Simple example:

>>> from telnetlib import Telnet
>>> t = Telnet("google.com", 80)
>>> t.write("GET / HTTP/1.1\r\n")        
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/py3k/Lib/telnetlib.py", line 280, in write
    self.sock.sendall(buffer)
TypeError: sendall() argument 1 must be string or buffer, not str
>>> t.write(b"GET / HTTP/1.1\r\n")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/py3k/Lib/telnetlib.py", line 277, in write
    if IAC in buffer:
TypeError: Type str doesn't support the buffer API

I think only bytes need to be allowed for write() and read*() because of
low-level nature of Telnet. I can create a patch later.

I think that telnet should only use bytes (and not characters). For an
HTTP connection, the charset is only known after parsing the HTTP
headers. So telnet should use bytes, and your HTTP browser will
convert bytes to characters using the charset from the HTTP headers.
My patch only uses bytes for internal buffering and special codes
(IAC, DONT, ENCRYPT, etc.).

Example to test the library (Starwars, telnet, ISO-8859-1):
from telnetlib import Telnet
from sys import stdout
ipv4 = "towel.blinkenlights.nl"
ipv6 = "2001:980:ffe:1::42"
t = Telnet(ipv6, 23)
while True:
command = t.read_some()
command = str(command, "ISO-8859-1")
stdout.write(command)

Example to test the library (Google, HTTP, ASCII):
from telnetlib import Telnet
t = Telnet("www.google.com", 80)
t.write(b'GET / HTTP/1.0\r\n\r\n')
answer = t.read_all()
answer = str(answer, "ASCII")
print(answer)

The patch looks pretty straightforward to me. If somebody else gives
their nod, I'll apply it.

Yes, the patch is good.

I think that documentation (both Doc\library\telnetlib.rst and the
docstrings in telnetlib.py) should reflect the change, at least the code
samples.

The patch is good. It's exactly what I told about in msg72132.

Fixed in r66904.