Re: [tcpm] 793bis: TCP Quiet Time

[Fernando Gont <[e-mail]>]

On 23/1/20 03:27, Joe Touch wrote:
> 
> 
>> On Jan 22, 2020, at 10:54 AM, Fernando Gont <[e-mail] 
>> <mailto:[e-mail]>> wrote:
>>
>>> On 8/28/2019 11:39 AM, Gorry Fairhurst wrote:
>>>> OLD, Section: "   The TCP Quiet Time Concept"
>>>> - Found this section quite amusing. Is this concept widely 
>>>> implemented in stacks?
>>
>> No.
> 
> It might be useful to consider this is relevant only for systems that 
> reboot in under 2 MSL anyway. That’s a good reason most systems don’t 
> need to care - but a good reason to worry about ignoring this issue as 
> we move forward.
> 
> I.e., of this concept and its utility, as financial investors advise:
> 
> Past performance is not be a predictor of future results.

I agree.

That said, in virtualized environments, and also in many (most?) modern 
envirnonments, you normally get to bootstrap in less than MSL.

As noted, I wouldn't remove the "quiet time concept", since it has to do 
with correctness of the spec (it is the only way to answer the question 
"how do you take care of stale segments from previous incarnations of a 
connection?").

That said, I find it quite unlikely that OSes try to implement the Quite 
Time Concept -- even for the cases where it would be required. 
Particularly if they can get away with it, for practical purposes, by 
means of TCP timestamps.

Thanks,
-- 
Fernando Gont
SI6 Networks
e-mail: [e-mail]
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492