Re: [quicwg/base-drafts] Remove ack_delay_exponent TP (#2670)

[MikkelFJ <[e-mail]>]

So I'm am one the few holding out for shift. I disagree with many of the arguments made on shift being imprecise, but I come to some of the same conclusion @kazuho has recently posted:

The encoded can choose its multiplier. It can choose one that it is a power of 2.

The native representation of time my require scaling anyway, so you might end up with a non-power of two. However, this is typically a constant, and division by a constant is multiplication if you do it compile time. It is a long multiplication which is slower, but not that much.

Therefore I am reverting on my stance and think that a multiplier is acceptable, as long as it can be chosen statically for the application, or it can choose a power of 2 dynamically.

More generally I still think it is a problem to guess the time scale and would like to have an adaptive scale without transport parameters, using 16 bit unsigned floats, but there is zero traction on that here.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/2670#issuecomment-509988126