Re: [Ice] 5245bis: STUN/TURN transaction timeout timer

[Ari Keränen <[e-mail]>]

Hi,

I had a look at the Ti timer in 5389 (https://tools.ietf.org/search/rfc5389#section-7.2.2) and it seems that it is defined to be in line with the time it takes for (UDP) STUN transaction that includes all re-transmissions to timeout:
>    Reliability of STUN over TCP and TLS-over-TCP is handled by TCP
>    itself, and there are no retransmissions at the STUN protocol level.
>    However, for a request/response transaction, if the client has not
>    received a response by Ti seconds after it sent the SYN to establish
>    the connection, it considers the transaction to have timed out.


However the HTO timer is for transaction pacing and is roughly what one successful connectivity check should take (2*RTT). 

Therefore these are two different things and I don't think it make sense to mention Ti here at all. I suggest we remove the note completely.


Cheers,
Ari (hat off)

> On 27 Feb 2018, at 2.25, Nils Ohlmeier <[e-mail]> wrote:
> 
> Yes you are right. Somehow I managed to read Christer’s Note the opposite way of what I now think he intends to point out.
> 
> Sorry for the confusion.
> 
>   Nils
> 
> 
>> On Feb 26, 2018, at 15:15, Black, David <[e-mail]> wrote:
>> 
>> Isn’t it the other way around – ICE HTO is much shorter than STUN or TURN Ti?
>>  
>> Thanks, --David
>>  
>> From: Nils Ohlmeier [mailto:[e-mail]] 
>> Sent: Monday, February 26, 2018 6:11 PM
>> To: Christer Holmberg <[e-mail]>
>> Cc: Black, David <[e-mail]>; [e-mail]; [e-mail]; [e-mail]; [e-mail]
>> Subject: Re: [Ice] 5245bis: STUN/TURN transaction timeout timer
>>  
>>  
>> On Feb 26, 2018, at 07:48, Christer Holmberg <[e-mail]> wrote:
>>  
>> Maybe adding the following note to the existing Timer HTO definition:
>>  
>> Timer HTO:  The timeout timer for a given STUN or TURN transaction.
>>  
>>   “NOTE: When STUN and TURN are used with ICE, timer HTO is used instead of timer Ti [RFC5389] as transaction timeout timer.”
>>  
>> My initial thought was: yes sounds good.
>>  
>> But one of the side effects know from real world deployments is that results in the end-of-candidates indication coming in after a long time if one of the STUN or TURN servers is not reachable.
>> I don’t want to make this a last minute change, but your indication that Ti explicitly got made shorter made me wonder if everyone in WG is aware of this usage of the long HTO value.
>>  
>> Best
>>   Nils Ohlmeier
>> 
>> 
>>  
>> Regards,
>>  
>> Christer
>>  
>> From: Black, David [mailto:[e-mail]] 
>> Sent: 26 February 2018 16:54
>> To: Christer Holmberg <[e-mail]>;[e-mail]; [e-mail]; [e-mail]
>> Cc: [e-mail]; Black, David <[e-mail]>
>> Subject: RE: 5245bis: STUN/TURN transaction timeout timer
>>  
>> That would be a fine thing to do, Thanks, --David
>>  
>> From: Christer Holmberg [mailto:[e-mail]] 
>> Sent: Monday, February 26, 2018 9:22 AM
>> To: Black, David <[e-mail]>; [e-mail]; [e-mail]; [e-mail]
>> Cc: [e-mail]
>> Subject: Re: 5245bis: STUN/TURN transaction timeout timer
>>  
>> Hi,
>>  
>> >> But, still, is there a reason we couldn’t use ‘Ti’ also in 5245bis, and point out the big value difference when used with ICE?
>> > 
>> >Given the nearly 2-orders-of-magnitude difference in the time periods, I’d be concerned that using the same name risks leaving an incorrect impression on an implementer who
>> >is familiar with one protocol, but new to the other.   Different names may also improve clarity in other documents that describe how STUN and ICE work together.
>>  
>> Fair enough. But, should we then point out that Ti isn’t used with ICE?
>>  
>> Regards,
>>  
>> Christer
>>  
>>  
>>  
>> From: Christer Holmberg [mailto:[e-mail]] 
>> Sent: Sunday, February 25, 2018 8:00 AM
>> To: [e-mail]; [e-mail]; [e-mail]
>> Cc: Black, David <[e-mail]>; [e-mail]
>> Subject: 5245bis: STUN/TURN transaction timeout timer
>>  
>> Hi,
>>  
>> In draft-5245bis, the name of the STUN/TURN transaction timeout timer is ‘HTO’.
>>  
>> As part of the IESG review, I have been asked what the ‘H’ stands for. After some digging in the mail archives (2016-09-14), I figured out it stands for “handshake”:
>>  
>> https://www.ietf.org/mail-archive/web/ice/current/msg00378.html
>>  
>>       “2.  A timeout for request packets, call it handshake timeout or HTO which SHOULD be 2*RTT if the RTT is known and 500ms otherwise.”
>>  
>> Now, in RFC 5389, the transaction timeout timer is called ‘Ti’. However, the default value for that SHOULD be 39,5 seconds – which is quite different from 500ms.
>>  
>> But, still, is there a reason we couldn’t use ‘Ti’ also in 5245bis, and point out the big value difference when used with ICE?
>>  
>> Regards,
>>  
>> Christer
>> _______________________________________________
>> Ice mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/ice
>