yet how these extensions work but removing these 3 lines from the handshake code fixes the issue.

WebSocket frame decoding in Python

1 Answers1