Re: #578: getting real-ish numbers for option 3

Hi,

The option 3 implementation I emailed earlier had D bit accidentally flipped
(so D is 1 for static table).  It was now fixed in github repository.

Best regards,
Tatsuhiro Tsujikawa

On Fri, Oct 24, 2014 at 7:33 PM, Mark Nottingham <mnot@mnot.net> wrote:

> Toy up at:
>   https://gist.github.com/mnot/434ab029a6e878b2af4c
>
> Cheers,
>
>
> > On 24 Oct 2014, at 6:50 pm, Willy Tarreau <w@1wt.eu> wrote:
> >
> > On Fri, Oct 24, 2014 at 05:56:37PM +1100, Mark Nottingham wrote:
> >>> Thus I think that we should define 3 "models" to test in fact :
> >>> - the "average" one as you describe above
> >>> - the "browser" one with a single custom header out of the 10
> >>> - the "partner" one with 9 out of the 10 custom headers
> >>>
> >>> That way we can see if one model shows an important deviation using
> one or
> >>> another encoding. In my opinion, an adequate encoding (I mean a safe
> one
> >>> for the future) should be reasonably good on all cases and show limited
> >>> variations around the average model.
> >>>
> >>> Once we're able to synthetize the requests for a given model, it's easy
> >>> to build the two other ones, so  think it should be done.
> >>>
> >>> Opinions ?
> >>
> >> Sure, with the proviso that actually interpreting what's a useful
> difference
> >> is still undefined, and likely to cause some debate.
> >
> > Sure, but I think it's not a big issue, because :
> >  - if we find important differences, there should be a rought consensus
> >    for the best solution
> >  - if there's no noticeale gain, that means it's best not to change a
> iota.
> >
> > And at least I hear that the persons less interested in changing are in
> > this thinking so that should not be a problem.
> >
> >> But let's go ahead and try, since the cost is relatively low. I'll
> write some
> >> Python this weekend (possibly tonight, subject to family stuff) to
> generate
> >> some header sets; if other folks can do the crunching code and have it
> ready,
> >> that'd be much appreciated.
> >
> > Great. I unfortunately cannot build deflatehd, I'm still trying to figure
> > why it fails while the rest is OK.
> >
> >> Unless I hear otherwise, I'm going to do HTTP/1 style header sets
> separated
> >> by double-newlines; e.g.,
> >>
> >> :scheme: https
> >> :authority: foo.com
> >> :path: /abc
> >> foo: bar
> >>
> >> :scheme: http
> >> :authority: bar.com
> >> :path: /def
> >> baz: bat
> >>
> >> and so on...
> >
> > Looks fine, it's how I've been handling headers as well till now.
> >
> >> I'll put it in a repo for inspection / pulls. Whoever does the other
> code
> >> should as well.
> >
> > Does anyone have an encoder which can be easily extracted from his/her
> > implementation and be fed with many requests like this ? It would save
> > a significant amount of work.
> >
> > Cheers,
> > Willy
> >
>
> --
> Mark Nottingham   http://www.mnot.net/
>
>
>
>
>