Re: Spin bit decision

[Lars Eggert <[e-mail]>]

Hi,

On 2018-10-2, at 9:05, [e-mail] wrote:
> After NYC, what are the remaining questions the WG is asking regarding the spin bit to educate the upcoming decision (Bangkok):
> 
>  (a) why is midpoint half-RTT measurement useful (in current practices) ?
>  (b) is the spin bit up to the job?
>  (c) isn't there any other way ?
>  (d) other
> 
> ?

the following is my personal feedback on this matter, i.e., I'm not speaking as chair:

I have little doubt that the single-bit spin bit variant is fit for purpose. I also think the DT has analyzed the privacy implications of the single-bit variant sufficiently for me to have no serious privacy concerns about the variant.

The key open question in my mind is this: on what fraction of the traffic would an operator need to see spinning and echoed bits in order to obtain useful network-wide or at least path specific latency information? In other words, how many browsers and servers-side deployments would need to opt in and spin or echo the bits? Would it be sufficient if one hyperscalar with a bowser and a server deployment opted in? Even if most of their traffic concentrated towards their peering links?

Regards,
Lars