Re: about TURN MAGIC-COOKIE

Assuming the transport protocol used on the data path created using TURN 
is RTP (or other protocols those are error-resilient) it would most 
likely to be harmless, considering the probability is rather small, but 
what if the transport protocol is not error-resilient, such as TFTP?
Please advise, thank you.

Toshihiro Sonoda wrote:

> Thank you for your comment.
> I agree on your opinion.
>
>> I do not think it is necessary to inspect data packets and, if it is
>> found that they contain the magic cookie string, wrap them in a DATA
>> attribute in a SEND.
>>
>> The magic cookie is a random 32 bit string. The probability of it
>> appearing in a data packet is approximately 1 in 4 billion. If that
>> probability is still too large, then we should just increase the size of
>> the magic cookie. However, 1 in 2 billion is likely much smaller than
>> the packet loss rate of the underlying network. A mis-clasification in
>> TURN would manifest itself as a packet loss. Since this packet loss rate
>> is less than the network packet loss rate, I don't see a real issue.
>>
>> Thanks,
>> Jonathan R.
>>
>> Kenneth Ho wrote:
>>
>>> Toshihiro Sonoda wrote:
>>>
>>>
>>>>> Toshihiro Sonoda wrote:
>>>>>
>>>>>
>>>>>> Why can a server disambiguate a TURN Request from a data packet by
>>>>>> looking for the MAGIC-COOKIE attribute, as described in Section 
>>>>>> 7.2.3.
>>>>>>
>>>>>> Can it guarantee that UDP packets of non-TURN Packets doesn't have
>>>>>> same value as MAGIC-COOKIE.
>>>>>>
>>>>>>
>>>>>
>>>>> My take is a TURN client library would need to inspect every packet
>>>>> before sending them to a TURN server.
>>>>> If a packet happens to have something resembling a magic-cookie, the
>>>>> library would have to wrap it in a SEND request.
>>>>
>>>>
>>>>
>>>> If a packet from client to turn server happens to have something
>>>> resembling a magic-cookie, then the packet would have to be wrapped
>>>> using SEND request.
>>>> I understand that.
>>>>
>>>> but, if a packet form turn server to client happens to have something
>>>> resembling a
>>>> magic-cookie, then should turn server wrap the packet using any
>>>> turn-request or response?
>>>
>>>
>>>
>>> You could wrap that packet in a DATA indication to client.
>>>
>>>
>>>>
>>>> _______________________________________________
>>>> midcom mailing list
>>>> [email protected]
>>>> https://www1.ietf.org/mailman/listinfo/midcom
>>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> midcom mailing list
>>> [email protected]
>>> https://www1.ietf.org/mailman/listinfo/midcom
>>>
>>
>> -- 
>> Jonathan D. Rosenberg, Ph.D.                   600 Lanidex Plaza
>> Director, Service Provider VoIP Architecture   Parsippany, NJ 07054-2711
>> Cisco Systems
>> [email protected]                              FAX:   (973) 952-5050
>> http://www.jdrosen.net                         PHONE: (973) 952-5000
>> http://www.cisco.com
>>
>
> _______________________________________________
> midcom mailing list
> [email protected]
> https://www1.ietf.org/mailman/listinfo/midcom
>