RE: FYI: Tools to evaluate header compression algorithms

Hi,



I just sent a patch allowing to do that.



The following command line using twice http1 (the second time with a dummy parameter) will generate an output containing both runs.



$ python compare_compressors.py -c http1 -c http1=dummy ../http_samples/mnot/wikipedia.org.har

408 req messages processed

                   compressed | ratio min   max   std

req       http1        89,284 | 1.00  1.00  1.00  0.00

req http1=dummy        89,284 | 1.00  1.00  1.00  0.00



408 res messages processed

                   compressed | ratio min   max   std

res       http1        85,817 | 1.00  1.00  1.00  0.00

res http1=dummy        85,817 | 1.00  1.00  1.00  0.00

Hervé.

From: James M Snell [mailto:jasnell@gmail.com]
Sent: lundi 14 janvier 2013 15:59
To: Mark Nottingham
Cc: Martin Nilsson; ietf-http-wg@w3.org
Subject: Re: FYI: Tools to evaluate header compression algorithms


What would be helpful is the ability to define multiple compression runs per compressor subdirectory... But the current design works fine too. Pretty straightforward really.
On Jan 14, 2013 3:47 AM, "Mark Nottingham" <mnot@mnot.net<mailto:mnot@mnot.net>> wrote:
Perhaps, but they're not really easy to disentangle; if people are interested in the different combinations, they can easily factor the compressor code to cover the space (as we're already starting to see with James' work).

Cheers,

(of course, if you want to tackle this, a pull request would be welcome)


On 10/01/2013, at 11:48 PM, Martin Nilsson <nilsson@opera.com<mailto:nilsson@opera.com>> wrote:

> On Sun, 06 Jan 2013 06:57:01 +0100, Mark Nottingham <mnot@mnot.net<mailto:mnot@mnot.net>> wrote:
>
>> Quick follow-up:
>>
>> I posted more about this here:
>>  http://www.mnot.net/blog/2013/01/04/http2_header_compression

>>
>> In particular, we have graphs for all of the HAR samples I took earlier:
>>  http://http2.github.com/http_samples/mnot/

>>
>
> One minor issue is that you are comparing both encoding and compression at the same time. I.e. you could have a (http, spdy) x (uncompressed, delta, delta-huffman, gzip) result matrix.
>
> /Martin Nilsson
>
> --
> Using Opera's revolutionary email client: http://www.opera.com/mail/

>

--
Mark Nottingham   http://www.mnot.net/