- From: Mark Nottingham <mnot@mnot.net>
- Date: Fri, 3 Nov 2017 10:30:43 +1100
- To: Willy Tarreau <w@1wt.eu>, Ilya Grigorik <ilya@igvita.com>
- Cc: Kazuho Oku <kazuhooku@gmail.com>, Poul-Henning Kamp <phk@phk.freebsd.dk>, HTTP Working Group <ietf-http-wg@w3.org>
Huh. Ilya, has there ever been discussion of using a format like "4:3" for DPR, to be more accurate? Cheers, > On 3 Nov 2017, at 9:48 am, Willy Tarreau <w@1wt.eu> wrote: > > On Fri, Nov 03, 2017 at 09:30:08AM +1100, Mark Nottingham wrote: >> On 2 Nov 2017, at 4:54 pm, Willy Tarreau <w@1wt.eu> wrote: >>> >>>> For floats, first let me state that I do not know if we even need to >>>> support it. To me it seems worth considering to drop it, as PHK >>>> suggested. >>> >>> +1. We can define that "q=" is specified in thousands for example and >>> that only 3 decimals have to be parsed there. >> >> I'm not so sure we can do that; it's not just QValues. E.g., >> >> http://httpwg.org/http-extensions/client-hints.html#dpr > > Interesting one. With the proposed format there's no way to reproduce > the exact value (eg: 4000x3000 will send 1.3333 but after how many '3' > is it enough ?). Ratios could be useful there (eg: 4000/3000) but some > implementations will always suffer from div by zero or the funnier div > by -1 which nobody cares to check against and which produces the same > result. Maybe we should accept some limits to the precision here, such > as 16 bit integral value and 16 bit fractional one for all decimal > numbers (as an example of course). > > Willy > -- Mark Nottingham https://www.mnot.net/
Received on Thursday, 2 November 2017 23:31:12 UTC