Re: Compression ratio of HPACK.

As mentioned before, but probably not enough, much of the value for this
part is expected to show up in the future as people adapt to the
compression by reusing header fields, and/or by using push.

-=R
On Jan 30, 2014 7:29 AM, "Adrian Cole" <adrian.f.cole@gmail.com> wrote:

> Really interesting, Kazu.  I would love to cut out gears if we are only
> saving a byte and a half.  Can someone else who profiled hpack review this
> data and conclusion?
>
> -A
> On Jan 29, 2014 9:09 PM, "Kazu Yamamoto" <kazu@iij.ad.jp> wrote:
>
>> Hi,
>>
>> I implemented 6 kinds of HPACK 05 encoding and calculated compression
>> ratio with the help from the Japanese HTTP/2 community. I wrote a blog
>> article on it.
>>
>>         http://d.hatena.ne.jp/kazu-yamamoto/20140129/1391057824
>>
>> In short, reference set saves only 1.57 byte per HEADERS frame in
>> average. So, I would suggest the designers to remove reference set.
>>
>> If you guys want to calculate compression ratio with other data set,
>> please let me know.
>>
>> --Kazu
>>
>>