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

[Kazuho Oku <[e-mail]>]

Doing a hum sounds good to me. Mp prefeence goes to arguing for dislikes as @ianswett says.

@martinthomson 
> As I understand the state of play, the main costs are bytes on the wire (potentially across a wide range of operating points: short RTT, fast processor to long RTT, slow processor), code complexity, and efficiency of calculating the values for encode (e.g., division vs. shift) or decode (e.g., multiply vs. shift).

FWIW, this is the table I have illustrating the differences of the encoder side. I do not think division is necessary even when adopt the multiplier proposal, because the encoder has the freedom to choose the multiplier. By choosing a multiplier that is 2-power of the timer granularity, a encoder can use a bit shift operation.

| design \ granularity | 1us | 2<sup>n</sup> us | 10<sup>n</sup> us |
|:---:|:---:|:---:|:---:|
|no scaling|noop|bit-shift|multiply|
|bit-shift|bit-shift|bit-shift|becomes inaccurate|
|multiplier|bit-shift|bit-shift|bit-shift|

For the decoder side, it's noop vs. shift vs. multiply.

-- 
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-509852239